Multi-agent broad reinforcement learning for intelligent traffic light control

Intelligent traffic light control (ITLC) aims to relieve congestion. Some multi-agent deep reinforcement learning (MADRL) algorithms have been proposed for ITLC, and most of them use neural networks make decisions. However, the abundant parameters structure lead time-consuming training process MADRL. Recently, a broad (BRL) approach has improve efficiency an agent. Unlike MADRL that architecture, BRL utilizes architecture. In this paper, we propose (MABRL) algorithm ITLC. The MABRL adopts network joint information updates using ridge regression. To increase effectiveness interaction among agents, design dynamic mechanism (DIM) based on attention mechanism. DIM enables agents aggregate particular intersections at appropriate moments. We conduct experiments three different datasets. results demonstrate outperforms several state-of-the-art in alleviating congestion with shorter time.