A Mixed Breadth-Depth First Search Strategy for Sequenced Group Trip Planning Queries

We study Sequenced Group Trip Planning Queries (SGTPQs). Consider a road network where some vertices represent Points of interest (POIs) and each POI belongs to exactly one Category of Interest (COI), e.g., A COI can be "Restaurants" and each POI in this COI is a specific instance of a restaurant. Given a group of users, each starting from a (possibly distinct) source location and going ultimately to a (possibly distinct) destination, as well as a ordered sequence of COIs, the SGTPQ finds, for each user, the route from his/her source location to his/her destination such that all users go through the same POIs, each one belonging to the specified sequence of COIs, while minimizing the total distance travelled by all users in the group. Different from previous work which investigated SGTPQs in Euclidean distance, we focus on SGTPQs in the more realistic case of road networks. The only existing algorithm for processing SGTPQs which may be also applicable in road networks, named IA, suffers from two drawbacks: it is not able to produce optimal answers and is computationally expensive. The first contribution of this paper is a small modification to IA, so that it can provide optimal solutions. The second and main contribution is a new approach, called Progressive Group Neighbour Exploration (PGNE) that delivers the optimal solution while being more efficient than IA. Our extensive experiments based on real and synthetic datasets show that PGNE is always faster than the modified IA and, in particular, typically twice as fast with respect to the number of users in the group travelling together, an important parameter for this type of query.