Dynamic Pricing Decision in a Duopolistic Retailing Market

Under the uncertain demand and variable environment, we studied the enterprises' dynamic decision problem on price strategies in duopolistic retailing market. In the game, two enterprises simultaneously choose their strategic variable in each period to maximize their expect revenue. We use Markov decision processes to build up the model and resolve it, and design two kinds of reinforcement learning methods which are named Nash Q-learning and Best-response Q-learning to simulate the model. Through the numerical study, we draw a conclusion that compared to the Nash Q-learning method the Best-response Q-learning is a better method to give a dynamic pricing decision in duopolistic retailing market