Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Preferences-Aware social ridesharing

Pagkalos Emmanouil

Full record

---

Summary

Ridesharing is a shared-economy transportation paradigm, in which people can catch rides in vehicles that are (possibly) privately owned and driven by others, in order to reduce transportation costs and possibly enjoy a more pleasant ride when compared to those offered by alternative means of transportation. Despite its touted advantages, e.g. in terms of increased socialization and potentially huge positive environmental impact, Ridesharing has not gained much popularity yet. Arguably, for this to happen Ridesharing needs to offer a clearly more-pleasant-than-others transportation experience to people who use it. In this thesis, we approach the Ridesharing problem via Artificial Intelligence solution concepts originating primarily in the Multiagent Systems, graph-theoretic, and game-theoretic research literature; and, importantly, we take into consideration the riders’ (agents) preferences about attributes of their co-riders. Taking into account such preferences, bears the potential to greatly impact positively the participants’ satisfaction. Our thesis offers an initial but complete framework for preferences-aware Ridesharing. We first employ the concept of a hypergraph to the set of agents, in order to create an initial clustering given the areas where they move. We then use a greedy algorithm and a branch and bound algorithm in order to distribute the agents into vehicles with the aim of (a) maintaining the drivers’ detours to as low levels as possible; and, at the same time, (b) satisfying most of the expressed agents’ preferences. Furthermore, we put forward a cost-sharing scheme to compensate drivers for their participation in the scheme and the extra costs this entails, and to create incentives for participation via overall drivers’ costs reduction. We studied systematically the performance of our Ridesharing framework via simulation scenarios run on maps of the four main cities of Crete: Heraklion, Chania, Rethymno, and Agios Nikolaos. Our results show that (a) our approach has high effectiveness in terms of covering the transportation needs of the (non-driving) commuters; (b) at the same time, the average extra distance that drivers need to cover when offering their services to passengers is kept to acceptable levels; (c) the costs of the drivers are substantially reduced when compared to their costs when not participating in Ridesharing; and (d) the agents’ preferences are satisfied to a large degree, a fact which arguably improves their whole Ridesharing experience. Notably, our simulation results are particularly encouraging for scenarios with much-lower-than-current vehicle ownership (drivers’ percentage), thus underscoring the potential of Ridesharing in a future, greener, largely free of privately owned vehicles world.

Available Files

Services