Independent Reinforcement Learning for Weakly Cooperative Multiagent Traffic Control Problem

Abstract

The adaptive traffic signal control (ATSC) problem can be modeled as a multiagent cooperative game among urban intersections, where intersections cooperate to optimize their common goal, i.e., the city's traffic conditions. The large scale of intersections in a real traffic scenario yield marked challenges for an algorithm to find an optimal joint control strategy by controlling multiple intersections at the same time. Recently, reinforcement learning (RL) has achieved marked successes in managing sequential decision making problems, which motivates us to apply RL in the traffic control problem. In particular, independent reinforcement learning (IRL) is typically used to solve a multiagent task, where IRL agents learn cooperation strategies independently according to specific rules, such as the 'optimistic' or 'lenient' principle, and treat other agents as part of the environment. Considering the large scale of intersections in an urban traffic environment, we use IRL to solve a complex traffic cooperative control problem in this study. One of the largest challenges of this problem is that the observation information of intersection is typically partially observable, which limits the learning performance of IRL algorithms. To mitigate this challenge, we model the traffic control problem as a partially observable weak cooperative traffic model (PO-WCTM) to optimize the overall traffic situation of a group of intersections. Different from a traditional IRL task that averages the returns of all agents in fully cooperative games, the learning goal of each intersection in PO-WCTM is to reduce the cooperative difficulty of learning, which is also consistent with the traffic environment hypothesis. To determine the optimal cooperative strategy of PO-WCTM, we also propose an IRL algorithm called Cooperative Important Lenient Double DQN (CIL-DDQN), which extends Double DQN (DDQN) algorithm using two mechanisms: the forgetful experience mechanism and the lenient weight training mechanism. The former mechanism decreases the importance of experiences stored in the experience reply buffer, which deals with the problem of experience failure caused by the strategy change of other agents. The latter mechanism increases the weight experiences with high estimation and 'leniently' trains the DDQN neural network, which improves the probability of the selection of cooperative joint strategies. Experimental results show that, in two real traffic scenarios and one simulated traffic scenario, CIL-DDQN outperforms other methods in almost all performance indicators of the traffic control problem.