Multiagent Reinforcement Learning for Integrated Network: Applying to a Part of the Road Network of Krasnoyarsk City

Abstract

The article examines a mathematical model of the selecting phases process of traffic light facilities of the road network section. A Markov decision process with a finite number of actions and states is used as a mathematical model, and the minimization problem is reduced to the Multiagent Reinforcement Learning for Integrated Network (MARLIN) problem. A Q-learning algorithm was implemented and a series of computational experiments were conducted in the Anylogic simulation system for a real section of the Krasnoyarsk road network to study the model.