Information-Driven Autonomous Intersection Control via Incentive Compatible Mechanisms

We propose a new information-driven intersection control to enhance the quality of transportation by using communication between vehicles and roadside units. The state-of-the-art solutions for intersection control only have access to the sensor data that is collected by vehicles or roadside units. However, congestion at intersections can have different impact on different drivers, and yet such an impact cannot be measured by sensors. An effective intersection control can consider such driver-exclusive differences based on the information reported by the drivers, which can substantially enhance the quality of transportation. However, such information is driver-exclusive, i.e., not verifiable easily, and therefore prone to be misreported strategically. We propose strategy-proof intersection control addressing such issues via a payment-based incentive-compatible mechanism. Particularly, vehicles at close proximity of the intersection report their driver-exclusive utility functions that they want to maximize (not necessarily truthfully), while the roadside unit seeks to maximize the sum of those utilities, i.e., social welfare, by scheduling intersection usage and charging each vehicle an amount of time-tokens corresponding to their impact on other drivers. This approach, based on the Vickrey-Clarke-Groove mechanism, guarantees truthful utility reporting by the vehicles and, correspondingly, maximizes the social welfare. The proposed scheme is universal such that it can be implemented based on various utility functions or intersection control constraints. We also provide a practical implementation to analyze the performance via numerical simulations.