Abstract:
Ridesharing has become a popular model that enables users to share their rides with others in recent years. Traditional ridesharing services arrange ridesharing trips to travel between a xed source and a xed destination locations. Users need to visit point of interests (POIs) such as a supermarket or a pharmacy for performing various daily activities while traveling between xed locations like an o ce and a home. In current ridesharing services, there is no guarantee that a user gets ridesharing options for the complete trip, e.g., for visiting from the o ce to a POI and then from a POI to the home. Again, the exibility in visiting a POI a little bit far away instead of visiting the nearest POI with respect to xed locations may increase the probability in getting ridesharing services. In this thesis, we introduce a novel type of ridesharing query, an Activity-aware Ridesharing Group Trip Planning (ARGTP) query that exhibit three novel features: (i) ensures a complete trip for visiting more than two locations, (ii) allows to visit both xed and exible locations, and (iii) provides true ridesharing services instead of a taxi like ridesourcing services by matching a group of riders' exible trips with a driver's xed trip. An ARGTP query considers the spatial proximity of the trips of the riders with that of a driver, and returns an optimal ridesharing group that minimizes the total cost of the ridesharing group. We develop the rst solution to process ARGTP queries in real time. The e ciency of the ARGTP query processing algorithms depends on the number of candidate riders and the number of POIs to be explored. We introduce novel pruning techniques to prune the riders and re ne the POI search space. We perform extensive experiments using both real and synthetic datasets to validate the e ciency and e ectiveness of our approach, and show that our approach outperforms a baseline approach with a large margin.
