On-demand mobility services are envisioned to extend the range of traditional public transport offers. They promise to provide flexible, efficient and finely tuned transportation by continuously adapting to user demand and the traffic situation. Deploying such services using CCAM vehicles further increases their potential in terms providing mobility and access 24h a day. However, when drivers are taken out of the equation, the vehicle movements need to be planned computationally. For that purpose, fleet dispatching algorithms are used. Such dispatching algorithms are in charge of deciding which available vehicle should be assigned to which waiting user. Some user requests may be delayed or even rejected if there are not sufficiently many available vehicles to satisfy the demand. The design of dispatch algorithms is a major area of research where techniques from optimization and artificial intelligence are employed to achieve performant strategies.
Mobility services adapted for heterogeneous demand
Most of the dispatch algorithms proposed in the literature are designed to minimize overall wait and travel times perceived by the users. They are evaluated considering a homogenous fleet and a homogenous set of users, where everyone presents the same characteristics and interacts similarly with the service.
However, these general optimizations can present issues in a real-life context where different (vulnerable) user-groups can have different needs when interacting with the service. For instance, a situation may arise where multiple requests are pending and one of them is emitted from a user that requires extra time to enter the vehicle. Given limited excess capacity at that time of day, only one traveller’s trip can be satisfied. A dispatch algorithm that strictly aims at minimizing travel times will then most likely discriminate against such users.
Within SINFONICA, IRT SystemX aims to shed light on this issue by evaluating existing algorithms on heterogenous travel demand. We explicitly take into consideration requests from vulnerable users and design services that address their needs. We rely on highly detailed simulations where each traveller is represented individually. This allows us to evaluate the service performance (wait and travel times, rejections) as perceived by each user-group.
After identifying where existing algorithms fail to consider all people, we investigate ways to mitigate these issues.
Keywords: CCAM fleet services, heterogeneous demand, operational strategies.
Author: Tarek Chouaki (IRT-SystemX)
Picture from Freepik
