A machine learning-driven two-phase metaheuristic for autonomous ridesharing operations

This paper contributes to the intersection of operations research and machine learning in the context of autonomous ridesharing. In this work, autonomous ridesharing operations are reproduced through an event-based simulation approach and are modeled as a sequence of static subproblems to be optimized. The optimization framework consists of a novel data -driven metaheuristic within a two phase approach. The first phase consists of a greedy insertion heuristic that assigns new online requests to vehicles. The second phase consists of a local-search based metaheuristic that iteratively revisits previously-made vehicle-trip assignments through intra-and inter-vehicle route exchanges. These exchanges are performed by selecting from a pool of destroy-repair operators using a machine learning approach that is trained offline on a large dataset composed of more than one and a half million examples of previously-solved autonomous ridesharing subproblems.Computational results are performed on multiple dynamic instances extracted from real ridesharing data published by Uber Technologies Inc. Results show that the proposed machine learning-based optimization approach outperforms benchmark state-of-the-art data-driven meta -heuristics by up to about nine percent, on average. Managerial insights highlight the correlation between selected vehicle routing features and the performance of the metaheuristics in the context of autonomous ridesharing operations.