Dynamic Taxi Service Planning by Minimizing Cruising Distance Without Passengers

It is a long-standing challenge to recommend routes to a group of taxis at different locations to minimize the overall mileage spent without customs. This paper studies route recommendation to a group of taxis by minimizing the overall mileage spent without customers. A new recommendation algorithm and evaluation model are proposed. The algorithm recommends the current optimal route and updates both the capacity and the probability along the route after getting a passenger. The new evaluation model is adapted to estimate the performance of each candidate route. A simulator that imitates the recommendation process on the real-world datasets and virtual taxis based on our recommendation system is developed. The experimental results demonstrate the effectiveness of the proposed evaluation model. The proposed method using the potential-cruising-distance model effectively reduces the global cruising distance of multiple taxis, especially in the applications with a large number taxis.