Smart Taxi Dispatch System Based On Real Traces

In daily life nowadays, taking taxis is becoming an importantchoice for citizens to travel around urban areas. Generally, taxi driversdrive their vehicles to find passengers based on personal experience or justrandomly along roads. However, due to the lack of passenger knowledge,this travelling is inefficient on economy, route and time. Therefore, it isimportant to design a dispatch system for taxis. However, existed taxidispatch systems have some drawbacks, leading to inefficiency of taxidispatch. In this paper, we introduce a novel taxi dispatch system, whichoffers routes for vacant taxis to find passengers more quickly. In oursystem, firstly, establishing a dynamical passenger appearance probabilitymodel based on Poisson process according to historical data; then, basedon this model, designing two dispatch algorithms, which both consider theimpact between different taxis and schedule effective routes for taxis tofind passengers more quickly; moreover, we build an effective feedbackmechanism by analyzing passenger appearance model, which enhances theaccuracy of predicting the probability of taxis encountering passengers andmeanwhile reduce the computing complexity. Besides, we probe intomethods of collecting vehicle travelling data and methods of recoveringreal traces of vehicles.