| Vehicle path planning has always been the research focus of intelligent transportation system. It is of great significance for easing growing urban traffic congestion, energy conservation, and the improving of people living efficiency. As the development of sensor, wireless communication, large amount of location data traces are easy to be collected., intelligent systems are developed to recommend optimal choice after mining these traces. In several fields, intelligent systems are developed to recommend optimal choice after mining these traces. It is so called recommender system. In recent years, recommender systems are increasingly advocated in transportation,leisure and business applicationsThis paper presents an online traffic system to recommend taxi drivers the fastest-path of picking passengers up. Simulation and information processing is one of the difficulties of our work, we collect GPS data from more than 4,000 taxi GPS through the Shanghai Dazhong Taxi Company. We build the Shanghai urban vehicle network (SUVnet) through the map-matching algorithm, path planning algorithm , real-time road segments speed calculating . Analyzing a real trace data including about 2000 taxis in a 22 square kilometers area in Shanghai, we get several hot terminals by k-means clustering taxi pick-up points, which have high probability of picking up passengers. With the online velocity measurement on each road, we provide a Potential Travel Time function (PPT)for evaluating each path of hot terminals sequence. Based on the monotone property of this function, the search space can be efficiently pruned. Then, we design a practical recommendation system to process the fastest-path selection. We use real trace to evaluate our recommender system. Experimental results show that our online system can quickly recommend the almost exact fastest-paths to taxi drivers for picking up passengers in real traces. At last, we assessed the effect of some pre-define parameters to our recommender system. |
PDF Full Text Request