| CPC B60W 60/001 (2020.02) [B60W 40/02 (2013.01); G01C 21/3461 (2013.01); G05B 13/027 (2013.01); G06N 3/045 (2023.01); G06N 3/0464 (2023.01); B60W 2554/00 (2020.02); B60W 2556/40 (2020.02); G08G 1/22 (2013.01)] | 14 Claims |
|
1. A method, comprising:
obtaining a map of an environment where a vehicle operates, the environment comprising a restricted traffic zone;
obtaining information of one or more objects in the restricted traffic zone; obtaining a temporal sequence of semantic grids of the restricted traffic zone, each semantic grid comprising information of at least part of the restricted traffic zone at a time;
generating an input dataset, the input dataset comprising the map of the environment, the information of one or more objects, and the temporal sequence of semantic grids;
inputting the input dataset into a neural network, the neural network generating an output that indicates one or more edges of the restricted traffic zone, wherein the output of the neural network further includes one or more polylines that divide the environment into a plurality of regions;
wherein the neural network comprises a convolutional neural network and a graph neural network;
wherein inputting the input dataset into the neural network comprises:
inputting the map of the environment and the information of the one or more objects into the graph neural network; and
inputting the temporal sequence of semantic grids into the convolutional neural network;
planning a trajectory of the vehicle through at least part of the environment based on the output of the neural network; and
causing the vehicle to maneuver through the at least part of the environment based on the trajectory, the maneuvering comprising at least one of controlling an engine throttle, controlling a motor speed, controlling brakes of the vehicle, or steering the vehicle.
|