Multi-agent path planning method based on deep reinforcement learning

Abstract

The invention discloses a multi-agent path planning method based on deep reinforcement learning. The method is a distributed path planning method, and comprises the following steps: inputting local observation information of a single agent into a neural network, transmitting information among the agents by using the graph neural network, training an approximate strategy function of the neural network, and outputting a movement strategy. Neural network parameters are trained by using a method of combining deep reinforcement learning and imitation learning, so that the convergence of a return function is faster. After training, a higher group path planning success rate in a four-neighborhood 2D grid map can be realized under the scale of thousands of agents, namely, a collision-free route from a starting point to an ending point is successfully planned for each agent within time limitation. The adaptability to the change of the map size and the barrier density is high.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a multi-agent path planning based on deep reinforcement learning. Background technique [0002] Multi-agent path planning is a problem of finding a set of non-conflicting paths for multiple agents from the starting position to the target position, while achieving optimal constraints: such as minimizing the sum of paths or action costs of all agents , throughput maximization, etc. Research on this problem has a large number of application scenarios in logistics, unmanned vehicles, military, security, games and other fields. [0003] There are many traditional algorithms for path planning of a single agent at home and abroad, such as A* algorithm, particle swarm algorithm, genetic algorithm, ant colony algorithm, simulated annealing algorithm, etc. With the improvement of industry and living standards, the work of a single agent often cannot meet the n...

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned deficiencies existing at present, the technical problem to be solved by the present invention lies in the lack of communication between agents in the prior art; the poor adaptability in the case of changing maps; Slow convergence of returns and slow training process due to framework design

Images