Taxi scheduling research based on Q-learning

Taxi, as one of the means of transportation in the city, are time-saving, comfortable, convenient and flexible. However, under the prevalence of taxis, there are still some problems. For example, the inability to realize real-time communication between the driver and potential passengers, resulting in the frequent occurrence of a large number of taxis in some places without load while passengers in another place have no taxis to take, which cannot well alleviate the urban traffic pressure. Consequently, in this paper, based on the driver's interests under the perspective of using Q-learning algorithm and Markov Decision process and through modeling to give drivers carrying passengers three options at any time and place: location, empty and waiting. Taxi drivers can learn the best form strategy from driving trajectory to increase income and improve the efficiency of taxi service as the goal. From the perspective of the driver, this research further solves the problem of the dynamic dispatch of taxis, and can also provide reference and technical support for the operation and management of taxis.