A Predictive Chance Constraint Rebalancing Approach to Mobility-on-Demand Services

Preprint, 2022

This paper considers the problem of supply-demand imbalances in Autonomous Mobility-on-Demand systems (AMoD) where demand uncertainty compromises both the service provider's and the customer objectives. The key idea is to include estimated stochastic travel demand patterns into receding horizon AMoD optimization problems. More precisely, we first estimate passenger demand using Gaussian Process Regression (GPR). GPR provides demand uncertainty bounds for time pattern prediction. Second, we integrate demand predictions with uncertainty bounds into a receding horizon AMoD optimization. In order to guarantee constraint satisfaction in the above optimization under estimated stochastic demand prediction, we employ a probabilistic constraining method with user defined confidence interval. Receding horizon AMoD optimization with probabilistic constraints thereby calls for Chance Constrained Model Predictive Control (CCMPC). The benefit of the proposed method is twofold. First, travel demand uncertainty prediction from data can naturally be embedded into AMoD optimization. Second, CCMPC can further be relaxed into a Mixed-Integer-Linear-Program (MILP) that can efficiently be solved. We show, through high-fidelity transportation simulation, that by tuning the confidence bound on the chance constraint close to "optimal" oracle performance can be achieved. The median wait time is reduced by 4% compared to using only the mean prediction of the GP.