کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Cooperative Coverage Path Planning Using Q-Learning and Sarsa in Two Environments

نویسندگان :

Alireza Nezamzadeh¹ Hamed Jalaly Bidgoly² Marzieh Kamali³

1- Isfahan University of Technology 2- Isfahan University of Technology 3- Isfahan University of Technology

کلمات کلیدی :

Multi-Agent system،Reinforcement Learning،Cooperative Control،Coverage Environment

چکیده :

This paper presents the coverage environment of two agents using reinforcement learning. When we talk about reinforcement learning, two popular algorithms, Q-learning and Sarsa, come up. One of the differences between these algorithms refers to the on-policy and off-policy methods for updating the rules of the algorithms. In this paper, we consider two environments in order to investigate cooperative coverage path planning by using these algorithms. Our goal is coverage environment with the cooperation of agents based on reducing the energy of the agents. At first, we consider the environment without an obstacle, then is considered an obstacle for that. The proposed algorithm expresses sharing experience between agents when they want to cover the environment with an obstacle and compares that in convergence speed with when there was no sharing experience between agents. The performance of the proposed algorithm is evaluated by various simulations.