A Simulation-Based Multiple-Objective Optimization for Designing K-Stacks Autonomous Valet Parking Lots

Abstract

Autonomous valet parking has drawn wide attention these years. The k-stacks layout, known for its ability to increase parking capacity by stacking vehicles more compactly, is of great practicality among all possible layout patterns. Although this layout can increase the capacity of a parking lot, it generates relocations, which let vehicles move additional distances and influence the lot’s peak hour service ability. For the sake of optimizing them all simultaneously, we propose a simulation-based multiple-objective optimization (SMOO) and use NSGA II to solve the problem, obtaining candidate solutions. Then, a nondominated sorting based on cumulative advantages (NSCA) method is put forward to select the most robust solution from all candidates, considering different demand scenarios. K-stacks parking lots optimized by the SMOO can provide 36%–59% more parking spaces than a traditional parking lot while keeping other evaluations fine. In addition, we specify high-demand and low-demand scenarios and discuss the impact of different aspect ratios. It is recommended to use k-stacks layouts when a lot’s length is close to its width.