DEPARTMENT OF GEOMATICS THE UNIVERSITY OF MELBOURNE Trip Quality In Peer-To-Peer Shared Ride Systems

In a peer-to-peer shared ride system, transportation clients with traffic demand negotiate with transportation hosts offering shared ride services for ad-hoc ride sharing in a continuously changing environment, using wireless geosensor networks. Due to the distinctive characteristic of this system—a complex and nondeterministic transportation network, and a local peer-to-peer communication strategy—clients will always have limited transportation knowledge, both from a spatial and a temporal perspective. Clients hear only from nearby hosts, and they do not know the future availability of current or new hosts. Clients can plan optimal trips prior to departure according to their current knowledge, but it is unlikely that these trips will be final optimal trip due to continuously changing traffic conditions. Therefore, it is necessary to evaluate the trip quality in this dynamic environment in order to assess different communication and wayfinding strategies. The goal of this thesis is to develop an objective and systematic approach to accessing trip quality in simulations of peer-to-peer shared ride systems and uses this approach to evaluate different methods of peer-to-peer shared ride trip planning. By generating and analyzing a global database from the simulations for a large number of simulation runs, global singleor multi-criteria optimal trips can be identified in each simulation. These global optimal trips can be used for evaluating the quality of suboptimal trips realized by clients who apply some ad-hoc trip planning strategies in the simulation. The space-time network was introduced to compute global optimal trips by applying Dijkstra’s algorithm and choosing travel time as a criterion. Multi-criteria optimization was presented for computing multi-criteria optimal trips in simulations considering travel time and trip fares as cost functions. The developed approach is implemented in a trip quality evaluation subsystem, which can be integrated into different versions of the multi-agent simulation of peer-to-peer shared ride systems. The experiments are conducted to test the proposed approaches to evaluating trip quality. The experiment results verify the hypothesis of the thesis: the trip quality can be determined in peer-to-peer shared ride system simulations.