Key Technologies In Real-Time Taxi Sharing

With the extensive use of smart phones and GPS devices, applications for Location-based service (LBS) are more various, which bring people convenience. Users get services by sending their locations via mobile phones. By analyzing big trajectory data, service a-gencies can understand human behavior so that they can make services more effective and responsible. Since the amount of taxis can not satisfy the demand and taxis are improp-erly distributed, passengers are difficult to hail taxis and drivers can not find passengers easily. Taxi-sharing is an efficient way to improve the utility of taxis by allowing multiple passengers to share a taxi. In some extent, it also helps to relieve traffic jams and air pollu-tion. We research key technologies in real-time taxi sharing, including analyzing massive taxi trajectory data by distributed computing platform, proposing a taxi sharing form of rapid matching partners and another form of personalized matching occupied taxis. Main contributions of this paper are as follows:· Off-line and On-line Taxi Sharing Framework Devise a two-phase general framework including off-line phase and on-line phase to deal with taxi sharing prob-lem. During off-line phase, we utilize both big trajectory data and road network data to compute related statistics for on-line searching. Based on the general framework, we propose two forms of taxi sharing and also devise their respective architectures.· Taxi Trajectory Preprocessing and Distributed Computing Framework Pro-pose a framework for preprocessing taxi trajectory data and computing statistics. The preprocessing includes noise filtering, anomaly detection and road segregation. Distributed computing is applying Map-Reduce paradigm to extract four types of important features. The statistics is basis of on-line searching.· Quick Matching among Passengers Propose the first form of taxi sharing to match among passengers. For on-line searching, we take advantage of off-line statistics to improve accuracy and devise a quick matching algorithm to find passengers with lowest travel time cost. We also evaluate the efficiency of the algorithm by experi-ments and develop a prototype system to simulate the real scenario.· Personalized Matching between Passengers and Occupied Taxis Define the second form of real-time personalized taxi sharing with the consideration of each passenger’s preference. We first define the satisfaction degree of each party in-volved and two goals (named MaxMin and MaxSum). Then, we propose several pruning rules for these tow goals to filter taxis as many as possible. Thereafter, we conduct extensive experiments on real datasets to verify the effectiveness and efficiency of the proposed method.