A Q-Learning-Based Approximate Solving Algorithm for Vehicular Route Game

Route game is recognized as an effective method to alleviate Braess’ paradox, which generates a new traffic congestion since numerous vehicles obey the same guidance from selfish route (such Google Maps). The conventional games have symmetry vehicles’ payoffs depend only on selected distribution but not who chose, leads precise Nash equilibrium being able be solved by constructing special potential function. However, with arrival of smart cities, real-time schemes more concerned engineers than absolute optimality in real traffic. It easy task re-construct functions due dynamic conditions. In this paper, compared hard-solvable function-based method, matched Q-learning algorithm designed generate approximate classic for An experimental study shows that coefficients generated Q-learning-based solving all converge 1.00, and still required convergence different parameters.