Towards sustainable ride-pooling algorithms for autonomous cars: a comparative study of passenger satisfaction, taxi fleet usage and emission metrics

Abstract

Reducing traffic-related CO2 emissions is one of the major factors in achieving city-level carbon neutrality. In this work, we study the problem of sustainable ride-pooling of autonomous cars with SUMO traffic simulation software as an environment for testing ride-pooling strategies and explicit emission modelling. We compare the scenarios without ride-pooling (baseline) and with ride-pooling for varying levels of demand, maximum occupancy of cars and penetration rates of autonomous taxis using three groups of metrics, namely, metrics of passenger satisfaction, taxi fleet usage and total emissions. The experimental study showed that simple heuristics (minimizing detours or maximizing the occupancy of the cars) may reduce emissions up to 15% compared to baseline but do not provide the balanced solution for multiple metrics. To overcome this problem, the future work will contain development of simulation-based multi-objective reinforcement learning algorithm for sustainable ride-pooling.