rdulmina/traffic-control-reinforcement-learing-agent

Deep Reinforcement Learning Agent to control traffic light providing emergency facilitation using real-time traffic data.

Minimize Traffic Congestion with Emergency Facilitation using DeepReinforcement Learning.

Abstract

In intelligent traffic light control, matrices derived from real-time traffic data are paramount for effi-ciency and performance. The rewards and state representations in previous studies could mislead a Reinforcement Learning agent in some cases. This paper examines the effectiveness of considering the Standard Deviation of vehicle’s Wait-ing Time (SDWT) on Deep Reinforcement Learning based traffic congestion control with emergency facilitation. The proposed method was selfevaluated by only considering average waiting time under both synthetic and Toronto real-world dataset. It has demonstrated that the proposed method was able to gain a significant impact on performance by considering the SDWT. More-over, the proposed method was able to reach zero waiting time for emergency vehicles.

Please run the Testing section of Agent.ipynb to run the agent

State representation Matrix M and Vector P

traffic-control-reinforcement-learing-agent's Issues

requirements

hi please what is the performance of your machine? And which version of TF and Keras do you use?

Paper

Hello
Thank you for sharing your code.

Could you provide the papers or tutorials that comes with this code?

do your code use gpu ?

I am encountering extended execution times while running reinforcement learning code on my system. Despite having an RTX 3060 GPU, 12 GB of RAM, and 64 GB of system memory, the processing time seems unusually high. I am seeking assistance in understanding the potential causes for this and exploring possible optimizations. noted i use the same parameter as you

action step and testing phase

hello sir i try your code but in training phase i get 1500 action onlly ow ever i didnt change anything and maximum waiting time was 140 in your code you have 12000 action and maximum waiting time for regular vehicle 140 too my qution why i didnt get same nbr of action.
anather quetion regarding testing phase it go worst with your model of DQN.h5 and it is not finishing how ever time of end similation is 50000 i dont konw why the waiting time become more and more biger am in 12 day on testing and nothing goes good how ever you mention that it stay onlly 60 s