Automated Valet Parking and Charging: A Dynamic Pricing and Reservation-Based Framework Leveraging Multi-Agent Reinforcement Learning
Gordon Owusu Boateng, Haonan Si, Huang Xia, Xiansheng Guo, Cheng Chen, Isaac Osei Agyemang, Nirwan Ansari
Abstract
Vehicle parking resource provisioning in major cities and urban areas has gradually become a challenging issue in Intelligent Transportation Systems (ITS) due to the upward trend in car ownership rates. Besides, the increasing attractiveness of Autonomous Electric Vehicle (AEV) technologies complicates the parking problem as self-parking and EV-charging solutions need to be integrated into existing parking infrastructure. Dynamic pricing and reservation-based automated parking and charging are envisioned to accommodate the increasing demand for parking and EV-charging services. This will minimize traffic congestion and enhance road safety. This paper proposes a novel intelligent framework based on dynamic pricing and in-advance parking and charging reservations for Automated Valet Parking and Charging (AVPC) scenarios. We formulate the dynamic pricing problem between a Parking Lot Manager (PLM) and multiple Autonomous Vehicles (AVs) as a twostage Stackelberg game in which the PLM, as the leader, sets its service price in the first stage to maximize its utility, and each AV, as a follower, determines its service demand in the second stage to maximize its utility. Then, we theoretically prove the existence and uniqueness of the Stackelberg Equilibrium (SE). Considering the stochastic nature of the parking traffic, we transform the game-based optimization problem into a Multi-Agent Markov Decision Process (MAMDP) and propose a Stackelberg Game-aided Multi-Agent Dueling Deep Q-Network (SG-MADDQN) algorithm to solve the problem. Comprehensive simulation results and analysis prove that the proposed algorithm achieves convergence and can best balance the pricing and demand strategies of the PLM and AVs compared with existing solutions.