Multi-agent behavior decision-making method and device, electronic device and storage medium

Abstract

The invention provides a multi-agent behavior decision method and device, an electronic device and a storage medium, and the method comprises the steps: constructing each agent and corresponding environment information into a graph based on a graph generation module in a multi-agent behavior model; based on an information transmission module in the multi-agent behavior model, encoding each agent and the corresponding environment information to obtain a joint encoding state corresponding to each agent; based on a strategy optimization module in the multi-agent behavior model, determining an initial decision of each agent in combination with the joint coding state corresponding to each agent; and based on a collision avoidance module in the multi-agent behavior model, performing variable step size control on the initial decision of each agent, and determining a final decision of each agent in combination with the repulsive force corresponding to each agent. According to the method, the problem that reinforcement learning is difficult to converge in a large-scale agent scene is solved, and high-performance autonomous decision-making ability and collision avoidance effect in a multi-agent system are realized.

Description

technical field [0001] The invention relates to the technical field of swarm intelligence, in particular to a multi-agent behavior decision-making method, device, electronic equipment and storage medium. Background technique [0002] Research has found that gregarious organisms such as "bee colonies" can emerge macroscopically intelligent behaviors through the coordinated behavior of many scattered individuals, forming complex group intelligence, which is a higher-level performance beyond individual behavior. With the development of emerging technologies such as autonomous control, the use of large-scale (homogeneous / heterogeneous) multi-agent "swarm" system swarm intelligence collaborative behavior has become a new way to solve major needs such as smart security, emergency rescue, and smart logistics. . [0003] The future development of the current "swarm" system can be roughly divided into three directions: [0004] (1) Quantitative development: to achieve a larger scal...

AI Technical Summary

Problems solved by technology

[0009] The invention provides a multi-agent behavior decision-making method, device, electronic equipment and storage medium to solve the defect of unstable learning of large-scale intelligent agents in the prior art

Images