A Self-optimization Algorithm of Multi-style Smart Parking Driven by Experience, Knowledge and Data

Abstract

In this paper, a new algorithm based on experience, knowledge, and data was proposed to address the problems that the traditional automatic parking algorithms, specifically for parallel parking, have large global computing requirements, high real-time computing power requirements, and limit the optimization space. Firstly, a parking operation model based on segmented logic and judgment criteria from the driver’s experience was built for the entire parking process. Secondly, a multi-style evaluation system based on parking knowledge and customer demand was established. It refers to the volatility boundary of the index and uses Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS), Rank Correlation Analysis (G1), and entropy methods to determine the weight distributing model. Finally, based on an optimization test, the data was analyzed and the parameter relationship between variable and index was fit. A rapid self-optimizing algorithm was designed by combining the relation-ship mentioned above with the Particle Swarm Optimization (PSO) algorithm. It has been verified that the algorithm has increased the parking scores of SUVs and light buses at different starting points by 28.35% and 16.68% in 5.77s and 4.72s. Compared with traditional optimization algorithms, the algorithm designed in this paper saves 84.47% and 85.10% of the time. Therefore, the efficiency and adaptability in multiple situations of the rapid self-optimization algorithm have been verified.