Understanding the impact of heterogeneous rider preferences on a shared autonomous vehicle system

In a ride pooling system, riders may have varied behaviors in seeking pooled or non-pooled rides. It is important to understand the effect of these rider behaviors on the system performance in order to formulate policies to guide ride pooling implementation. Existing literature modeling ride pooling systems using agent-based models only considers the extreme cases in which riders either all participate or not participate in pooling. However, the pooling behaviors could be more complex. This study segments the rides in the system into five types (non-pooling only, non-pooling preferred, indifferent, pooling preferred, and pooling only). We use an agent-based model to simulate these preferences in a system of pooled autonomous vehicles. We use mixture experiments to vary the proportion of riders within these five types and build models to study the interactions among the rider types in terms of the system’s service quality and environmental performance. The results show that higher service level is achieved when all riders in the system are open to pooling, with 30% of pooling only riders and 70% of pooling preferred or indifferent riders providing the maximum value. The results can help formulate incentives and policies to promote ride pooling participation to improve ride pooling system performance.