METHOD TO ENABLE VEHICLES TO AVOID WEIGH STATIONS
Method for controlling weighing of a vehicle travelling on a road by means of a weigh station alongside the road includes determining weight of the vehicle using a processor on-board the vehicle and transmitting from the vehicle using a telematics device, the determined weight of the vehicle to the weigh station. Determining the weight of the vehicle may entail processing inertial property data from an inertial measurement unit (IMU) into an indication of the weight of the vehicle. The inertial property data from the IMU may be multiple sets of inertial property data obtained over time. The IMU may be calibrated, using the processor, based differential motion of the vehicle over a period of time as determined by a location positioning system. A Kalman filter may also be used. A vehicle transmitting its determined weight avoids stopping at the weigh station.
This application is a divisional of U.S. patent application Ser. No. 13/649,807 filed Oct. 11, 2012, which is a divisional application of U.S. patent application Ser. No. 12/028,956 filed Feb. 11, 2008, now abandoned, which is:
1. a CIP of U.S. patent application Ser. No. 11/082,739 filed Mar. 17, 2005, now U.S. Pat. No. 7,421,321, which is a CIP of U.S. patent application Ser. No. 10/701,361 filed Nov. 4, 2003, now U.S. Pat. No. 6,988,026, which is a CIP of U.S. patent application Ser. No. 10/638,743 filed Aug. 11, 2003, now U.S. Pat. No. 7,284,769;
2. a CIP of U.S. patent application Ser. No. 11/131,623 filed May 18, 2005, now U.S. Pat. No. 7,481,453, which is a CIP of U.S. patent application Ser. No. 10/638,743 filed Aug. 11, 2003, now U.S. Pat. No. 7,284,769;
3. a CIP of U.S. patent application Ser. No. 11/833,033 filed Aug. 2, 2007, now abandoned, which is a CIP of U.S. patent application Ser. No. 10/638,743 filed Aug. 11, 2003, now U.S. Pat. No. 7,284,769; and
4. a CIP of U.S. patent application Ser. No. 11/833,052 filed Aug. 2, 2007, now U.S. Pat. No. 8,060,282, which is a CIP of U.S. patent application Ser. No. 10/638,743 filed Aug. 11, 2003, now U.S. Pat. No. 7,284,769.
All of the above applications and patents, and any applications, publications and patents mentioned below, are incorporated by reference herein in their entirety and made a part hereof.
FIELD OF THE INVENTIONThe present invention relates to methods to enable vehicles to avoid stopping at weigh stations to be weighed and ensuring safe travel of a vehicle or safety of an occupant of the vehicle.
BACKGROUND OF THE INVENTIONIn the United States, taxes are often collected from trucking companies and truck owners based in part on weight of the truck. To determine taxes owed, weight/inspection stations are utilized on most highways, which are also useful to identify trucks exceeding legal weight limits for the road on which the weigh station is alongside. Signs direct trucks to pull into the stations to have their weight checked to ensure that the trucks are in compliance with federal and state weight regulations. These stations use static scales, which require that the truck pull onto the scale and stop while being weighed.
The weighing process at a weigh station can take a few minutes or longer during peak truck travel times. This delay can represent a significant cost to trucking operations, particularly in situations where “just-in-time” shipping is being utilized and delays can result in reduced revenues.
An example of a prior art electronic screening system is “PrePass” (www.prepass.com). As described in U.S. Pat. No. 6,980,093, PrePass is a system that allows participating transponder-equipped commercial vehicles to bypass designated weigh stations and other such facilities. A vehicle participating in the PrePass system is identified in a database proprietary to the PrePass system, as part of the pre-certification process conducted when the vehicle is registered in the system. The database contains weight information and “credential information” regarding the vehicle and correlates this information with a PrePass transponder ID number that corresponds to a transponder carried in the vehicle. As a vehicle approaches a PrePass-equipped weight/inspection station, it comes into the range of an Automatic Vehicle Identification (AVI) antenna, which communicates with the transponder to identify the transponder ID number, thereby giving the PrePass system access to the saved data for that vehicle. At the same time, the vehicle passes over a WIM scale, and the weight data obtained from the scale is also transmitted back to the PrePass system. This allows the PrePass system to verify that the vehicle should be able to bypass the inspection station. Assuming everything is verified, a signal is sent to the transponder causing it to issue an audible signal and “go” indication (e.g., a “green light”) directing the driver to pass the station without needing to stop.
As also described in the '093 patent, an e-screening system concept that complies with the architecture of the “CVISN” architecture prescribed by the Federal Motor Carrier Safety Administration is described in “Introductory Guide to CVISN”, section 2.7 The CVISN e-screening concept has many advantages because of its use of a standardized national database that is shared among the states with data and methods of exchange that are standardized according to CVISN architecture. While having many advantages when compared to PrePass, both of these suffer from some disadvantages. Mainline screening alone, based upon AVI, is largely ineffective because it cannot reach the vast majority of trucks that do not operate with a transponder. Mainline screening systems must send all the vehicles that do not have transponders into the weigh station. At many stations, queuing backups would not be alleviated until at least 30-50% of the mainline vehicles were bypassed. Currently, on the order of 1-2% of commercial vehicles carry transponders, so e-screening systems designed around mainline screening alone cannot be effective. Additionally, mainline (i.e., highway-based) WIM scales are inherently inaccurate because the trucks are operating at highway speeds when being weighed using the mainline WIM scale. Vehicle dynamics generated by bumps in the highway road surface and the path of the vehicle contribute to inaccuracies when using mainline WIM scales. As a result, even the transponder-equipped vehicles tend to be directed into the weigh/inspection station to be subjected to the more rigorous and time-consuming static weighing system and detailed inspection process, only to be found in compliance and redirected back to the highway after significant (and unnecessary) delay.
The '093 asserts that it is desirable to have an e-screening system that can conduct a secondary screening process, based upon AVI and alternative vehicle-identification technology after an initial (primary) screening process to reduce the number of vehicles that are subjected to the time-consuming static-scale weighing process improperly due to the inaccuracies inherent in mainline WIM scale measurements.
To this end, the '093 patent describes a system for electronically screening vehicles traveling on a road having an exit ramp along which is situated a vehicle weigh station, the vehicle weigh station having a static scale configured to make static weight measurements of the vehicles. The system includes a first weigh-in-motion (WIM) scale, positioned along the road in the proximity of the vehicle weigh station, configured to make a first weight measurement of the vehicles; a second WIM scale, positioned along the exit ramp and associated with the weigh station, configured to make a second weight measurement of the vehicles; one or more indicator signals, positioned near the first and second WIM scales so as to be perceivable by drivers of the vehicles, which, when activated, direct the traveling vehicles to either pull onto or bypass the exit ramp and/or the static scale; and a processor, coupled to the first and second WIM scales, the static scales and the one or more indicator signals. The processor is configured to (i) correct the first and second WIM weight measurements based on the static weight measurements; and (ii) activate the one or more indicator signals based on the corrected first and second weight measurements.
SUMMARY OF THE INVENTIONA method for controlling weighing of a vehicle travelling on a road by means of a weigh station alongside the road in accordance with the invention includes determining weight of the vehicle using a processor on-board the vehicle and transmitting from the vehicle using a telematics device, the determined weight of the vehicle to the weigh station. Determining the weight of the vehicle may entail processing inertial property data from an inertial measurement unit (IMU) into an indication of the weight of the vehicle. The inertial property data from the IMU may be multiple sets of inertial property data obtained over time. The IMU may be calibrated, using the processor, based differential motion of the vehicle over a period of time as determined by a location positioning system. A Kalman filter may also be used. A vehicle transmitting its determined weight avoids stopping at the weigh station.
Another method for controlling travel of vehicles on a road having a weigh station alongside the road at which vehicles must provide their vehicle weight at least for some designated times and designated vehicle types in accordance with the invention includes for those vehicles equipped with an on-board weight determining unit, determining weight of the vehicle using a processor on-board the vehicle and transmitting from the vehicle using a telematics device, the determined weight of the vehicle to the weigh station. A vehicle transmitting its determined weight avoids stopping at the weigh station. The on-board weight determining unit includes an inertial measurement unit (IMU) and the weight of the vehicle may then be determined by processing inertial property data from the IMU into an indication of the weight of the vehicle. The inertial property data from the IMU may be multiple sets of inertial property data obtained over time. The IMU may be calibrated, using the processor, based differential motion of the vehicle over a period of time as determined by a location positioning system. A Kalman filter may also be used.
A method for managing road information in accordance with the invention includes obtaining road properties from vehicles during travel on the vehicle on a road at an off-vehicle location, the road properties being obtained using an inertial measurement unit on board each vehicle and transmitted from the vehicle using a telematics device, storing the road properties from the vehicles at the off-vehicle locations in a data storage device, and selectively distributing from the data storage device to the telematics device on vehicles traveling on the road, at least part of the collected road properties to the vehicles traveling on the road. In one embodiment, the road properties are associated with the weather in the area of the road at the off-vehicle location, in which case, the collected road properties is distributed to the vehicles traveling on the road based on the weather.
The following drawings are illustrative of embodiments of the invention and are not meant to limit the scope of the invention as encompassed by the claims.
1.1 Pattern Recognition Approach to Crash Sensing
Throughout much of the discussion herein, the neural network will be used as an example of a pattern recognition technique or algorithm since the neural network is one of the most developed of such techniques. However, it has limitations that are now being addressed with the development of newer pattern recognition techniques as well as better neural network techniques such as combination or modular neural networks. These limitations involve the difficulty in describing the process used in classifying patterns with the result that there is a fear that a pattern that was not part of the training set might be missed. Also, the training process of the neural network does not guarantee that convergence to the best solution will result. One such example is the local minimum problem wherein the training algorithm converges on a result that is not the best overall or global solution. These problems are being solved with the development of newer pattern recognition techniques such as disclosed in various U.S. patents and technical papers. One invention disclosed herein is the use of pattern recognition techniques including neural networks, regardless of the particular technique, to provide a superior smart airbag system. In particular, genetic algorithms are being applied to aid in selecting the best of many possible choices for the neural network architecture. The use of genetic algorithms helps avoid the local minimum situation since several different architectures are tried and the best retained.
The pattern recognition algorithm, which forms an integral part of the crash sensor described herein, can be implemented either as an algorithm using a conventional microprocessor, FPGA or ASIC or through a neural computer. In the first case, the training is accomplished using a neural pattern recognition program and the result is a computer algorithm frequently written in the C computer language, although many other computer languages such as FORTRAN, assembly, Basic, etc. could be used. In the last case, the same neural computer can be used for the training as used on the vehicle. Neural network software for use on a conventional microcomputer is available from several sources such as International Scientific Research, Panama City, Panama. An example of a neural network-based crash sensor algorithm produced by ISR software after being trained on a crash library created by using data supplied by an automobile manufacturer for a particular model vehicle plus additional data created by using the techniques of crash and velocity scaling is:
Neural computers on a chip are now available from various chip suppliers. These chips make use of massively parallel architecture and allow all of the input data to be processed simultaneously. The result is that the computation time required for a pattern to be tested changes from the order of milliseconds for the case of the microprocessor-implemented system to the order of tens to hundreds of microseconds for the neural computer. With this computational speed, one neural computer can easily be used for several pattern recognition implementations simultaneously even during the crash event including dynamic out-of-position and crash sensing. A discussion of the structure of such a neural computer can be found on page 382 of Digital Neural Networks, by Kung, S. Y., PTR Prentice Hall, Englewood Cliffs, N.J., 1993.
As an example of an algorithm produced by such software after being trained on a crash library created by using data supplied by an automobile manufacturer for a particular model vehicle plus additional data created by using the techniques of crash and velocity scaling, the network was trained to give a value of 1 for triggering the airbag and 0 for not triggering. In the instant case, this value would depend on the type of gas control module that is used and in general would vary continuously from 0 to 1 with the particular value indicative of the action to be taken by the gas control module, such as adding more gas to the airbag.
Examples of neural networks in several forms will be discussed in more detail below in several sections of this application.
1.2 Electronic Crash Sensors
An airbag electronic sensor and diagnostic module (SDM) is typically mounted at a convenient location in the passenger compartment such as the transmission tunnel or firewall.
A tri-axial accelerometer is a device that includes three accelerometers and measures accelerations in three orthogonal directions that are typically the longitudinal, lateral and vertical directions, although there are sometimes reasons to use a different orientation. Such a different orientation can be useful to remove some of the bias errors in the accelerometers by, for example, allowing each accelerometer to be partially influenced by gravity. Also, in some applications, the tri-axial accelerometer is intentionally rotated relative to the vehicle to expose different accelerometers to gravity again for accuracy calibration purposes. An alternate method is to electronically test the acceleration sensing elements by exposing them to an electric field and measure their response. Such an accelerometer is called a “testable” accelerometer.
The circuit board of the SDM 55 also optionally contains a capacitor 61 as a backup power supply, other electronic components 58 and various circuitry. The SDM is connected to the airbags 51, 52 with wires 53 and 54 (shown in dotted lines in
For frontal impacts, for example, a signal is sent through wires 53 and 54 to initiate deployment of airbags 51 and 52 and to control the gas flow into and/or out of each airbag 51, 52 through the gas control modules (not shown) for each airbag. The ADC 57 is connected to the acceleration sensor, in this case the tri-axial accelerometer 56, and converts an analog signal generated by one or more of the accelerometers 56 representative of the acceleration thereof, and thus the vehicle, into a digital signal. In one embodiment, the ADC 57 may derive the digital signal from the integral of the analog signal. Naturally, many of the components of the printed circuit board can be incorporated into an ASIC as is obvious to those skilled in the art.
The tri-axial accelerometer and/or gyroscopes 56 (or IMU) are mounted by suitable mounting structure to the vehicle and can be mounted in a variety of positions to sense, e.g., frontal impacts, side impacts, rear impacts and/or rollovers. In another embodiment described below, the microprocessor 59 may include a detection system for detecting when the occupant to be protected by the deployable airbags 51, 52 is out-of-position and thereupon to suppress deployment thereof. Also, the detection system may be applied to detect the presence of a rear-facing child seat positioned on a passenger seat and thereupon to suppress deployment of the airbag. In each case, the microprocessor or neural computer 59 performs an analysis on signals received from appropriate sensors and corresponding ADCs. Recent advances in computational theory suggest that a form of computation using analog data rather than digital data may become viable. One example is the use of optical correlators for object detection and identification in the military where the optical signal from a video scene is converted to its Fourier transform using diffraction techniques.
The pattern recognition crash sensor described and illustrated in
A diagram of one example of a neural network used for a crash sensor designed based on the teachings of this invention is shown in
During the training phase, an output node value is assigned for every setting of the gas control module corresponding to the desired gas flow for that particular crash as it has occurred at a particular point in time. As the crash progresses and more acceleration values appear on the input nodes, the value of the output node may change. In this manner, as long as the crash is approximately represented in the training set, the gas flow can be varied at each one or two milliseconds depending on the system design to optimally match the quantity of gas in the airbag to the crash as it is occurring. Similarly, if an occupant sensor and a weight sensor are present, that information can additionally be fed into a set of input nodes so that the gas module can optimize the quantity of gas in the airbag taking into account both the crash deceleration and also the position, velocity, size and/or weight of the occupant to optimally deploy the airbag to minimize airbag induced injuries and maximize the protection to the occupant. Details of the manner in which a neural network process operates and is trained are known and will not be presented in detail here.
A time step, such as two milliseconds, is selected as the period in which the ADC pre-processes the output from the accelerometers and feeds data to input node 1. Thus, using this time step, at time equal to 2 milliseconds from the start of the process, node 1 contains a value obtained from the ADC and the remaining input nodes have a random value or a value of 0. At time equal 4 milliseconds, the value that was on node 1 is transferred to node 2 (or the node numbering scheme is advanced) and a new value from the ADC is fed into node 1. In a similar manner, data continues to be fed from the ADC to node 1 and the data on node 1 is transferred to node 2 whose previous value was transferred to node 3 etc. The actual transfer of data to different memory locations need not take place but only a redefinition of the location that the neural network should find the data for node 1. For one preferred embodiment of this invention, a total of one hundred input nodes were used representing two hundred milliseconds of acceleration data. At each step, the neural network is evaluated and if the value at the output node exceeds some value such as 0.5, then the airbags are deployed by the remainder of the electronic circuit. In this manner, the system does not need to know when the crash begins, that is, there is no need for a separate sensor to determine the start of the crash or of a particular algorithm operating on the acceleration data to make that determination.
In the example above, one hundred input nodes were used, along with twelve hidden layer nodes and one output layer node. Accelerations from only the longitudinal direction were considered. If other data such as accelerations from the vertical or lateral directions or the output from a number of gyroscopes were also used, then the number of input layer nodes would increase. If the neural network is to be used for sensing rear impacts, or side impacts, 2 or 3 output nodes might be used, one for each gas control module, headrest control module etc. Alternately, combination, modular or even separate neural networks can be used. The theory for determining the complexity of a neural network for a particular application is the subject of many technical papers and will not be presented in detail here. Determining the requisite complexity for the example presented herein can be accomplished by those skilled in the art of neural network design and is discussed briefly below. In another implementation, the integral of the acceleration data is used and it has been found that the number of input nodes can be significantly reduced in this manner.
The neural network described above defines a method of sensing a crash and determining whether to begin inflating a deployable occupant protection device, and at what rate, and comprises:
(a) obtaining one or more acceleration signals from one or more accelerometers mounted on a vehicle;
(b) converting the acceleration signal(s) into a digital time series which may include pre-processing of the data;
(c) entering the digital time series data into the input nodes of a neural network;
(d) performing a mathematical operation on the data from each of the input nodes and inputting the operated-on data into a second series of nodes wherein the operation performed on each of the input node data prior to inputting the operated on value to a second series node is different from that operation performed on some other input node data;
(e) combining the operated-on data from all of the input nodes into each second series node to form a value at each second series node;
(f) performing a mathematical operation on each of the values on the second series of nodes and inputting the operated-on data into an output series of nodes wherein the operation performed on each of the second series node data prior to inputting the operated on value to an output series node is different from that operation performed on some other second series node data;
(g) combining the operated on data from all of the second series nodes into each output series node to form a value at each output series node; and,
(h) initiating gas flow into an airbag if the value on one output series node is within a selected range signifying that a crash requiring the deployment of an airbag is underway; and
(i) causing the amount of gas flow into or out of the airbag to depend on the value on that one output series node.
The particular neural network described and illustrated above contains a single series of hidden layer nodes. In some network designs, more than one hidden layer is used although only rarely will more than two such layers appear. There are of course many other variations of the neural network architecture illustrated above which appear in the literature.
The implementation of neural networks can have at least two forms, an algorithm programmed on a digital microprocessor or in a neural computer. Neural computer chips are now available and neural computers can be incorporated into ASIC designs. As more advanced pattern recognition techniques are developed, specially designed chips can be expected to be developed for these techniques as well.
Airbags have traditionally been designed based on the assumption that 30 milliseconds of deployment time is available before the occupant, as represented by an unbelted dummy corresponding to the average male, has moved five inches. An occupant can be seriously injured or even killed by the deployment of the airbag if he or she is too close to the airbag when it deploys and in fact many people, particularly children and small adults, have now been killed in this manner. It is known that this is particularly serious when the occupant is leaning against the airbag when it deploys which corresponds to about 12 inches of motion for the average male occupant, and it is also known that he will be uninjured by the deploying airbag when he has moved less than 5 inches when the airbag is completely deployed. These dimensions are based on the dummy that represents the average male, the so-called 50% male dummy, sitting in the mid-seating position.
The threshold for significant injury is thus somewhere in between these two points and thus for the purposes of this table, two benchmarks have been selected as being approximations of the threshold of significant injury. These benchmarks are, based on the motion of an unrestrained occupant, (i) if the occupant has already moved 5 inches at the time that deployment is initiated, and (ii) if the occupant has moved 12 inches by the time that the airbag is fully deployed. Both benchmarks really mean that the occupant will be significantly interacting with the airbag as it is deploying. Other benchmarks could of course be used; however, it is believed that these two benchmarks are reasonable lacking a significant number of test results to demonstrate otherwise, at least for the 50% male dummy.
One additional feature, which results from the use of the neural network crash sensor of this invention, is that at the time the decision is made to deploy the airbag and even for as long afterward as the sensor is allowed to run, in the above example, 200 milliseconds of crash data is stored in the network input nodes. This provides a sort of “black box” which can be used later to accurately determine the severity of the crash as well as the position of the occupant at the time of the crash. If some intermediate occupant positions are desired, they could be stored on a separate non-volatile memory.
Above, the sensing of frontal impacts has been discussed using a neural network derived algorithm. A similar system can be derived for rear and side impacts especially if an anticipatory sensor is available as will be discussed below. An IMU located at a single location in a vehicle can do an excellent job of monitoring the motions of the vehicle that could lead to accidents including pre-crash braking, excessive yaw or pitching or roll which could lead to a rollover event. If the vehicle also has a GPS system, then the differential motion of the vehicle over a period of one second as measured by the GPS can be used to calibrate the IMU eliminating all significant errors. This is done using a Kalman filter. If a DGPS system is also available along with an accurate map, then the vehicle will also know its precise position within centimeters. This however is not necessary for calibrating and thereby significantly improving the accuracy of the IMU and thus the vehicle motion can be known approximately 100 times better than systems that do not use such a GPS-calibrated IMU. This greatly enhances the ability of vehicle systems to avoid skidding, rollover and other out-of-control situations that frequently lead to accidents, injuries and death. This combination of an inexpensive perhaps MEMS-based IMU with GPS and a Kalman filter has previously not been applied to a vehicle for safety and vehicle control purposes although the concept has been used with a DGPS system for farm tractors for precision farming.
With an accurate IMU, as mentioned above, the weight of a variably loaded vehicle can be determined and sent by telematics to a weigh station thereby eliminating the need for the vehicle to stop and be weighed. Referring to
Such an accurate IMU can also be used to determine the inertial properties of a variably loaded vehicle such as a truck or trailer. In this case, the IMU output can be analyzed by appropriate equations of a neural network, and with assumed statistical road properties plus perhaps some calibration for a particular vehicle, to give the center of mass of the vehicle as well as its load and moments of inertia. With this knowledge plus even a crude digital map, a driver can be forewarned that he might wish to slow down due to an upcoming curve. If telematics are added, then the road properties can be automatically accumulated at an appropriate off-vehicle location and the nature of the road under all weather conditions can be made available to trucks traveling the road to minimize the chance of accidents. This information plus the output of the IMU can significantly reduce truck accidents. The information can also be made available to passing automobiles to warn them of impending potential problems. Similarly, if a vehicle is not behaving appropriately based on the known road geometry, for example if the driver is wandering off the road, traveling at an excessive speed for conditions or generally driving in an unsafe manner, the off-vehicle site can be made aware of the fact and remedial action taken.
There are many ways to utilize one or more IMUs to improve vehicle safety and in particular to prevent rollovers, out-of-control skidding, jack-knifing etc. In a simple implementation, a single IMU is placed at an appropriate location such as the roof of a truck or trailer and used to monitor the motion over time of the truck or trailer. Based on the assumption that the road introduces certain statistically determinable disturbances into the vehicle, such monitoring over time can give a good idea of the mass of the vehicle, the load distribution and its moments of inertia. It can also give some idea as to the coefficient of friction on the tires against the roadway. If there is also one or more IMUs located on the vehicle axle or other appropriate location that moves with the wheels, then a driving function of disturbances to the vehicle can also be known leading to a very accurate determination of the parameters listed above especially if both a front and rear axle are so equipped. This need not be prohibitively expensive as IMUs are expected to break the $100 per unit level in the next few years.
As mentioned above, if accurate maps of information from other vehicles are available, the IMUs on the axles may not be necessary as the driving function would be available from such sources. Over the life of the vehicle, it would undoubtedly be driven empty and full to capacity so that if an adaptive neural network is available, the system can gradually be trained to quickly determine the vehicle's inertial properties when the load or load distribution is changed. It can also be trained to recognize some potentially dangerous situations such as loads that have become lost resulting in cargo that shifts during travel.
If GPS is not available, then a terrain map can also be used to provide some corrections to the IMU. By following the motion of the vehicle compared with the known geometry of the road, a crude deviation can be determined and used to correct IMU errors. For example, if the beginning and end of a stretch of a road is known and compared with the integrated output of the IMU, then corrections to the IMU can be made.
The MEMS gyroscopes used in a typical IMU are usually vibrating tuning forks or similar objects. Another technology developed by the Sciras Company of Anaheim, Calif., (The μSCIRAS multisensor, a Coriolis Vibratory Gyro and Accelerometer IMU) makes use of a vibrating accelerometer and shows promise of making a low cost gyroscope with improved accuracy. A preferred IMU is described in U.S. Pat. No. 4,711,125. One disclosed embodiment of a side impact crash sensor for a vehicle in accordance with the invention comprises a housing, a mass within the housing movable relative to the housing in response to accelerations of the housing, and structure responsive to the motion of the mass upon acceleration of the housing in excess of a predetermined threshold value for controlling an occupant protection apparatus. The housing is mounted by an appropriate mechanism in such a position and a direction as to sense an impact into a side of the vehicle. The sensor may be an electronic sensor arranged to generate a signal representative of the movement of the mass and optionally comprise a microprocessor and an algorithm for determining whether the movement over time of the mass as processed by the algorithm results in a calculated value that is in excess of the threshold value based on the signal. In the alternative, the mass may constitute part of an accelerometer, i.e., a micro-machined acceleration sensing mass. The accelerometer could include a piezo-electric element for generating a signal representative of the movement of the mass.
An embodiment of a side impact airbag system for a vehicle in accordance with an invention herein comprises an airbag housing defining an interior space, one or more inflatable airbags arranged in the interior space of the system housing such that when inflating, the airbag(s) is/are expelled from the airbag housing into the passenger compartment (along the side of the passenger compartment), and an inflator mechanism for inflating the airbag(s). The inflator mechanism may comprise an inflator housing containing propellant. The airbag system also includes a crash sensor for controlling inflation of the airbag(s) via the inflator mechanism upon a determination of a crash requiring inflation thereof, e.g., a crash into the side of the vehicle along which the airbag(s) is/are situated. The crash sensor may thus comprise a sensor housing arranged within the airbag housing, external of the airbag housing, proximate to the airbag housing and/or mounted on the airbag housing, and a sensing mass arranged in the sensor housing to move relative to the sensor housing in response to accelerations of the sensor housing resulting from, e.g., the crash into the side of the vehicle. Upon movement of the sensing mass in excess of a threshold value, the crash sensor controls the inflator to inflate the airbag(s). The threshold value may be the maximum motion of the sensing mass required to determine that a crash requiring deployment of the airbag(s) is taking place.
The crash sensor of this embodiment, or as a separate sensor of another embodiment, may be an electronic sensor and the movement of the sensing mass may be monitored. The electronic sensor generates a signal representative of the movement of the sensing mass that may be monitored and recorded over time. The electronic sensor may also include a microprocessor and an algorithm for determining whether the movement over time of the sensing mass as processed by the algorithm results in a calculated value that is in excess of the threshold value based on the signal.
In some embodiments, the crash sensor also includes an accelerometer, the sensing mass constituting part of the accelerometer. For example, the sensing mass may be a micro-machined acceleration sensing mass in which case, the electronic sensor includes a micro-processor for determining whether the movement of the sensing mass over time results in an algorithmic determined value which is in excess of the threshold value based on the signal. In the alternative, the accelerometer includes a piezo-electric element for generating a signal representative of the movement of the sensing mass, in which case, the electronic sensor includes a micro-processor for determining whether the movement of the sensing mass over time results in an algorithmic determined value which is in excess of the threshold value based on the signal.
1.3 Crash Severity Prediction
In the particular implementation described above, the neural network could be trained using crash data from approximately 25 crash and non-crash events. In addition, the techniques of velocity and crash scaling were used to create a large library of crashes representing many events not staged by the automobile manufacturer. The resulting library, it is believed, represents the vast majority of crash events that occur in real world accidents for the majority of automobiles. Thus, the neural network algorithm comes close to the goal of a universal electronic sensor usable on most if not all automobiles as further described in U.S. Pat. No. 5,684,701. The results of this algorithm as reported in the '701 patent for a matrix of crashes created by the velocity and crash scaling technique appears in FIGS. 7 and 8 of that patent.
The '701 patent describes the dramatic improvement achievable through the use of pattern recognition techniques for determining whether the airbag should be deployed. Such a determination is really a forecasting that the eventual velocity change of the vehicle will be above an amount, such as about 12 mph, which requires airbag deployment. The instant invention extends this concept to indirectly predict what the eventual velocity change will in fact be when the occupant, represented by an unrestrained mass, impacts the airbag. Furthermore, it does so not just at the time that the deployment decision is required but also, in the preferred implementation, at all later times until adding or removing additional gas from the airbag will have no significant injury reducing effect. The neural network can be trained to predict or extrapolate this velocity but even that is not entirely sufficient. What is needed is to determine the flow rate of gas into and/or out of the airbag to optimize injury reduction which depends not only on the prediction or extrapolation of the velocity change at a particular point in time but must take into account the prediction that was made at an earlier point when the decision was made to inject a given amount of gas into the airbag. Also, the timing of when the velocity change will occur is a necessary parameter since gas is usually not only flowing into but out of the airbag and both flows must be taken into account. It is thus unlikely that an algorithm, which will perform well in all real world crashes, can be mathematically derived.
The neural network solves the problem by considering all of the acceleration up to the current point in the crash and therefore knows how much gas has been put into the airbag and how much has flowed out. It can be seen that even if this problem could be solved mathematically for all crashes, the mathematical approach becomes hopeless as soon as the occupant properties are added.
Once a pattern recognition computer system is implemented in a vehicle, the same system can be used for many other pattern recognition functions such as the airbag system diagnostic. Testing that the pattern of the airbag system during the diagnostic test on vehicle startup, as represented by the proper resistances appearing across the wires to the various system components, for example, is an easy task for a pattern recognition system. The system can thus do all of the functions of the conventional SDM, sensing and diagnostics, as well as many others.
2. DiagnosticsA smart airbag system is really part of a general vehicle diagnostic system and many of the components that make up the airbag system and the rest of the vehicle diagnostic system can be shared. Therefore, we will now briefly discuss a general vehicle diagnostic system focusing on the interaction with the occupant restraint system. This description is taken from U.S. Pat. No. 6,484,080.
For the purposes herein the following terms are defined as follows:
The term “component” refers to any part or assembly of parts that is mounted to or a part of a motor vehicle and which is capable of emitting a signal representative of its operating state that can be sensed by any appropriate sensor. The following is a partial list of general automobile and truck components, the list not being exclusive:
Occupant restraints; engine; transmission; brakes and associated brake assembly; tires; wheel; steering wheel and steering column assembly; water pump; alternator; shock absorber; wheel mounting assembly; radiator; battery; oil pump; fuel pump; air conditioner compressor; differential gear; exhaust system; fan belts; engine valves; steering assembly; vehicle suspension including shock absorbers; vehicle wiring system; and engine cooling fan assembly.
The term “sensor” as used herein will generally refer to any measuring, detecting or sensing device mounted on a vehicle or any of its components including new sensors mounted in conjunction with the diagnostic module in accordance with the invention. A partial, non-exhaustive list of common sensors mounted on an automobile or truck is:
airbag crash sensor; accelerometer; microphone; camera; antenna; capacitance sensor or other electromagnetic wave sensor; stress or strain sensor; pressure sensor; weight sensor; magnetic field or flux sensor; coolant thermometer; oil pressure sensor; oil level sensor; air flow meter; voltmeter; ammeter; humidity sensor; engine knock sensor; oil turbidity sensor; throttle position sensor; steering wheel torque sensor; wheel speed sensor; tachometer; speedometer; other velocity sensors; other position or displacement sensors; oxygen sensor; yaw, pitch and roll angular sensors; clock; odometer; power steering pressure sensor; pollution sensor; fuel gauge; cabin thermometer; transmission fluid level sensor; gyroscopes or other angular rate sensors including yaw, pitch and roll rate sensors; coolant level sensor; transmission fluid turbidity sensor; break pressure sensor; tire pressure sensor; tire temperature sensor; tire acceleration sensor; GPS receiver; DGPS receiver; coolant pressure sensor; occupant position sensor; and occupant weight sensor.
The term “actuator” as used herein will generally refer to a device that performs some action upon receiving the proper signal. Examples of actuators include:
window motor; door opening and closing motor; electric door lock; deck lid lock; airbag inflator initiator; fuel injector; brake valves; pumps; relays; and steering assist devices.
The term “signal” as used herein will generally refer to any time varying output from a component including electrical, acoustic, thermal, or electromagnetic radiation, or mechanical vibration.
Sensors on a vehicle are generally designed to measure particular parameters of particular vehicle components. However, frequently these sensors also measure outputs from other vehicle components. For example, electronic airbag crash sensors currently in use contain an accelerometer for determining the accelerations of the vehicle structure so that the associated electronic circuitry of the airbag crash sensor can determine whether a vehicle is experiencing a crash of sufficient magnitude so as to require deployment of the airbag.
An IMU using up to three accelerometers and up to three gyroscopes can also be used. This accelerometer continuously monitors the vibrations in the vehicle structure regardless of the source of these vibrations. If a wheel is out-of-balance or delaminating, or if there is extensive wear of the parts of the front wheel mounting assembly, or wear in the shock absorbers, the resulting abnormal vibrations or accelerations can, in many cases, be sensed by the crash sensor accelerometer. There are other cases, however, where the sensitivity or location of the airbag crash sensor accelerometer is not appropriate and one or more additional accelerometers and/or gyroscopes or IMU may be mounted onto a vehicle for the purposes of this invention. Some airbag crash sensors are not sufficiently sensitive accelerometers or have sufficient dynamic range for the purposes herein.
Every component of a vehicle emits various signals during its life. These signals can take the form of electromagnetic radiation, acoustic radiation, thermal radiation, electric or magnetic field variations, vibrations transmitted through the vehicle structure, and voltage or current fluctuations, depending on the particular component. When a component is functioning normally, it may not emit a perceptible signal. In that case, the normal signal is no signal, i.e., the absence of a signal. In most cases, a component will emit signals that change over its life and it is these changes that contain information as to the state of the component, e.g., whether failure of the component is impending. Usually components do not fail without warning. However, most such warnings are either not perceived or if perceived are not understood by the vehicle operator until the component actually fails and, in some cases, a breakdown of the vehicle occurs. In a few years, it is expected that various roadways will have systems for automatically guiding vehicles operating thereon. Such systems have been called “smart highways” and are part of the field of intelligent transportation systems (ITS). If a vehicle operating on such a smart highway were to breakdown, serious disruption of the system could result and the safety of other users of the smart highway could be endangered.
Accelerometers are routinely used mounted outside of the crush zone for sensing the failure of the vehicle, that is, a crash of the vehicle. Looking at this in general terms, there is synergy between the requirements of sensing the status of the whole vehicle as well as its components and the same sensors can often be used for multiple purposes. The output of a microphone mounted in the vehicle could be used to help determine the existence and severity of a crash, for example.
In accordance with the invention, each of these signals emitted by the vehicle components is converted into electrical signals and then digitized (i.e., the analog signal is converted into a digital signal) to create numerical time series data that is then entered into a processor. Pattern recognition algorithms are then applied in the processor to attempt to identify and classify patterns in this time series data. For a particular component, such as a tire for example, the algorithm attempts to determine from the relevant digital data whether the tire is functioning properly and/or whether it requires balancing, additional air, or perhaps replacement. Future systems may bypass the A/D conversion and operate directly on the analog signals. Optical correlation systems are now used by the military that create the Fourier transform of an image directly using diffraction gratings and compare the image with a stored image.
Frequently, the data entered into the computer needs to be pre-processed before being analyzed by a pattern recognition algorithm. The data from a wheel speed sensor, for example, might be used as is for determining whether a particular tire is operating abnormally in the event it is unbalanced, whereas the integral of the wheel speed data over a long time period (integration being a pre-processing step), when compared to such sensors on different wheels, might be more useful in determining whether a particular tire is going flat and therefore needs air.
In some cases, the frequencies present in a set of data are a better predictor of component failures than the data itself. For example, when a motor begins to fail due to worn bearings, certain characteristic frequencies began to appear. In most cases, the vibrations arising from rotating components, such as the engine, will be normalized based on the rotational frequency as disclosed in a recent NASA TSP. Moreover, the identification of which component is causing vibrations present in the vehicle structure can frequently be accomplished through a frequency analysis of the data. For these cases, a Fourier transformation of the data is made prior to entry of the data into a pattern recognition algorithm. Optical correlations systems using Fourier transforms can also be applicable.
Other mathematical transformations are also made for particular pattern recognition purposes in practicing the teachings of this invention. Some of these include shifting and combining data to determine phase changes for example, differentiating the data, filtering the data, and sampling the data. Also, there exist certain more sophisticated mathematical operations that attempt to extract or highlight specific features of the data. This invention contemplates the use of a variety of these preprocessing techniques, and combinations thereof, and the choice of which one or ones is left to the skill of the practitioner designing a particular diagnostic module.
Another technique that is contemplated for some implementations of this invention is the use of multiple accelerometers and/or microphones that allow the system to locate the source of any measured vibrations based on the time of flight, or time of arrival of a signal at different locations, and/or triangulation techniques. Once a distributed accelerometer installation has been implemented to permit this source location, the same sensors can be used for smarter crash sensing as it will permit the determination of the location of the impact on the vehicle. Once the impact location is known, a highly tailored algorithm can be used to accurately forecast the crash severity making use of knowledge of the force vs. crush properties of the vehicle at the impact location.
When a vehicle component begins to change its operating behavior, it is not always apparent from the particular sensors, if any, which are monitoring that component. Output from any one of these sensors can be normal even though the component is failing. By analyzing the output of a variety of sensors, however, the pending failure can be diagnosed. For example, the rate of temperature rise in the vehicle coolant, if it were monitored, might appear normal unless it were known that the vehicle was idling and not traveling down a highway at a high speed. Even the level of coolant temperature which is in the normal range could in fact be abnormal in some situations signifying a failing coolant pump, for example, but not detectable from the coolant thermometer alone.
Pending failure of some components is difficult to diagnose and sometimes the design of the component requires modification so that the diagnosis can be more readily made. A fan belt, for example, frequently begins failing by a cracking of the inner surface. The belt can be designed to provide a sonic or electrical signal when this cracking begins in a variety of ways. Similarly, coolant hoses can be designed with an intentional weak spot where failure will occur first in a controlled manner that can also cause a whistle sound as a small amount of steam exits from the hose. This whistle sound can then be sensed by a general purpose microphone, for example.
In
All of the sensors illustrated on
As shown in
Important to this invention is the manner in which the diagnostic module 266 determines a normal pattern from an abnormal pattern and the manner in which it decides what data to use from the vast amount of data available. This is accomplished using pattern recognition technologies such as artificial neural networks and training. The theory of neural networks including many examples can be found in books on the subject. The neural network pattern recognition technology is one of the most developed of pattern recognition technologies. The neural network will be used here to illustrate one example of a pattern recognition technology but it is emphasized that this invention is not limited to neural networks. Rather, the invention may apply any known pattern recognition technology including sensor fusion and various correlation technologies. A brief description of the neural network pattern recognition technology is set forth below.
Neural networks are constructed of processing elements known as neurons that are interconnected using information channels call interconnects. Each neuron can have multiple inputs but generally only one output. Each output however is connected to all other neurons in the next layer. Neurons in the first layer operate collectively on the input data as described in more detail below. Neural networks learn by extracting relational information from the data and the desired output. Neural networks have been applied to a wide variety of pattern recognition problems including automobile occupant sensing, speech recognition, optical character recognition, and handwriting analysis.
To train a neural network, data is provided in the form of one or more time series that represents the condition to be diagnosed as well as normal operation. As an example, the simple case of an out-of-balance tire will be used. Various sensors on the vehicle can be used to extract information from signals emitted by the tire such as an accelerometer, a torque sensor on the steering wheel, the pressure output of the power steering system, a tire pressure monitor or tire temperature monitor. Other sensors that might not have an obvious relationship to an unbalanced tire are also included such as, for example, the vehicle speed or wheel speed. Data is taken from a variety of vehicles where the tires were accurately balanced under a variety of operating conditions also for cases where varying amounts of unbalance was intentionally introduced. Once the data has been collected, some degree of preprocessing or feature extraction is usually performed to reduce the total amount of data fed to the neural network. In the case of the unbalanced tire, the time period between data points might be chosen such that there are at least ten data points per revolution of the wheel. For some other application, the time period might be one minute or one millisecond.
Once the data has been collected, it is processed by a neural network-generating program, for example, if a neural network pattern recognition system is to be used. Such programs are available commercially, e.g., from NeuralWare of Pittsburgh, Pa. The program proceeds in a trial and error manner until it successfully associates the various patterns representative of abnormal behavior, an unbalanced tire, with that condition. The resulting neural network can be tested to determine if some of the input data from some of the sensors, for example, can be eliminated. In this way, the engineer can determine what sensor data is relevant to a particular diagnostic problem. The program then generates an algorithm that is programmed onto a microprocessor, microcontroller, neural processor, or DSP (herein collectively referred to as a microprocessor or processor). Such a microprocessor appears inside the diagnostic module 266 in
Once trained, the neural network, as represented by the algorithm, will now operationally recognize an unbalanced tire on a vehicle when this event occurs. At that time, when the tire is unbalanced, the diagnostic module 266 will output a signal indicative of the unbalanced tire, such as a signal to be sent to an output device which provides a message to the driver indicating that the tire should be now be balanced as described in more detail below. The message to the driver is provided by an output device coupled to or incorporated within the module 266 and may be, e.g., a light on the dashboard, a vocal tone or any other recognizable indication apparatus. Messages can also be transmitter to others outside of the vehicle such as other vehicles or to a vehicle dealer. In some cases, control of the vehicle may be taken over by a vehicle system in response to a message. In some cases, the vehicle component failure portends an oncoming accident and one or more parts of the restraint system can be deployed.
It is important to note that there may be many neural networks involved in a total vehicle diagnostic system. These can be organized either in parallel, series, as an ensemble, cellular neural network or as a modular neural network system. In one implementation of a modular neural network, a primary neural network identifies that there is an abnormality and tries to identify the likely source. Once a choice has been made as to the likely source of the abnormality, another of a group of neural networks is called upon to determine the exact cause of the abnormality. In this manner, the neural networks are arranged in a tree pattern with each neural network trained to perform a particular pattern recognition task.
Operation of a neural network is well understood by those skilled in the art. Neural networks are the most well-known of the pattern recognition technologies based on training, although neural networks have only recently received widespread attention and have been applied to only very limited and specialized problems in motor vehicles. Other non-training based pattern recognition technologies exist, such as fuzzy logic. However, the programming required to use fuzzy logic, where the patterns must be determined by the programmer, render these systems impractical for general vehicle diagnostic problems such as described herein. Therefore, preferably the pattern recognition systems that learn by training are used herein. On the other hand, the combination of neural networks and fuzzy logic, such as in a Neural-Fuzzy system, are applicable and can result in superior results.
The neural network is the first highly successful of what will be a variety of pattern recognition techniques based on training. There is nothing that suggests that it is the only or even the best technology. The characteristics of all of these technologies which render them applicable to this general diagnostic problem include the use of time-based input data and that they are trainable. In all cases, the pattern recognition technology learns from examples of data characteristic of normal and abnormal component operation.
A diagram of one example of a neural network used for diagnosing an unbalanced tire, for example, based on the teachings of this invention is shown in
Each of the input nodes is connected to each of the second layer nodes, h-1, h-2, . . . , h-n, called the hidden layer, either electrically as in the case of a neural computer, or through mathematical functions containing multiplying coefficients called weights. At each hidden layer node, a summation occurs of the values from each of the input layer nodes, which have been operated on by functions containing the weights, to create a node value. Similarly, the hidden layer nodes are in like manner connected to the output layer node(s), which in this example is only a single node O representing the decision to notify the driver of the unbalanced tire. During the training phase, an output node value of 1, for example, is assigned to indicate that the driver should be notified and a value of 0 is assigned to not providing an indication to the driver.
In the example above, twenty input nodes were used, five hidden layer nodes and one output layer node. In this example, only one sensor was considered and accelerations from only one direction were used. If other data from other sensors such as accelerations from the vertical or lateral directions were also used, then the number of input layer nodes would increase. Again, the theory for determining the complexity of a neural network for a particular application has been the subject of many technical papers and will not be presented in detail here. Determining the requisite complexity for the example presented here can be accomplished by those skilled in the art of neural network design. For an example of the use of a neural network crash sensor algorithm, see U.S. Pat. No. 5,684,701. Note that the inventors of this invention contemplate all combinations of the teachings of the '701 patent and those disclosed herein.
It is also possible to apply modular neural networks in accordance with the invention wherein several neural network are trained, each having a specific function relating to the detection of the abnormality in the operation of the component. The particular neural network(s) used, i.e., those to which input is provided or from which output is used, can be determined based on the measurements by one or more of the sensors.
Briefly, the neural network described above defines a method, using a pattern recognition system, of sensing an unbalanced tire and determining whether to notify the driver and comprises:
(a) obtaining an acceleration signal from an accelerometer mounted on a vehicle;
(b) converting the acceleration signal into a digital time series;
(c) entering the digital time series data into the input nodes of the neural network;
(d) performing a mathematical operation on the data from each of the input nodes and inputting the operated on data into a second series of nodes wherein the operation performed on each of the input node data prior to inputting the operated on value to a second series node is different from (e.g. may employ a different weight) that operation performed on some other input node data;
(e) combining the operated on data from all of the input nodes into each second series node to form a value at each second series node;
(f) performing a mathematical operation on each of the values on the second series of nodes and inputting this operated on data into an output series of nodes wherein the operation performed on each of the second series node data prior to inputting the operated on value to an output series node is different from that operation performed on some other second series node data;
(g) combining the operated on data from all of the second series nodes into each output series node to form a value at each output series node; and
(h) notifying a driver or taking some other action if the value on one output series node is within a selected range signifying that a tire requires balancing.
This method can be generalized to a method of predicting that a component of a vehicle will fail comprising:
(a) sensing a signal emitted from the component;
(b) converting the sensed signal into a digital time series;
(c) entering the digital time series data into a pattern recognition algorithm;
(d) executing the pattern recognition algorithm to determine if there exists within the digital time series data a pattern characteristic of abnormal operation of the component; and
(e) notifying a driver or taking some other action, including, in some cases, deployment of an occupant restraint system, if the abnormal pattern is recognized.
The particular neural network described above contains a single series of hidden layer nodes. In some network designs, more than one hidden layer is used, although only rarely will more than two such layers appear. There are of course many other variations of the neural network architecture illustrated above which appear in the referenced literature. For the purposes herein, therefore, “neural network” will be defined as a system wherein the data to be processed is separated into discrete values which are then operated on and combined in at least a two-stage process and where the operation performed on the data at each stage is, in general, different for each discrete value and where the operation performed is at least determined through a training process.
Implementation of neural networks can take on at least two forms, an algorithm programmed on a digital microprocessor, DSP or in a neural computer. In this regard, it is noted that neural computer chips are now becoming available.
In the example above, only a single component failure was discussed using only a single sensor since the data from the single sensor contains a pattern which the neural network was trained to recognize as either normal operation of the component or abnormal operation of the component. The diagnostic module 266 contains preprocessing and neural network algorithms for a number of component failures. The neural network algorithms are generally relatively simple, requiring only a few hundred lines of computer code. A single general neural network program can be used for multiple pattern recognition cases by specifying different coefficients for various terms, one set for each application. Thus, adding different diagnostic checks has only a small affect on the cost of the system. Also, the system has available to it all of the information available on the data bus. During the training process, the pattern recognition program sorts out from the available vehicle data on the data bus or from other sources, those patterns that predict failure of a particular component. Sometimes more than one data bus is used. For example, in some cases, there is a general data bus and one reserved for safety systems. Any number of data buses can of course be monitored.
In
Sensor 1 is a crash sensor having an accelerometer (alternately one or more dedicated accelerometers can be used), sensor 2 is represents one or more microphones, sensor 3 is a coolant thermometer, sensor 4 is an oil pressure sensor, sensor 5 is an oil level sensor, sensor 6 is an air flow meter, sensor 7 is a voltmeter, sensor 8 is an ammeter, sensor 9 is a humidity sensor, sensor 10 is an engine knock sensor, sensor 11 is an oil turbidity sensor, sensor 12 is a throttle position sensor, sensor 13 is a steering torque sensor, sensor 14 is a wheel speed sensor, sensor 15 is a tachometer, sensor 16 is a speedometer, sensor 17 is an oxygen sensor, sensor 18 is a pitch/roll sensor, sensor 19 is a clock, sensor 20 is an odometer, sensor 21 is a power steering pressure sensor, sensor 22 is a pollution sensor, sensor 23 is a fuel gauge, sensor 24 is a cabin thermometer, sensor 25 is a transmission fluid level sensor, sensor 26 is a yaw sensor, sensor 27 is a coolant level sensor, sensor 28 is a transmission fluid turbidity sensor, sensor 29 is brake pressure sensor and sensor 30 is a coolant pressure sensor. Other possible sensors include a temperature transducer, a pressure transducer, a liquid level sensor, a flow meter, a position sensor, a velocity sensor, a RPM sensor, a chemical sensor and an angle sensor, angular rate sensor or gyroscope.
If a distributed group of acceleration sensors or accelerometers are used to permit a determination of the location of a vibration source, the same group can, in some cases, also be used to measure the pitch, yaw and/or roll of the vehicle eliminating the need for dedicated angular rate sensors. In addition, such a suite of sensors can also be used to determine the location and severity of a vehicle crash and additionally to determine that the vehicle is on the verge of rolling over. Thus, the same suite of accelerometers optimally performs a variety of functions including inertial navigation, crash sensing, vehicle diagnostics, roll over sensing etc.
Consider now some examples. The following is a partial list of potential component failures and the sensors from the list on
Several interesting facts can be deduced from a review of the above list. First, all of the failure modes listed can be at least partially sensed by multiple sensors. In many cases, some of the sensors merely add information to aid in the interpretation of signals received from other sensors. In today's automobile, there are few if any cases where multiple sensors are used to diagnose or predict a problem. In fact, there is virtually no failure prediction undertaken at all. Second, many of the failure modes listed require information from more than one sensor. Third, information for many of the failure modes listed cannot be obtained by observing one data point in time as is now done by most vehicle sensors. Usually, an analysis of the variation in a parameter as a function of time is necessary. In fact, the association of data with time to create a temporal pattern for use in diagnosing component failures in automobile is believed to be unique to this invention as is the combination of several such temporal patterns. Fourth, the vibration measuring capability of the airbag crash sensor, or other accelerometer, is useful for most of the cases discussed above yet, at the time of this invention, there was no such use of accelerometers except as non-crush zone mounted crash sensors. The airbag crash sensor is used only to detect crashes of the vehicle. Fifth, the second most-used sensor in the above list, a microphone, does not currently appear on any automobiles yet sound is the signal most often used by vehicle operators and mechanics to diagnose vehicle problems. Another sensor that is listed above which also did not currently appear on automobiles at the time of this invention is a pollution sensor. This is typically a chemical sensor mounted in the exhaust system for detecting emissions from the vehicle. It is expected that this and other chemical sensors will be used more in the future.
In addition, from the foregoing depiction of different sensors which receive signals from a plurality of components, it is possible for a single sensor to receive and output signals from a plurality of components which are then analyzed by the processor to determine if any one of the components for which the received signals were obtained by that sensor is operating in an abnormal state. Likewise, it is also possible to provide for a multiplicity of sensors each receiving a different signal related to a specific component which are then analyzed by the processor to determine if that component is operating in an abnormal state. Note that neural networks can simultaneously analyze data from multiple sensors of the same type or different types.
The discussion above has centered on notifying the vehicle operator of a pending problem with a vehicle component. Today, there is great competition in the automobile marketplace and the manufacturers and dealers who are most responsive to customers are likely to benefit by increased sales both from repeat purchasers and new customers. The diagnostic module disclosed herein benefits the dealer by making him instantly aware, through the cellular telephone system, or other communication link, coupled to the diagnostic module or system in accordance with the invention, when a component is likely to fail.
As envisioned, on some automobiles, when the diagnostic module 266 detects a potential failure, it not only notifies the driver through a display 278, but also automatically notifies the dealer through a vehicle cellular phone 279. The dealer can thus contact the vehicle owner and schedule an appointment to undertake the necessary repair at each party's mutual convenience. The customer is pleased since a potential vehicle breakdown has been avoided and the dealer is pleased since he is likely to perform the repair work. The vehicle manufacturer also benefits by early and accurate statistics on the failure rate of vehicle components. This early warning system can reduce the cost of a potential recall for components having design defects. It could even have saved lives if such a system had been in place during the Firestone tire failure problem. The vehicle manufacturer will thus be guided toward producing higher quality vehicles thus improving his competitiveness. Finally, experience with this system will actually lead to a reduction in the number of sensors on the vehicle since only those sensors that are successful in predicting failures will be necessary.
For most cases, it is sufficient to notify a driver that a component is about to fail through a warning display. In some critical cases, action beyond warning the driver may be required. If, for example, the diagnostic module detected that the alternator was beginning to fail, in addition to warning the driver of this eventuality, the module could send a signal to another vehicle system to turn off all non-essential devices which use electricity thereby conserving electrical energy and maximizing the time and distance that the vehicle can travel before exhausting the energy in the battery. Additionally, this system can be coupled to a system such as ONSTAR® or a vehicle route guidance system, and the driver can be guided to the nearest open repair facility or a facility of his or her choice.
The diagnostic module of this invention assumes that a vehicle data bus exists which is used by all of the relevant sensors on the vehicle. Most vehicles manufactured at the time of this invention did not have a data bus although it was widely believed that most vehicles will have one in the near future. A vehicle safety bus has been considered for several vehicle models. Relevant signals can be transmitted to the diagnostic module through a variety of coupling systems other than through a data bus and this invention is not limited to vehicles having a data bus. For example, the data can be sent wirelessly to the diagnostic module using the Bluetooth or WiFi specification. In some cases, even the sensors do not have to be wired and can obtain their power via RF from the interrogator as is well known in the RFID (radio frequency identification) field. Alternately, an inductive or capacitive power transfer system can be used.
As can be appreciated, the invention described herein brings several new improvements to automobiles including, but not limited to, use of pattern recognition technologies to diagnose potential vehicle component failures, use of trainable systems thereby eliminating the need of complex and extensive programming, simultaneous use of multiple sensors to monitor a particular component, use of a single sensor to monitor the operation of many vehicle components, monitoring of vehicle components which have no dedicated sensors, and notification to the driver and possibly an outside entity of a potential component failure in time so that the failure can be averted and vehicle breakdowns substantially eliminated. Additionally, improvements to the vehicle stability, crash avoidance, crash anticipation and occupant protection are available.
To implement a component diagnostic system for diagnosing the component utilizing a plurality of sensors not directly associated with the component, i.e., independent of the component, a series of tests are conducted. For each test, the signals received from the sensors are input into a pattern recognition training algorithm with an indication of whether the component is operating normally or abnormally (the component being intentionally altered to provide for abnormal operation). Data from the test is used to generate the pattern recognition algorithm, e.g., a neural network, so that in use, the data from the sensors is input into the algorithm and the algorithm provides an indication of abnormal or normal operation of the component. Also, to provide a more versatile diagnostic module for use in conjunction with diagnosing abnormal operation of multiple components, tests may be conducted in which each component is operated abnormally while the other components are operating normally, as well as tests in which two or more components are operating abnormally. In this manner, the diagnostic module may be able to determine based on one set of signals from the sensors during use that either a single component or multiple components are operating abnormally. Of course, crash tests are also run to permit crash sensing.
Furthermore, the pattern recognition algorithm may be trained based on patterns within the signals from the sensors. Thus, by means of a single sensor, it would be possible to determine whether one or more components are operating abnormally. To obtain such a pattern recognition algorithm, tests are conducted using a single sensor, such as a microphone, and causing abnormal operation of one or more components, each component operating abnormally while the other components operate normally and multiple components operating abnormally. In this manner, in use, the pattern recognition algorithm may analyze a signal from a single sensor and determine abnormal operation of one or more components. In some cases, simulations can be used to analytically generate the relevant data.
The invention is also particularly useful in light of the foreseeable implementation of smart highways. Smart highways will result in vehicles traveling down highways under partial or complete control of an automatic system, i.e., not being controlled by the driver. The on-board diagnostic system will thus be able to determine failure of a component prior to and/or upon failure thereof and inform the vehicle's guidance system to cause the vehicle to move out of the stream of traffic, i.e., onto a shoulder of the highway, in a safe and orderly manner. Moreover, the diagnostic system may be controlled or programmed to prevent movement of the disabled vehicle back into the stream of traffic until repair of the component is satisfactorily completed.
In a method in accordance with this embodiment, the operation of the component would be monitored and if abnormal operation of the component is detected, e.g., by any of the methods and apparatus disclosed herein (although other component failure systems may of course be used in this implementation), the vehicle guidance system which controls the movement of the vehicle would be notified, e.g., via a signal from the diagnostic module to the guidance system, and the guidance system would be programmed to move the vehicle out of the stream of traffic, or off of the restricted roadway, possibly to a service station or dealer, upon reception of the particular signal from the diagnostic module. The automatic guidance systems for vehicles traveling on highways may be any existing system or system being developed, such as one based on satellite positioning techniques or ground-based positioning techniques. Since the guidance system may be programmed to ascertain the vehicle's position on the highway, it can determine the vehicle's current position, the nearest location out of the stream of traffic, or off of the restricted roadway, such as an appropriate shoulder or exit to which the vehicle may be moved, and the path of movement of the vehicle from the current position to the location out of the stream of traffic, or off of the restricted roadway. The vehicle may thus be moved along this path under the control of the automatic guidance system. In the alternative, the path may be displayed to a driver and the driver can follow the path, i.e., manually control the vehicle. The diagnostic module and/or guidance system may be designed to prevent re-entry of the vehicle into the stream of traffic, or off of the restricted roadway, until the abnormal operation of the component is satisfactorily addressed.
In general, sensors 300-306 provide a measurement of the state of the sensor, such as its velocity, acceleration, angular orientation or temperature, or a state of the location at which the sensor is mounted. Thus, measurements related to the state of the sensor 300-306 would include measurements of the acceleration of the sensor, measurements of the temperature of the mounting location as well as changes in the state of the sensor and rates of changes of the state of the sensor. As such, any described use or function of the sensors 300-306 above is merely exemplary and is not intended to limit the form of the sensor or its function.
Each of the sensors 300-306 may be single axis, double axis or triaxial accelerometers and/or gyroscopes typically of the MEMS type. MEMS stands for microelectromechanical system and is a term known to those skilled in the art. These sensors 300-306 can either be wired to the central control module or processor directly wherein they would receive power and transmit information, or they could be connected onto the vehicle bus or, in some cases, using RFID technology, the sensors can be wireless and would receive their power through RF from one or more interrogators located in the vehicle. RFID stands for radio frequency identification wherein sensors are each provided with an identification code and designed to be powered by the energy in a radio frequency wave containing that code which is emitted by the interrogator. In this case, the interrogators can be connected either to the vehicle bus or directly to control module. Alternately, an inductive or capacitive power and information transfer system can be used.
One particular implementation will now be described. In this case, each of the sensors 300-306 is a single or dual axis accelerometer. They are made using silicon micromachined technology such as disclosed in U.S. Pat. No. 5,121,180 and U.S. Pat. No. 5,894,090. These are only representative patents of these devices and there exist more than 100 other relevant U.S. patents describing this technology. Commercially available MEMS gyroscopes such as from Systron Doner have accuracies of approximately one degree per second. In contrast, optical gyroscopes typically have accuracies of approximately one degree per hour. Unfortunately, the optical gyroscopes are prohibitively expensive for automotive applications. On the other hand, typical MEMS gyroscopes are not sufficiently accurate for many control applications.
Referring now to
Thus, in connection with the control of parts of the vehicle, location information may be obtained from the GPS receiver 312 and input to a pattern recognition system for consideration when determining a control signal for the part of the vehicle. Position information from the IMU 311 could alternatively or additionally be provided to the pattern recognition system. The location determination by the GPS receiver 312 and IMU 311 may be improved using the Kalman filter embodied in processor 313 in conjunction with the pattern recognition system to diagnose, for example, the state of the vehicle.
Another way to use the IMU 311, GPS receiver 312 and Kalman filter embodied in processor 313 would be to use the GPS receiver 312 and Kalman filter in processor 313 to periodically calibrate the location of the vehicle as determined by the IMU 311 using data from the GPS receiver 312 and the Kalman filter embodied in processor 313. A DGPS receiver 314 could also be coupled to the processor 313 in which case, the processor 313 would receive information from the DGPS receiver 314 and correct the determination of the location of the vehicle as determined by the GPS receiver 312 or the IMU 311.
The angular rate function can be obtained through placing accelerometers at two separated, non-co-located points in a vehicle and using the differential acceleration to obtain an indication of angular motion and angular acceleration. From the variety of accelerometers shown on
Instead of using two accelerometers at separate locations on the vehicle, a single conformal MEMS-IDT gyroscope may be used. A MEMS-IDT gyroscope is a microelectromechanical system-interdigital transducer gyroscope. Such a conformal MEMS-IDT gyroscope is described in a paper by V. K. Varadan, Conformal MEMS-IDT Gyroscopes and Their Comparison With Fiber Optic Gyro, incorporated in its entirety herein. The MEMS-IDT gyroscope is based on the principle of surface acoustic wave (SAW) standing waves on a piezoelectric substrate. A surface acoustic wave resonator is used to create standing waves inside a cavity and the particles at the anti-nodes of the standing waves experience large amplitude of vibrations, which serves as the reference vibrating motion for the gyroscope. Arrays of metallic dots are positioned at the anti-node locations so that the effect of Coriolis force due to rotation will acoustically amplify the magnitude of the waves. Unlike other MEMS gyroscopes, the MEMS-IDT gyroscope has a planar configuration with no suspended resonating mechanical structures.
The system of
Similarly, the tendency of the vehicle to slide or skid can be considerably more accurately determined and again the steering, braking and throttle systems commanded to minimize the unstable vehicle behavior.
Thus, through the sample deployment of inexpensive accelerometers at a variety of locations in the vehicle, significant improvements are many in the areas of vehicle stability control, crash sensing, rollover sensing, and resulting occupant protection technologies.
Finally, the combination of the outputs from these accelerometer sensors and the output of strain gage weight sensors in a vehicle seat, or in/on a support structure of the seat, can be used to make an accurate assessment of the occupancy of the seat and differentiate between animate and inanimate occupants as well as determining where in the seat the occupants are sitting. This can be done by observing the acceleration signals from the sensors of
For this embodiment, sensor 307 in
This application is one in a series of applications covering safety and other systems for vehicles and other uses. The disclosure herein goes beyond that needed to support the claims of the particular invention that is claimed herein. This is not to be construed that the inventors are thereby releasing the unclaimed disclosure and subject matter into the public domain. Rather, it is intended that patent applications have been or will be filed to cover all of the subject matter disclosed above.
The inventions described above are, of course, susceptible to many variations, modifications and changes, all of which are within the skill of the art. It should be understood that all such variations, modifications and changes are within the spirit and scope of the inventions and of the appended claims. Similarly, it will be understood that applicant intends to cover and claim all changes, modifications and variations of the examples of the preferred embodiments of the invention herein disclosed for the purpose of illustration which do not constitute departures from the spirit and scope of the present invention as claimed.
Although several preferred embodiments are illustrated and described above, there are possible combinations using other geometries, materials and different dimensions for the components and different forms of the neural network implementation that perform the same functions. Also, the neural network has been described as an example of one pattern recognition system. Other pattern recognition systems exist and still others are under development and will be available in the future. Such a system can be used to identify crashes requiring the deployment of an occupant restraint system and then, optionally coupled with additional information related to the occupant, for example, create a system that satisfies the requirements of one of the Smart Airbag Phases. Also, with the neural network system described above, the input data to the network may be data which has been pre-processed rather than the raw acceleration data either through a process called “feature extraction”, as described in Green (U.S. Pat. No. 4,906,940) for example, or by integrating the data and inputting the velocity data to the system, for example. This invention is not limited to the above embodiments and should be determined by the following claims.
Claims
1. A method for controlling weighing of a vehicle travelling on a road by means of a weigh station alongside the road, comprising:
- determining weight of the vehicle using a processor on-board the vehicle; and
- transmitting from the vehicle using a telematics device, the determined weight of the vehicle to the weigh station.
2. The method of claim 1, wherein the step of determining the weight of the vehicle comprises processing inertial property data from an inertial measurement unit (IMU) into an indication of the weight of the vehicle.
3. The method of claim 2, wherein the inertial property data from the IMU is multiple sets of inertial property data obtained over time.
4. The method of claim 2, further comprising calibrating the IMU, using the processor, based differential motion of the vehicle over a period of time as determined by a location positioning system.
5. The method of claim 4, wherein the IMU calibrating step comprises using a Kalman filter.
6. The method of claim 1, wherein a vehicle transmitting its determined weight avoids stopping at the weigh station.
7. A method for controlling travel of vehicles on a road having a weigh station alongside the road at which vehicles must provide their vehicle weight at least for some designated times and designated vehicle types, comprising:
- for those vehicles equipped with an on-board weight determining unit, determining weight of the vehicle using a processor on-board the vehicle; and transmitting from the vehicle using a telematics device, the determined weight of the vehicle to the weigh station, whereby a vehicle transmitting its determined weight avoids stopping at the weigh station.
8. The method of claim 7, wherein the on-board weight determining unit includes an inertial measurement unit (IMU) and the step of determining the weight of the vehicle comprises processing inertial property data from the IMU into an indication of the weight of the vehicle.
9. The method of claim 8, wherein the inertial property data from the IMU is multiple sets of inertial property data obtained over time.
10. The method of claim 8, further comprising calibrating the IMU, using the processor, based differential motion of the vehicle over a period of time as determined by a location positioning system.
11. The method of claim 10, wherein the IMU calibrating step comprises using a Kalman filter.
12. A method for managing road information, comprising:
- obtaining road properties from vehicles during travel on the vehicle on a road at an off-vehicle location, the road properties being obtained using an inertial measurement unit on board each vehicle and transmitted from the vehicle using a telematics device;
- storing the road properties from the vehicles at the off-vehicle locations in a data storage device; and
- selectively distributing from the data storage device to the telematics device on vehicles traveling on the road, at least part of the collected road properties to the vehicles traveling on the road.
13. The method of claim 12, further comprising associating the road properties with the weather in the area of the road at the off-vehicle location, the at least part of the collected road properties being distributed to the vehicles traveling on the road based on the weather.
Type: Application
Filed: Jun 30, 2014
Publication Date: Oct 16, 2014
Inventor: David S. Breed (Boonton, NJ)
Application Number: 14/319,642
International Classification: G01G 19/02 (20060101); B60W 40/06 (20060101); G07C 5/00 (20060101);
