AGV path planning method and system based on ant colony algorithm and multi-agent q learning

Abstract

The invention discloses an AGV path planning method and system based on ant colony algorithm and multi-agent Q learning, which improves the ability of global optimization, and introduces multi-agent Q learning into AGV path planning research to realize AGV in interaction Learn how to avoid obstacles, and better utilize the autonomy and learning ability of AGV. The technical solution is: according to the static environment, use the grid method to model the AGV operating environment, set the starting point and target point; generate the global optimal path through the ant colony algorithm according to the starting point and target point coordinates of the AGV; AGV according to the global optimization The path moves to the target point. When a dynamic obstacle is detected within the minimum distance, the multi-agent Q learning corresponds to the environment state to select the obstacle avoidance strategy, and then make the corresponding obstacle avoidance action. After the obstacle avoidance is completed, return to the original Continue on the path.

Description

technical field [0001] The invention relates to an AGV path planning technology, in particular to an AGV (Automated Guided Vehicle) path planning method and system based on an improved ant colony algorithm and optimized multi-agent Q learning in a dynamic environment. Background technique [0002] Path planning is an important issue in AGV planning and scheduling. As more and more AGVs are put into use, many new problems have been brought about, such as the establishment of dynamically changing environmental models; the path planning of multiple AGVs and the problem of obstacle avoidance. ; The problem of the learning and intelligence of the multi-AGV system. The existing AGV path planning methods include swarm intelligent bionic algorithm, A*, D* algorithm, etc. Such a single global planning method often cannot achieve the global optimum, and requires high prior knowledge of the environment and requires a large When encountering a complex and dynamic environment, the effic...

AI Technical Summary

Problems solved by technology

Among them, the Q-learning algorithm, as a reinforcement learning algorithm, learns from the environment state to the action mapping. The agent adopts the optimal strategy according to the maximum reward value. The principle of the Q-learning algorithm is easy to understand, easy to combine with reality, and suitable for use in unknown environments. AGV path planning research, but there are more than one agent in a multi-agent system, so when applying Q-learning, it needs to be different from single-agent Q-learning. Multi-agents need to consider the decisions and influences of other agents. When the environment is complex Unknown, when the number of agents is too large, the dimension of the entire state space will increase rapidly, resulting in the disaster of dimensionality

Images