Abstract: This application is directed to augmenting training data used for vehicle driving modelling. A computer system obtains a first image of a road and identifies a drivable area of the road within the first image. The computer system obtains an image of an object and generates a second image from the first image by overlaying the image of the object over the drivable area. The second image is added to a corpus of training images to be used by a machine learning system to generate a model for facilitating driving of a vehicle (e.g., at least partial autonomously). In some embodiments, the computer system applies machine learning to train a model using the corpus of training images and distributes the model to one or more vehicles. In use, the model processes road images captured by the one or more vehicles to facilitate vehicle driving.

Abstract: A safety case discovery system includes a scenario framework and safety protocols for edge cases. The safety case discovery system receives sensor data generated by at least one sensor during operation of a vehicle and stores the sensor data in a data warehouse. The data warehouse can be queried based on a predefined scenario description to produce a subset of records which are ranked based on a relevancy of the records to the predefined scenario description. The safety case discovery system deduplicates the ranked results to produce edge cases and updates the safety framework. The safety case discovery system can train a machine learning model vehicle based on the edge cases.

Abstract: The present teaching relates to method, system, medium, and implementations for trailer pose estimation. At least a dimension of a trailer connected to a vehicle is obtained where the vehicle and the trailer have respective opposing surfaces facing one another. A first set of feature points is identified from a first image of a target present on one the opposing surfaces acquired at a first time instant by a sensor deployed on the other the opposing surfaces. A second set of corresponding feature points is identified from a second image of the target acquired at a subsequent time instant by the sensor. Based on the first and second sets of feature points, a first transformation is determined and used to compute a second transformation that the trailer undergoes from the first to the second time instants. A pose of the trailer at the second time instant is estimated based on the second transformation and the dimension.

Abstract: A system for calibrating stereo cameras using time-of-flight (ToF) includes a stereo camera system and a ToF sensor onboard a vehicle. The system receives stereo image data depicting a scene and ToF sensor data captured during stationary or moving operations of the vehicle. The system associates the ToF sensor data to the stereo image data by projection onto the stereo image data to filter out stereo depth errors. The system converts distance measurements based on the ToF sensor data to disparities in the stereo image data. The system further filters out the disparities between the ToF sensor and the stereo camera system using a voting scheme to populate a histogram with disparities. The system can select the peak of the histogram and correct the stereo image data with the disparity results in the peak, to correct stereo image data during operation of the vehicle.

Abstract: The present teaching relates to method and system for sensor protection. A sensor protection system includes an assembly having a first structure embedded in a second structure, wherein a sensor is firmly placed in the first structure for sensing information through the assembly and protected from negative impact of outside environment, and one or more devices mounted on the second structure for providing means to clean a slanted surface of the second structure through which the sensor senses the surrounding information, wherein the one or more devices can be individually or jointly activated as needed to clean the slanted surface in order to prevent degradation of information sensed by the sensor through the assembly.

Abstract: The present teaching relates to different configurations for facilitating calibration of multiple sensors of different types. A plurality of fiducial markers are arranged in space for simultaneously calibrating multiple sensors of different types. Each of the plurality of fiducial markers has a feature point thereon and is provided to enable the multiple sensors to calibrate by detecting the features points and estimating their corresponding 3D coordinates with respect to respective coordinate systems of the multiple sensors.

Abstract: The present teaching relates to method, system, and implementations for glare control. A sensor housing assembly is provided on an autonomous vehicle to house at least one sensor therein and protect the at least one sensor from negative impact from an environment including glare from the environment that prevents the at least one sensor from capturing accurate information regarding a field of view with respect to the environment to facilitate autonomous driving of the autonomous vehicle. The negative impact on the at least one sensor is determined by a controller, which then activates a glare blocking mechanism in response to the determined negative impact. The glare blocking mechanism being mounted on the sensor housing assembly and configured to assist in blocking glare with respect to the at least one sensor.

Abstract: Methods, systems, and non-transitory computer-readable media are configured to perform operations comprising: determining whether a first type of condition associated with an input is satisfied; based at least in part on satisfaction of the first type of condition, generating an estimated value of a parameter associated with operation of a vehicle; and controlling the vehicle based at least in part on the estimated value of the parameter.

Abstract: Embodiments provide a vehicle computer coupled to a vehicle. The vehicle computer may be configured to compute (e.g., generate) a first (e.g., regular, main) trajectory and a second (e.g., fail-safe, minimal risk maneuver) trajectory for the vehicle. Embodiments may provide a fault detection interval during which the second trajectory may be activated due to missing updated trajectory input. Embodiments may also provide a recovery detection interval during which an updated regular trajectory may be generated and followed instead of the fail-safe trajectory if a valid trajectory update is received. Accordingly, embodiments allow for an early activation of a fail-safe trajectory compared to conventional systems, while also allowing the system to recover (e.g., revert back to a modified version of the first trajectory) if a valid trajectory update is received during a recovery detection interval.

Abstract: In some embodiments, an apparatus, comprises an unmanned vehicle configured to be disposed with a vehicle and a processor operatively coupled to the unmanned vehicle. The processor is configured to receive a first signal indicating a stop of the vehicle without a user request and a second signal indicating a location of the vehicle. The processor is configured to determine, based on the location of the vehicle, a target location in a pre-defined area surrounding the vehicle. The processor is configured to send a third signal to the unmanned vehicle to instruct the unmanned vehicle to move to the target location to alert other vehicles via a warning regarding occurrence of the stop.

Abstract: The present teaching relates to method and system for an assembly for protecting a sensor. The assembly includes a first structure having a rectangular prism and connected with a truncated tube, wherein the rectangular prism and the truncated tube have parallel longitudinal axes and the tube is truncated to yield a cross section having a norm that forms an angle with the longitudinal axis of the truncated tube and a second structure of a trapezoid prism having a longitudinal axis parallel to the longitudinal axes and with the first structure embedded therein. The second structure has a slanted surface in the front that meets the cross section of the truncated tube and has the norm of the cross section.

Abstract: Embodiments provide reinforcement learning (RL) techniques for transferring machine learning obtained in one domain to another. A neural network (NN) of a source domain may be trained using training data associated with that domain. In some cases, the training data is inaccessible to a target domain for which a similar NN is desired. A number of target NNs may be initialized with a portion of the parameters transferred from the source NN. A computing agent may utilize RL techniques to train the target NNs using training data of the target domain. The target NN with the most beneficial subset of transferred parameters may be selected. By using the RL techniques to identify the most advantageous combination of transferred parameters, a similarly accurate NN may be provided in the target domain at a fraction of the otherwise needed time and without accessing the original training data.

Abstract: Embodiments provide a vehicle computer coupled to a vehicle. The vehicle computer may be configured to compute (e.g., generate) an automated lane change maneuver moving the vehicle from a first lane into a second, adjacent lane. The lane change maneuver may have commenced while the maneuver was safe to conduct, but a threat vehicle (or another threat object) may be subsequently detected moving into the same target lane. The option to immediately abort the maneuver and return to the original lane may not be appropriate after some time into the lane change maneuver. The vehicle computer may control the vehicle in a lateral position hold along the lane demarcation line for a predetermined amount of time before moving into the second lane when the threat vehicle clears the second lane or returning to the first lane when the threat vehicle does not clear the second lane.

Abstract: In one or more embodiments, a method comprises receiving, at a processor, sensor data from a sensor at a vehicle that is moving while in an autonomous mode. A position and an orientation of the vehicle based on the sensor data is determined at the processor. A lateral deviation of the vehicle from a planned path based on the position and the orientation of the vehicle while the vehicle is moving in the autonomous mode is calculated at the processor. In response to the lateral deviation exceeding a predefined lateral deviation threshold, the autonomous mode is disengaged.

Abstract: The present teaching relates to method, system, medium, and implementation for sensor protection. It is determined whether a cleaning operation needs to be applied to a portion of a sensor assembly deployed on a vehicle and hosting a sensor therein to acquire information of a surrounding through the portion to facilitate autonomous driving. If the cleaning operation is to be applied, at least one type of cleaning is selected to be carried out by one or more devices mounted on the sensor assembly with respect to the portion. For each of the devices, needed control signal is generated for activating the device to perform a cleaning task the device is configured for and sent to the device to clean the certain portion of the sensor assembly in order to prevent degradation in the information acquired by the sensor through the portion.

Abstract: A method comprises: receiving, at a processor, a first image from a first camera from a stereo camera pair and a second image from a second camera from the stereo camera pair. The method also includes determining, at the processor using a machine learning model, a first set of objects in the first image. The processor determines an object type. The processor identifies a second set of objects in the second image associated with the first plurality of objects. The method also includes calculating, at the processor, a set of disparity values between the first image and the second image based on (1) an object from the first set of objects, (2) an object from the second set of objects and associated with the object from the first set of objects, and (3) an object type of the object from the first set of objects.

Abstract: The present teaching relates to method, system, medium, and implementation of automatic motion planning for an autonomous driving vehicle. Information is obtained with respect to a current location of the autonomous driving vehicle operating in a current vehicle motion, wherein the information is to be used to estimate the operational capability of the autonomous driving vehicle with respect to the current location. Sensor data are obtained, via one or more sensors in one or more media types. A reaction of a passenger present in the vehicle with respect to a current vehicle motion is estimated based on the sensor data. Motion planning is performed based on the information with respect to the current location and the reaction of the passenger to the current vehicle motion.

Abstract: A non-transitory processor-readable medium stores code representing instructions to be executed by the processor. The code comprises code to cause the processor to receive a first image and a second image from a stereo camera pair disposed with a vehicle. The code causes the processor to detect, using a machine learning model, an object based on the first image, the object located within a pre-defined area within a vicinity of the vehicle. The code causes the processor to determine a distance between the object and the vehicle based on disparity between the first image and the second image. The code causes the processor to determine a longitudinal value of the vehicle based on the distance and a height of the vehicle. The code causes the processor to send an instruction to facilitate driving of the vehicle based on a road profile associated with the longitudinal value.

Abstract: The present teaching relates to method, system, medium, and implementation of human-like vehicle control for an autonomous vehicle. Information related to a target motion to be achieved by the autonomous vehicle is received, wherein the information includes a current vehicle state of the autonomous vehicle. A first vehicle control signal is generated with respect to the target motion and the given vehicle state in accordance with a vehicle kinematic model. A second vehicle control signal is generated in accordance with a human-like vehicle control model, with respect to the target motion, the given vehicle state, and the first vehicle control signal, wherein the second vehicle control signal modifies the first vehicle control signal to achieve human-like vehicle control behavior.

Abstract: An apparatus includes a processor configured to be disposed with a vehicle and a memory coupled to the processor. The memory stores instructions to cause the processor to receive, at least two of: radar data, camera data, lidar data, or sonar data. The sensor data is associated with a predefined region of a vicinity of the vehicle while the vehicle is traveling during a first time period. At least a portion of the vehicle is positioned within the predefined region during the first time period. The method also includes detecting that no other vehicle is present within the predefined region. An environment of the vehicle during the first time period is classified as one state from a set of states that includes at least one of dry, light rain, heavy rain, light snow, or heavy snow, based on at least two of the sensor data to produce an environment classification. An operational parameter of the vehicle based on the environment classification is modified.