Multi-agent Reinforcement Learning for Urban Projects Planning

Abstract

Actually, planners always need more and more updated plans to satisfy eventual changes. Those plans usually bring immediate and even sustainable solutions. According to the nature of problems, and decision-makers yearnings, the allocated time to establish plans is still tight. Therefore, adequate technics and methods are suitable to tackle this problem. Recently, a primer research field called Machine learning (ML), whose technics are based on learning by studying data or by applying known rules to categorize things, to predict outcomes, to identify patterns, or to detect unexpected behaviors. Reinforcement learning (RL) is an active research field of ML, based on learning how to map situations to actions, so as to maximize a numerical reward. By employing RL methods, the aim is to provide better plans for urban projects, wherein they are modeled to form a multi-agents system, acting cooperatively and optimally.