A Data-Driven Near-Optimization Approach for Smart Parking Management Platforms

Abstract

This paper addresses an allocation optimization problem of parking slots in a real-time parking reservation platform in which parking demands randomly show up. For each decision period of a finite horizon, the reservation platform allocates the available parking slots at hand for the random demands with different types specified by the parking duration to maximize the total revenue over the whole horizon. This paper presents a real-time reservation framework and formulates the problem as a stochastic programming model, considering the different type demands with unknown probability distributions. We propose a data-driven near-optimization approach entitled a two-sample average approximation (2-SAA) to determine the allocation scheme over the horizon. In the 2-SAA approach, the confidence interval of the total revenue is established by the out-of-sample resampling method. The results of the numerical experiment validate the performance of the proposed 2-SAA algorithm.