Waiting or Moving? A Crossroad Network-Based Markov Decision Process Approach to Catch Vacant Taxis

Abstract

Taxi services play a critical role in the public transportation system in our cities. However, we usually find it difficult to catch vacant taxis based on our experience alone in the random taxi-waiting mode, especially on the streets unfamiliar to us, which may greatly influence users' taxi service experience. Therefore, how to recommend appropriate waiting locations for passengers becomes meaningful, and the available large-scale taxi trajectory data have helped with the right recommendation. Recent researches have focused on the one-location recommendation for the passengers without considering the recommendation failure situation where they find no vacant taxis available after waiting a long time. In response to this deficiency, we designed a Crossroad Network-based Markov Decision Process (CN-MDP) scheme to recommend a waiting location sequence for a current passenger whose cumulative probability of catching a vacant taxi getting close to 100%. Further, our scheme changes the recommended locations from the road segments to the crossroads, as we discovered that passengers are more likely to catch vacant taxis at the crossroads connected to multi-road segments. In addition, the multi-passengers competing strategy for vacant taxis at the same location and in the same time slot is also involved in our scheme by dynamically updating the pass rate of vacant taxis at each crossroad and road segment. Some evaluations on a real taxi data set from a major city in China have shown that our recommendation scheme works well and has a higher probability of catching vacant taxis than that of our previous approach and other ones. Our scheme further improves the user experience of taxi services. INDEX TERMS Markov decision process, waiting location sequence, pass rate of vacant taxis, multi-passengers competition, taxi services.

Towards the significance of taxi recommender systems in smart cities

“…20,37,200 • More concentration is required on taxi service behaviors with the human decision-making process. 214,225,245,246 • Majorly publication of taxis is taking GPS based, but more concentration is needed to utilize cloud computing, RFID, and other devices in cross-domain. 211,214,247 • Security is still the issue in taxi recommendations such as fraud, other routes, payment, extra fuel charges done by passengers and taxi drivers are still yet to solve.…”

“…The outcomes of this article highlighted the crucial suggestions and future work: Based on published papers, taxi recommendation publications are still expanding and trending research areas 20,37,200 More concentration is required on taxi service behaviors with the human decision‐making process 214,225,245,246 Majorly publication of taxis is taking GPS based, but more concentration is needed to utilize cloud computing, RFID, and other devices in cross‐domain 211,214,247 …”

A Spatio-Temporal Schedule-Based Neural Network for Urban Taxi Waiting Time Prediction

“…Elsewhere, the XGBoost algorithm is applied to predict static waiting times of taxis [18]. A crossroad network-based Markov decision process scheme is suggested to recommend taxi waiting place [19]. Recently, the importance of spatio-temporal features in traffic flow predictions is receiving more and more attention.…”