Dynamic Taxi Ride-Sharing Through Adaptive Request Propagation Using Regional Taxi Demand and Supply

Abstract

Taxi ride-sharing is an emerging public transportation model that provides several benefits in terms of cost, environmental impact, and road congestion. It is further popularized through market available app-based systems, such as Uber, Lyft, Didi, etc. However, those systems are limited due to their centralized architecture, high cost sharing with the driver, and proprietary business model. Distributed ride-sharing, on the other hand, involves only passengers and drivers and operates in a peer-to-peer manner. But, distributed ride-sharing systems often suffer due to Spatio-temporal constraints associated with taxi demand and supply as well as broadcast message storms. While we have observed dynamic and distributed ride-sharing systems which address the Spatio-temporal issues, there is hardly any effort to reduce their message overhead. In this paper, we present a hybrid model of ride-sharing where a central server adaptively calculates transmission range for passenger request propagation using Spatio-temporal information of ride-sharing success rate for the past 30-minute. Passengers use the adaptive transmission range to find the best shared-ride using a distributed manner. Our extensive empirical evaluation shows that our proposed approach increases the overall ride-sharing success rate and taxi utilization while significantly reducing the communication overhead, request processing time, and passenger waiting time.