.png){kind=icon}
.png){kind=icon}
Return
A multi-agent deep distribution approximation strategy optimization algorithm with multi-threaded parallel computing mechanism suitable for large-scale and complex urban road networks
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.110999.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In order to alleviate the traffic congestion of urban road networks, a multi-agent hierarchical heterogeneous control framework with cooperation mechanism is proposed in this paper. Secondly, the maximum vehicle delay is introduced as the road congestion level judgment standard, and the road congestion level is discretely divided. Thirdly, a multi-agent joint reward mechanism is designed to enhance the multi-agent collaboration ability and improve the cumulative income of the system. Finally, a multi-agent deep distribution approximation strategy optimization algorithm based on the regional adaptive green wave control mechanism (RAG-MADPPO) is proposed to solve the problems of the long vehicle waiting time in the process of largescale traffic signal control. Moreover, the multi-threaded parallel computing strategy is proposed to improve the training efficiency of the RAG-MADPPO algorithm. In addition to testing the different strategies on the Manhattan road network scenario, the RAG-MADPPO algorithm is applied to the Oakland road network to verify its effectiveness further. The results show that the proposed RAG-MADPPO algorithm has significant advantages in evaluation indicators and model performance.
Keywords:
Multi-agent reinforcement learning
Adaptive traffic signal control
Intelligent transportation system
Large-scale and complex scenario
Journal
IF:
8
Papers:
5.3K
Citations:
3.5W