Gm Global Technolgoy Operations, Inc.

Sayyed Rouhollah Jafari Tafti

Syed B Mehdi

Pinaki Gupta

Daniel Aguilar Marsillach

Brent Navin Roger Bacchus

Michael A Losh

Rasoul Salehi

Fan Bai

Shuqing Zeng

Bakhtiar Brian Litkouhi

Nikolai K Moshchuk

Mohammad Naserian

Allan K Lewis

Reza Zarringhalam

Curtis L Hay

Wen-Chiao Lin

Soheil Samii

Kamran Ali

Jacob Alan Bond

Edward T Heil

Alireza Esna Ashari Esfahani

Seyedalireza Kasaiezadeh Mahabadi

Praveen Palanisamy

Brent N Bacchus

Marcus J Huber

Jordan Ford

Stefano Bonasera

A method of adaptively tuning parameters in action planning for automated driving of a vehicle to a destination is provided. The method comprises receiving sensor data from a vehicle sensor, lane data of a road plan to the destination, and a plurality of first hyperparameters in a reinforcement learning agent at an initial state and an initial timestamp. The reinforcement learning agent has a planning policy. The method comprises adjusting the plurality of first hyperparameters via the planning policy having at least one first activation function to define an output defining a plurality of second hyperparameters. The method comprises determining a baseline trajectory action based on a trajectory reward value at a final state and a final timestamp. The method comprises modifying the baseline trajectory action between the initial state and the final state to define a refined trajectory action and controlling the vehicle based on the refined trajectory action.

System and method of adaptively tuning parameters in action planning for automated driving

A preemptive maneuvering system for viewing a status of a traffic control device by one or more cameras of an autonomous vehicle includes one or more controllers, where a moving obstruction is in front of the autonomous vehicle. The one or more controllers execute instructions to instruct the autonomous vehicle to execute one or more preemptive maneuvers to position the autonomous vehicle at the target following distance from the moving obstruction along the roadway.

Preemptive maneuvering system for viewing a traffic control device

A system and method for trajectory planning selects a trajectory from a plurality of candidate trajectories to direct an autonomous vehicle. The system includes a controller in communication with a global positioning system (GPS) and at least one sensor. The GPS receives a plurality of global scene information. The at least one sensor collects local scene information. The controller is programmed to receive global scene information and local scene information, identify the location of the occluding obstacle. Additionally, the controller is programmed to determine an occluded portion of the roadway, identify a location of the scene actors on the roadway, and calculate a contracted occluded portion within the occluded portion. Furthermore, the system generates candidate trajectories for the autonomous vehicle to travel along, score the candidate trajectories, and select one of the candidate trajectories based on the assigned score.

System and method for trajectory planning

A system for planning motion of a vehicle includes a controller having a processor and tangible, non-transitory memory on which instructions are recorded. The vehicle is capable of an automated driving operation. A planning module is executable by the controller to generate a vehicle trajectory constrained by a set of planning parameters. During a degraded driving situation, an adjustment module is executable by the controller to generate at least one adjusted parameter in the set of planning parameters. The controller is adapted to register occurrence of the degraded driving situation when the planning module is unable to generate a feasible path for the vehicle constrained by the set of planning parameters. The planning module is adapted to generate a modified trajectory plan based on the at least one adjusted parameter, with the automated driving operation being based on the modified trajectory plan during the degraded driving situation.

Adjustment of vehicle planning parameters in a degraded driving situation

A method generating planning-based attention signals includes receiving driving-scene data. The driving-scene data is indicative of a driving scene around a host vehicle. The driving-scene data includes map data and localization data. The driving scene includes a plurality of actors. The method further includes converting the driving-scene data into a scene-graph. The method further includes inputting the scene-graph into a deep neural network (DNN). Further, the method includes determining an attention score for each of the plurality of actors using the DNN and the scene-graph. The attention score of each of the plurality of actors represents a priority given to each of the plurality of actors in the driving scene. The method further includes commanding the host vehicle to autonomously drive according to a trajectory determined by taking into account the attention score of each of the plurality of actors.

Algorithm to generate planning-based attention signals

A method for mitigating an adversarial attack includes receiving input data. The input data includes sensor data from a plurality of sensors and map data. The method further includes monitoring, in real time, an environment around an autonomous vehicle to identify a region that is possibly subject to an adversarial attack and determining a probability of the adversarial attack in the region. The method further includes determining whether the probability of the adversarial attack in the region that is possibly subject to the adversarial attack is greater than a predetermined threshold and, in response, planning a motion of the autonomous vehicle by taking into account the adversarial attack to generate a planned motion. The method further includes controlling a host vehicle to move in accordance with the planned motion.

Algorithm to detect and mitigate real-time perceptual adversial attacks on autonomous vehicles

A system comprises a computer including a processor and a memory. The memory includes instructions such that the processor is programmed to: receive sensor data representing a perceived driving environment, select a reinforcement learning agent from a plurality of reinforcement learning agents based on a challenge score calculated using the sensor data and a desired driving style, and generate, via the selected reinforcement learning agent, a driving action based on the sensor data.

Automated driving system with desired level of driving aggressiveness

A vehicle trajectory planning system includes a perception system of a host vehicle collecting information from multiple sources and communicating with a computer. A fusion module fuses scene information from a map and perception items identified by the perception system. A behavior planning module receives an output of the fusion module and produces a host vehicle baseline trajectory. A trajectory and motion planning module receives an output of the fusion module in parallel with the behavior planning module. The trajectory and motion planning module determines a reference trajectory and operation corridor for a host vehicle. A disturbance and reachability refiner module receives an output of the trajectory and motion planning module including the reference trajectory and operation corridor. An algorithm is applied to adjust and re-plan the host vehicle baseline trajectory to be robust to a range of inclement weather conditions.

Robust trajectory planning using data-driven reachability corridors

A method for efficient autonomous driving planning includes receiving a current driving-scene data and a predicted driving-scene data. The current driving-scene data is indicative of a current driving scene around a host vehicle. The predicted driving-scene data is indicative of a predicted driving scene around the host vehicle. The predicted driving scene around the host vehicle is different from the current driving scene around the host vehicle. The method further includes converting the current driving-scene data and the predicted driving-scene data into a first scene-graph and a second scene-graph, respectively. The method further includes determining a plurality of scene change metrics using the first scene-graph and the second scene-graph. The method further includes selecting between a first trajectory planning process and a second trajectory planning process based on the plurality of scene change metrics.

System and method for efficient planning under uncertainty for autonomous vehicles

A method for generating an occlusion map using crowdsourcing includes receiving occlusion data from a host vehicle. The occlusion data includes a plurality of virtual detection lines each extending from the sensor and intersecting an off-road area. The method further includes marking all non-occluded points that are located inside the off-road area and that are located within a predetermined distance from each plurality of virtual detection lines. The method further includes determining a non-occluded region inside the off-road area using the non-occluded points that have been previously marked inside the off-road area, wherein the non-occluded region is part of non-occlusion data. The method further includes generating an occlusion map using the non-occlusion data. The method further includes transmitting the occlusion map to the host vehicle. The method can also determine an occlusion caused by an elevation peak.

System and method for generating an occlusion map using crowdsourcing

A vehicle and a system method of operating the vehicle is disclosed. The system includes a monitoring module and a mitigation module operating on a processor. The monitoring module is configured to measure a degradation in an operation parameter of the vehicle, the vehicle operating in a first state based on a first value of a set of adaptive parameters. The mitigation module is configured to determine a threat to the vehicle due to operating the vehicle in the first state with the degradation in the operation parameter and adjust the set of adaptive parameters from the first value to a second value that mitigates the threat to the vehicle, wherein the processor operates the vehicle in a second state based on the second value.

Method and process for degradation mitigation in automated driving

An autonomous vehicle, system and method of navigating the autonomous vehicle. The system includes one or more sensors for obtaining data with respect to a remote stationary vehicle, and a processor. The processor is configured to classify the remote stationary vehicle into an object hypothesis based on the data, determine an actionable behavior of the autonomous vehicle based on a probability for the object hypothesis, and navigate the autonomous vehicle with respect to the remote vehicle via the actionable behavior.

Approach to maneuver planning for navigating around parked vehicles for autonomous driving

A process for sensor sharing for an autonomous lane change is provided. The process includes, within a dynamic controller of a host vehicle, monitoring sensors of the host vehicle, establishing communication between the host vehicle and a confederate vehicle on a same roadway as the host vehicle, monitoring sensors of the confederate vehicle, within the dynamic controller of the host vehicle, utilizing data from the sensors of the host vehicle and data from the sensors of the confederate vehicle to initiate a lane change maneuver for the host vehicle, and executing the lane change maneuver for the host vehicle.

Process and system for sensor sharing for an autonomous lane change

A system or method implemented by an autonomous vehicle involves determining a path plan to reach a destination from an origin. The path plan includes two or more path steps indicating tasks to be completed to reach the destination. The method includes, during traversal of the path plan by the autonomous vehicle, evaluating one or more of the two or more path steps of a planning horizon to determine a behavior plan for the planning horizon. The planning horizon is based on a current position of the autonomous vehicle, the behavior plan includes a speed and a trajectory, and the evaluating includes performing a cost analysis using a parallelized tree-based decision scheme at each of two or more simulation intervals within the planning horizon. The evaluating and the determining the behavior plan is repeated at two or more positions of the autonomous vehicle from the origin to the destination.

Parallelized tree-based decision scheme for autonomous vehicle

In various embodiments, methods, systems, and vehicles are provided for executing a lane change for a host vehicle. In various embodiments, one or more sensors obtain sensor data pertaining to target vehicles in proximity to the host vehicle; and a processor at least facilitates: obtaining, using the sensor data, predictions as to future positions and movement of the target vehicle; identifying a plurality of gaps through which the host vehicle may accomplish the lane change, based on the predictions; calculating a cost for each of the plurality of gaps; selecting, a selected gap of the plurality of gaps, having a minimized cost; and executing the lane change for the host vehicle via the selected gap.

Growing community of inventors

Troy, MI, United States of America

Sayyed Rouhollah Jafari Tafti