Ant-Taxi to Pie-Passenger: Optimizing Routes and Time for Distributed Taxi Ride Sharing

Sustainability of local co-operative taxi networks in the face of competition from corporate behemoths is a problem worthy of investigation. Lack of infrastructure background support and venture capital demands alternative strategies for survival. A viable solution requires novel modeling approach and computational framework. Such framework needs to be fully distributed and therefore is sub-optimal. We propose an intra-city taxi service that leverages individual rides and local network for ride management. This, in turn, demands a bunch of decision variables and objectives to be optimized. Some of these objectives are conflicting and therefore the problem requires a multi objective optimization formulation. We propose to exploit the flexible structure and multi-agent like behavior of Ant colony optimization to tackle the diversity of objectives in the ride share problem. We consider the scenario of ride share only since that is the hardest part in the taxi business. The implementation is multi-modular and provides an alternative approach to commonly used centralized taxi network. We define novel performance metrics and present results to supplement the modeling approach. Our results show that spatio-temporal diversity is the biggest road block to wide scale ride sharing. Nonetheless, our solution ensures up to 77% acceptance in shared rides.