Evaluation of Anticipatory Decision-Making in Ride-Sharing Services

In recent years, innovative ride-sharing services have gained significant attention. Such services require dynamic decisions on the acceptance of arriving trip requests and vehicle routing to ensure the fulfillment of requests. Decision support for acceptance and routing must be made under uncertainty of future requests. In this paper, we highlight that state-of-the-art approaches focus on anticipatory decision-making for either acceptance or routing decisions. Our aim is to evaluate the potential of different levels of anticipation in ride-sharing services. Up to now, it is unclear how the value of information differs between none, partial, or fully anticipatory decision-making processes. To this end, we define and solve variants of the underlying dial-a-ride problem, which differ in the information available about future requests. Using a large neighborhood search, our experimental results demonstrate that ride-sharing services can highly benefit from anticipatory decision-making, while the favorable level of anticipation depends on particular characteristics of the service, esp. the demand-to-service ratio.