Abstract: Aspects of the disclosure provide for controlling an autonomous vehicle to enter a driveway. As an example, an instruction to pickup or drop off a passenger at a location may be received. It may be determined that the vehicle is arriving in a lane on an opposite side of a street as the location, the lane having a traffic direction opposite to a traffic direction of a lane adjacent to the location. A difficulty score to maneuver the vehicle to the lane adjacent to the location may be determined, and the difficulty score may be compared to a predetermined difficulty score. Based on the comparison, it may be determined to an available driveway on the same side of the street as the location. The vehicle may be controlled to enter the available driveway.

Abstract: Recent decades have brought tremendous advances in communication systems which have given rise to the global proliferation of smartphones and other mobile communication systems. A signal processing system implements technical solutions for determining building footprints from inputs including signal data associated with mobile communication devices.

Abstract: Autonomous driving system methods and devices which trigger vehicular actions based on the monitoring of one or more occupants of a vehicle are presented. The methods, and corresponding devices, may include identifying a plurality of features in a plurality of subsets of image data detailing the one or more occupants; tracking changes over time of the plurality of features over the plurality of subsets of image data; determining a state, from a plurality of states, of the one or more occupants based on the tracked changes; and triggering the vehicular action based on the determined state.

Abstract: A vehicle includes a camera to recognize at least one sign installed on a road and a controller to determine a type of the at least one sign, in a case in which the at least one sign corresponds to a road freezing sign based on referring to a memory, determine whether a temperature sensed by a temperature sensor installed outside the vehicle is a freezing danger temperature, and in a case in which the temperature sensed by the temperature sensor corresponds to the freezing danger temperature, determine whether a drive mode of the vehicle is set to a winter mode. The controller displays a winter mode automatic switching notification message through an instrument panel and to transmit a command for switching to the winter mode to a drive mode controller in a case in which the drive mode of the vehicle is not set to the winter mode.

Abstract: An operation of an AV in a scene is simulated. The AV includes sensors detecting the scene and generating sensor data. The sensor data can be input into a control model that outputs control signals, in accordance with which the AV operates. A learning cost, i.e., a cost of training or applying the control model, is determined. A performance of the AV during the simulation is evaluated. A simulation cost, i.e., a cost of running the simulation is determined. The control model can be optimized based on the learning cost and the performance of the AV. Settings of the sensors can be optimized based on the simulation cost and the performance of the AV. The optimization of the control model or the settings of the sensors can be done through reinforcement learning.

Abstract: A system and a method for generating a trajectory of a target device from a current position to a goal position within an environment is provided. The method may include: inputting physical workspace information associated with the environment in which the target device, to a first neural network to obtain a set of weights representing a cost-to-go function that defines a cost-to-go function relating to a length of a collision-free path from one position to the goal position; configuring a second neural network based on the set of weights; identifying a next position of the target device based on the current position and a motion control input of the target; and inputting the identified next position of the target device and the goal position to the second neural network to identify the trajectory to the goal position, and the motion control input corresponding to the trajectory.

Abstract: A control device performs: setting a moving area including a trajectory along which a mobile object is scheduled to move with behavior of the mobile object and a surrounding area when the mobile object moves straight, turns right, or turns left; controlling an information provision device configured to provide a driver with information such that the driver is notified of presence of a first type of object when a first type of object is present in the surrounding area and the driver does not see the first type of object; and controlling the information provision device such that the driver is re-notified of the first type of object when the first type of object is present in the moving area and the driver does not see the first type of object after the driver has seen the first type of object through the notification.

Abstract: Disclosed are embodiments that provide a transportation environment data service. The transportation environmental data service includes harvesting services that crawl roadside infrastructure solutions to obtain sensor data collected from sensors physically positioned at the roadside infrastructure. In some cases, the roadside infrastructure solutions perform additional processing on the sensor data. For example, some roadside infrastructure performs object detection and/or object recognition. When encountering these solutions, the edge or harvesting service also collects the object detection and/or object recognition information. Customers can subscribe to various data services provided by the transportation environment data service. For example, some subscribers indicate an interest in any updates of environmental data for a particular region. Other subscribers are interested in video data associated with any vehicular accidents detected by the transportation environment data service.

Abstract: A system and corresponding method for autonomous driving of a vehicle are provided. The system comprises at least one neural network (NN) that generates at least one output for controlling the autonomous driving. The system further comprises a main data path that routes bulk sensor data to the at least one NN and a low-latency data path with reduced latency relative to the main data path. The low-latency data path routes limited sensor data to the at least one NN which, in turn, employs the limited sensor data to improve performance of the at least one NN's processing of the bulk sensor data for generating the at least one output. Improving performance of the at least one NN's processing of the bulk sensor data enables the system to, for example, identify a safety hazard sooner, enabling the autonomous driving to divert the vehicle and avoid contact with the safety hazard.

Abstract: Systems and apparatuses for receiving data from a plurality of sensors and using the data, as well as other data, to generate a parking recommendation for an autonomous vehicle and instruct the autonomous vehicle to travel to the recommended parking location are provided. Data may be received from a plurality of sensors within a first autonomous vehicle, as well as from other vehicles and/or structures. Historical parking data associated with the first autonomous vehicle may also be extracted. In some examples, an expected future trip of the first autonomous vehicle may be determined. The system may then evaluate the data to generate a parking recommendation for the first autonomous vehicle. The system may generate and transmit instructions for traveling from a current location to the recommended parking location and may cause the first autonomous vehicle to travel to the recommended parking location.

Abstract: System and techniques for vehicle operation safety model (VOSM) compliance measurement are described herein. A subject vehicle is tested in a vehicle following scenario against VOSM parameter compliance. The test measures the subject vehicle activity during phases of the following scenario in which a lead vehicle slows and produces log data and calculations that form the basis of a VOSM compliance measurement.

Abstract: A method for operating a plurality of vehicles is disclosed. The method comprises receiving (101) a first set of vehicle data and a second set of vehicle data, the vehicle data comprising information about each vehicle of the plurality of vehicles, each vehicle operating along at least one fixed route, receiving (102) a first set of environmental data and a second set of environmental data, the environmental data comprising information about each fixed route, and estimating (103), by means of the global self-learning model and each local-self learning model, a schedule parameter for each vehicle of the plurality of vehicles based on the received first set of vehicle data, the received first set of environmental data, the received second set of vehicle data, the received second set of environmental data, and a predefined interaction model between the global self-learning model and each local-self learning model.

Abstract: A method of implementing distributed AI or ML learning for autonomous vehicles is disclosed. An AI or ML model specific to a location or a type of the location is generated at a vehicle. In response to a detection that the vehicle is within a proximity to a road side unit (RSU) associated with the location or the type of the location or the vehicle is within a proximity to an additional vehicle that is present or anticipated to be present at the location or the type of the location, causing an AI or ML model transmission to the additional vehicle or the RSU. Based on the causing the AI or ML model reception, causing deployment of an additional AI or ML model in the vehicle to optimize the vehicle for the location or the type of the location.

Abstract: An approach is provided for map updates using a machine learning-based persistence filter. The approach involves, for example, processing sensor data collected from vehicles to determine positive observations and/or negative observations of a map or geographic feature. The method also comprises providing the positive and/or observations as an input sequence to a machine learning model. The machine learning model, for instance, includes an input layer that feeds the input sequence to a long short-term memory (LSTM) layer or other type of recurrent neural network (RNN) layer. The LSTM or RNN layer connects to a fully connected layer and then to an output layer. The output layer outputs a predicted sequence of positive and/or negative observations. The method further comprises initiating an update of a geographic database to add, remove, or update the geographic feature based on the predicted sequence.

Abstract: A method is provided for identifying an obstacle during a work operation of a tractor and attached implement combination. The method includes detecting an object during the work operation, and identifying the detected object as an obstacle in dependence on the predefined classification parameter for classification of the detected object or a predefined work parameter of the tractor or the implement.

Abstract: Systems, methods, and non-transitory computer-readable media can determine a first utility metric associated with a region and first autonomous vehicle eligibility criteria, wherein the first utility metric is determined based on a first plurality of rides and a subset of the first plurality of rides that can be successfully executed within the region based on the first autonomous vehicle eligibility criteria. A second utility metric associated with the region and second autonomous vehicle eligibility criteria can be determined, wherein the second utility metric is determined based on a second plurality of rides and a subset of the second plurality of rides that can be successfully executed within the region based on the second autonomous vehicle eligibility criteria. An autonomous vehicle associated with the first autonomous vehicle eligibility criteria can be selected to drive in the region based on a comparison of the first utility metric and the second utility metric.

Abstract: The present invention provides a system (100) for extending voice commands and determining a user's location. The system (100) includes a plurality of interconnected user devices (10a, 10b, 10c, 10d, 10e, 10f, 10g, 10h, 10i, 10j, . . . ) forming a meshed network (1000), to receive a voice command from a user (1) for extending the range of voice command detection. Further, the system (100) includes a base unit (20) connected to one of the user devices, which receives the voice command from the user device (10) closest to the base unit (20) and calculates ranging data. A master controller (30), connected to multiple base units (20a, 20b, 20c, 20d, . . . ), calculates the user's location based on ranging data, processes the voice command to determine the user's intent, and performs an action based on the determined intent and sends the user's location to a central server (40), to visualize the user's location on map.

Abstract: The present disclosure provides a traffic prediction apparatus, system, method and program capable of predicting the number of vehicles and the speed of vehicles in a predetermined time and in a predetermined range based on waterfall data of vehicles. The traffic prediction apparatus comprises acquisition means for acquiring waterfall data comprising a generation position of a vibration, a generation time of the vibration and an amplitude of the vibration generated by a vehicle traveling on a road, pre-processing means for transforming the acquired waterfall data, generation means for training a portion of the plurality of processed waterfall data and at least one corresponding ground truth traffic property used as plurality of labels to generate a trained model, wherein the ground truth may be obtained from a secondary acquisition means, and prediction means for predicting at least one traffic property for a processed waterfall data.

Abstract: Aspects of the technology relate to exception handling for a vehicle. For instance, a current trajectory for the vehicle and sensor data corresponding to one or more objects may be received. Based on the received sensor data, projected trajectories of the one or more objects may be determined. Potential collisions with the one or more objects may be determined based on the projected trajectories and the current trajectory. One of the potential collisions that is earliest in time may be identified. Based on the one of the potential collisions, a safety-time-horizon (STH) may be identified. When a runtime exception occurs, before performing a precautionary maneuver to avoid a collision, waiting no longer than the STH for the runtime exception to resolve.

Abstract: A system and corresponding method for autonomous driving of a vehicle are provided. The system comprises at least one neural network (NN) that generates at least one output for controlling the autonomous driving. The system further comprises a main data path that routes bulk sensor data to the at least one NN and a low-latency data path with reduced latency relative to the main data path. The low-latency data path routes limited sensor data to the at least one NN which, in turn, employs the limited sensor data to improve performance of the at least one NN's processing of the bulk sensor data for generating the at least one output. Improving performance of the at least one NN's processing of the bulk sensor data enables the system to, for example, identify a safety hazard sooner, enabling the autonomous driving to divert the vehicle and avoid contact with the safety hazard.

Abstract: Systems and methods for real time routing of vehicles. In particular, systems and methods are provided for real time routing of vehicles based on encounters with vulnerable road users. Vehicle perception and communication systems can provide live feedback of vulnerable road user locations. Various vulnerable road user behaviors can be identified by vehicle perception systems and flagged for avoidance in a central mapping system. The feedback can be used to route other fleet vehicles around the vulnerable road users.

Abstract: Aspects of the invention may be related to a computer system and a computerized method of controlling a marine vessel. Embodiments may include: receiving, by a controller, a position of the marine vessel from a positioning system; receiving, by the controller, geographical data, from at least one database, the geographical data may include at least a map of seabed depths; receiving, by the controller, a heading direction and speed of the marine vessel; calculating, by the controller, a safety zone ahead of the marine vessel based on the position, the heading direction and the speed of the marine vessel; identifying a location of a critical seabed depth inside the safety zone based on the received, the geographical data; and changing, by the controller, a state of a propelling unit of the marine vessel when a critical seabed depth was identified inside the safety zone.

Abstract: A vehicle controller includes a memory configured to store map information including information on a predetermined area used in autonomous driving control of a vehicle; and a processor configured to: refer to the map information to identify an autonomous-driving practicable point at which a planned travel route from the current position of the vehicle outside the predetermined area to a destination of the vehicle enters the predetermined area, calculate a manual-driving section length from the current position of the vehicle to the autonomous-driving practicable point along the planned travel route, and notify the calculated manual-driving section length to a driver of the vehicle via a notification device provided in the interior of the vehicle.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for controlling a robot to perform a custom real-time action that uses a callback function. One of the methods comprises receiving a definition of a custom real-time control function that specifies a custom callback function, an action, and a custom reaction that references the custom callback function; providing a command to initiate the action; repeatedly executing, by the control layer of the real-time robotics control framework, the custom real-time control function at each tick of a real-time robotics system driving one or more physical robots, including: obtaining current values of one or more state variables, evaluating the custom reaction specified by the custom real-time control function according to the current values of the one or more state variables, and whenever the one or more conditions of the custom reaction are satisfied, invoking the custom callback function.

Abstract: Methods and systems are provided for detecting and responding to dangerous road users. In some aspects, a process can include steps for receiving sensor data of a detected object from an autonomous vehicle, determining whether the detected object is exhibiting a dangerous attribute, generating output data based on the determining of whether the detected object is exhibiting the dangerous attribute, and updating a machine learning model based on the output data relating to the dangerous attribute.

Abstract: A first distinct vehicle operational scenario is identified for an autonomous vehicle (AV). A first set of candidate vehicle control actions are received from a model that provides a first solution to the first distinct vehicle operational scenario. An action is selected from the first set of candidate vehicle control actions. The AV is controlled based on the action. The first solution is obtained offline in a first idealized situation that is decoupled from a current context of the AV.

Abstract: A framework for offline learning from a set of diverse and suboptimal demonstrations operates by selectively imitating local sequences from the dataset. At least one embodiment recovers performant policies from large manipulation datasets by decomposing the problem into a goal-conditioned imitation and a high-level goal selection mechanism.

Abstract: Autonomous driving system methods and devices which trigger vehicular actions based on the monitoring of one or more occupants of a vehicle are presented. The methods, and corresponding devices, may include identifying a plurality of features in a plurality of subsets of image data detailing the one or more occupants; tracking changes over time of the plurality of features over the plurality of subsets of image data; determining a state, from a plurality of states, of the one or more occupants based on the tracked changes; and triggering the vehicular action based on the determined state.

Abstract: An autonomous off-road vehicle, upon encountering an obstruction while navigating a route, can apply a first machine-learned model to identify the obstruction. In the event that the first machine-learned model cannot identify the obstruction, the autonomous off-road vehicle can apply a second machine-learned model configured to determine whether or not the obstruction can be ignored, for instance based on dimensions of the obstruction. If the obstruction can be ignored, the autonomous off-road vehicle can continue navigating the route. If the obstruction cannot be ignored, the autonomous off-road vehicle can modify the route, can stop, can flag the obstruction to a remote human operator, can modify an interface of a human operator to display a notification or a video feed from the vehicle, and the like.

Abstract: A driving analysis server may be configured to receive vehicle operation data from mobile devices respectively disposed within the vehicles, and may use the data to group the vehicles into multiple groups. A driving pattern for each vehicle may be established and compared against a group driving pattern of its corresponding group to identify outliers. A driver score the outliers may be adjusted positively or negatively based on whether the outlier behaved in a manner more or less safe than its group. Further, unsafe driving events performed by the outlier that were the result of another vehicle's unsafe driving event may be ignored or positively accounted for in determining or adjusting the outlier's driver score.

Abstract: Aspects of the disclosure provide for controlling an autonomous vehicle to enter a driveway. As an example, an instruction to pickup or drop off a passenger at a location may be received. It may be determined that the vehicle is arriving in a lane on an opposite side of a street as the location, the lane having a traffic direction opposite to a traffic direction of a lane adjacent to the location. A difficulty score to maneuver the vehicle to the lane adjacent to the location may be determined, and the difficulty score may be compared to a predetermined difficulty score. Based on the comparison, it may be determined to an available driveway on the same side of the street as the location. The vehicle may be controlled to enter the available driveway.

Abstract: A method may include obtaining, by an autonomous driving system, input information relating to an autonomous vehicle (AV). The method may include determining, by the autonomous driving system, one or more driving signals that describe operations of the AV. The method may include instructing the AV to move according to the driving signals and simulating, by a virtual driving system, movement of the AV based on the driving signals, wherein the simulating the movement of the AV occurs concurrently with the AV moving according to the driving signals as instructed.

Abstract: Disclosed are an autonomous driving apparatus and method for an ego vehicle that autonomously travels. The autonomous driving apparatus includes a first sensor to detect a nearby vehicle nearby an ego vehicle, a memory to store map information, and a processor including a driving trajectory generator to generate a first driving trajectory of the ego vehicle and a second driving trajectory of the nearby vehicle based on the map information stored in the memory and a control processor configured to control autonomous driving of the ego vehicle based on the first and second driving trajectories generated by the driving trajectory generator.

Abstract: Sensor data collected from an autonomous vehicle data can be labeled using sensor data collected from an additional vehicle. The additional vehicle can include a non-autonomous vehicle mounted with a removable hardware pod. In many implementations, removable hardware pods can be vehicle agnostic. In many implementations, generated labels can be utilized to train a machine learning model which can generate one or more control signals for the autonomous vehicle.

Abstract: A method for safely determining a position information of a train on a track includes an on-board system determining appearance characteristics, current distances relative to the train and current angular positions relative to the train of passive trackside structures with a first sensor arrangement of a first localization stage of the on-board system. The on-board system stores a map data base in which georeferenced locations and appearance characteristics of the passive trackside structures are registered. A first position information about the train is derived from a comparison of determined current distances and current angular positions and the registered locations of allocated passive trackside structures by the first localization stage. A second position information about the train is derived from satellite signals determined by a second sensor arrangement of a second localization stage. The first and second position information undergo a data fusion resulting in a consolidated position information.

Abstract: A driving analysis server may be configured to receive vehicle operation data from mobile devices respectively disposed within the vehicles, and may use the data to group the vehicles into multiple groups. A driving pattern for each vehicle may be established and compared against a group driving pattern of its corresponding group to identify outliers. A driver score the outliers may be adjusted positively or negatively based on whether the outlier behaved in a manner more or less safe than its group. Further, unsafe driving events performed by the outlier that were the result of another vehicle's unsafe driving event may be ignored or positively accounted for in determining or adjusting the outlier's driver score.

Abstract: A trajectory for a vehicle can be generated using a lateral offset bias. The vehicle, such as an autonomous vehicle (AV), may be directed to follow reference trajectory for through an environment. The AV may determine a segment associated with the reference trajectory based on curvatures of the reference trajectory, determine a lateral offset bias associated with the segment based at least in part on, for example, one or more of a speed or acceleration of the vehicle, and determine a candidate trajectory for the autonomous vehicle based at least in part on the lateral offset bias. The candidate trajectory may then be used to control the autonomous vehicle.

Abstract: In one embodiment, a method includes accessing a plurality of content objects, generating a plurality of voxelized representations for the plurality of content objects, respectively, generating one or more building blocks based on one or more sparse convolutions, which includes determining one or more active sites for each of the plurality of content objects based on the voxelized representation of each of the plurality of content objects and applying the one or more sparse convolutions to the one or more active sites, and training a machine-learning model based on a convolutional network including the one or more building blocks.

Abstract: A virtual teaching system includes a semi-autonomous device to perform a task such as cleaning a target environment. The semi-autonomous device includes one or more sensors configured to record environmental data from the target environment that can be used to construct a virtual environment. The semi-autonomous device is operably coupled to an analysis system. The analysis system includes a processor to perform multiple functions, such as constructing the virtual environment from the recorded environmental data and supporting operation of a user interface. The user interface can be operably coupled to the processor, allowing a human operator to teach a virtual device in the virtual environment to perform an action sequence. Once the virtual device has been taught an action sequence in the virtual environment, the analysis system can transfer the recorded action sequence to the semi-autonomous device for use in the target environment.

Abstract: A system and corresponding method for autonomous driving of a vehicle are provided. The system comprises at least one neural network (NN) that generates at least one output for controlling the autonomous driving. The system further comprises a main data path that routes bulk sensor data to the at least one NN and a low-latency data path with reduced latency relative to the main data path. The low-latency data path routes limited sensor data to the at least one NN which, in turn, employs the limited sensor data to improve performance of the at least one NN's processing of the bulk sensor data for generating the at least one output. Improving performance of the at least one NN's processing of the bulk sensor data enables the system to, for example, identify a safety hazard sooner, enabling the autonomous driving to divert the vehicle and avoid contact with the safety hazard.

Abstract: The parking assist apparatus comprises a first vehicle stop apparatus operated for a purpose of stopping a vehicle, a controller configured to be capable of performing parking assist control for parking a vehicle in a parking position and to finish the parking assist control by performing control for instructing a driver of the vehicle to operate the first vehicle stop apparatus or by controlling the first vehicle stop apparatus, and a second vehicle stop apparatus operated for a purpose of stopping the vehicle. When it is determined while the parking assist control is being performed that there is a malfunction in an apparatus for detecting whether or not operation of the first vehicle stop apparatus has been performed, the controller is configured to finish the parking assist control by performing control for instructing the driver to operate the second vehicle stop apparatus or by controlling the second vehicle stop apparatus.

Abstract: A Light Detection And Ranging (LIDAR) apparatus includes a pulsed light source to emit optical signals, a detector array comprising single-photon detectors to output respective detection signals indicating times of arrival of a plurality of photons incident thereon, and processing circuitry to receive the respective detection signals.

Abstract: An autonomous driving vehicle (ADV) receives instructions for a human test driver to drive the ADV in manual mode and to collect a specified amount of driving data for one or more specified driving categories. As the user drivers the ADV in manual mode, driving data corresponding to the one or more driving categories is logged. A user interface of the ADV displays the one or more driving categories that the human driver is instructed collect data upon, and a progress indicator for each of these categories as the human driving progresses. The driving data is uploaded to a server for machine learning. If the server machine learning achieves a threshold grading amount of the uploaded data to variables of a dynamic self-driving model, then the server generates an ADV self-driving model, and distributes the model to one or more ADVs that are navigated in the self-driving mode.

Abstract: A training system for training a trained model for use by a navigating robot to perform visual navigation includes memory including N base virtual training environments, each of the N base virtual training environments including a field of view at a location within an indoor space, where N is an integer greater than 1. A randomization module is configured to generate N varied virtual training environments based on the N base virtual training environments, respectively, by varying at least one characteristic of the respective N base virtual training environments. A training module is configured to train the trained model for use by the navigating robot to perform visual navigation based on a training set including: the N base virtual training environments; and the N varied virtual training environments.

Abstract: A device receives network constraints associated with a network connected to a vehicle device, data collection constraints, and vehicle device constraints. The device determines a first sampling rate, a first time period, a second sampling rate, and a second time period for collecting vehicle data based on the network constraints, the data collection constraints, and the vehicle device constraints, wherein the first sampling rate is different than the second sampling rate. The device receives first vehicle data, provided at the first sampling rate and for the first time period, and second vehicle data, provided at the second sampling rate and for the second time period, and processes the first and second vehicle data, with a machine learning model, to identify a driving behavior or a vehicle issue. The device performs one or more actions based on the driving behavior or the vehicle issue.

Abstract: In one embodiment, a system perceives an environment surrounding an ADV including a vehicle in front of the ADV. The system determines whether the vehicle in front is slipping backwards based on the perception. The system determines whether the vehicle in front is situated on a road with a slope based on map information. The system determines whether a tail light or a brake light of the vehicle in front is turned on based on the perception. If it is determined that the vehicle in front is situated on a sloped road, is slipping backwards, and the tail light or the brake light is not turned on, the system calculates a time to impact or a distance to impact based on the distance and speed of the vehicle in front.

Abstract: An autonomous mobile device (AMD) moves through a physical space without human intervention. Data from sensors on the AMD are used to determine an occupancy map indicative of obstacles and their locations in the physical space. A navigation map is created from the occupancy map by inflating the apparent size of the obstacles indicated by the occupancy map. In one implementation the inflation process dynamically skips some portions of the occupancy map based on the presence of an obstacle and a designated inflation distance. The extent of the inflation is based on a size of the AMD. The navigation map is then used for path planning.

Abstract: A driving simulation system may perform simulations using one or more of a virtual driving simulator or a log-based driving simulator. A log-based simulator may replay data from logs to test whether a control algorithm is successfully capable of navigating a scenario. However, when such log-based simulations are invalid (e.g., based on the results of the simulations or evaluations of conditions of the simulation), a virtual simulation may be generated for the vehicle control system during which new log data may be captured. Such a virtual simulation may comprise sensor simulations and more sophisticated object control for ensuring convergence to a scenario to be tested. The new log data may be used for additional log-based simulations, thereby improving the durability and flexibility for testing and/or validation, while reducing the computational overhead of driving simulation scenarios.

Abstract: In one embodiment, an open space model is generated for a system to plan trajectories for an ADV in an open space. The system perceives an environment surrounding an ADV including one or more obstacles. The system determines a target function for the open space model based on constraints for the one or more obstacles and map information. The system iteratively, performs a first quadratic programming (QP) optimization on the target function based on a first trajectory while fixing a first set of variables, and performs a second QP optimization on the target function based on a result of the first QP optimization while fixing a second set of variables. The system generates a second trajectory based on results of the first and the second QP optimizations to control the ADV autonomously according to the second trajectory.

Abstract: A method performed by an apparatus is described. The method includes receiving remote data. The method also includes adapting model processing based on the remote data. The method further includes determining a driving decision based on the adapted model processing. In some examples, adapting model processing may include selecting a model, adjusting model scheduling, and/or adjusting a frame computation frequency.