Unmanned driving hierarchical motion decision control method based on deep reinforcement learning

Abstract

The invention discloses an unmanned driving hierarchical motion decision control method based on deep reinforcement learning. The control method is realized by taking a meta motion decision-vehicle control hierarchical motion decision control model as a carrier. Through abstract decomposition of driving behaviors and analysis of environmental factors influencing the driving behaviors, a motion decision control process is decomposed into a mode of meta-motion decision-vehicle control, the meta-motion decision belongs to a discrete decision problem, and an end-to-end model from dynamic driving information to element actions is established by using a DQN deep reinforcement learning algorithm. Vehicle control belongs to continuous action output, a DDPG deep reinforcement learning algorithm is adopted to establish an end-to-end model which maps road information and element actions to accelerator, brake and steering wheel control quantities, a PyTorch deep learning framework is used for establishment of a neural network, a selected development language is Python, and the model outputs the control quantity of the vehicle by receiving the driving behavior instruction and combining the environment state information.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, the field of unmanned driving and vehicle control technology, in particular to an unmanned vehicle motion control scheme implemented based on deep reinforcement learning algorithms in scenes requiring interactive and coordinated driving of multiple traffic participants such as urban roads. Background technique [0002] The rapid development of the automobile industry has made automobiles an indispensable means of transportation for people to travel. The subsequent urban road traffic safety, traffic congestion, traffic pollution, and energy consumption are important issues facing the development of urban transportation today. In recent years, the sales volume of automobiles has maintained the first place in the world, and the number of automobiles has increased rapidly. These social problems mentioned above are particularly prominent. The application of unmanned vehicles is an importa...

AI Technical Summary

Problems solved by technology

[0005] The main purpose of the present invention is to propose an unmanned vehicle motion decision-making control scheme for dynamic driving scenarios, aiming to solve the problem of completing driving behavior based on the end-to-end motion decision-making control model based on learning, so that unmanned vehicles can fully consider the road structure and dynamic traffic participants and other environmental information to realize vehicle behavior changes

Images