Reinforcement Learning Intersection Controller

Exploring cutting-edge solutions for optimizing traffic control, a fundamental challenge in modern society, using the power of reinforcement learning algorithms

During my bachelor studies, I investigated the following research questions:

Can a single road intersection be effectively managed by a system based on reinforcement learning?
What would be the optimal choice of a model and a reward function to ensure the minimal congestion at the road intersection at all times?
What kind of real-time information would be enough for a model to learn the optimal behavior patterns?
How can we test this approach? Would a simulator be a good enough proxy to the real-world scenarios?
How can we extend this approach to a system of multiple road intersections?

References

Books & Conferences

2018

IEEEXplore

Reinforcement Learning Intersection Controller

Gulnur Tolebi, Nurlan S. Dairbekov, Daniyar Kurmankhojayev, and Ravil Mussabayev

Abs HTML Code

This paper presents an online model-free adaptive traffic signal controller for an isolated intersection using a Reinforcement Learning (RL) approach. We base our solution on the Q-learning algorithm with action-value approximation. In contrast with other studies in the field, we use the queue length in adddition to the average delay as a measure of performance. Also, the number of queuing vehicles and the green phase duration in four directions are aggregated to represent a state. The duration of phases is a precise value for the nonconflicting directions. Therefore, cycle length is non-fixed. Finally, we analyze and update the equilibrium and queue reduction terms in our previous equation of an immediate reward. Also, the delay based reward is tested in the given control system. The performance of the proposed method is compared with an optimal symmetric fixed signal plan.