Research On Intelligent Outide-area Parking Guidance Strategy Based On Reinforcement Learning

The number of vehicles has increased sharply,and the serious mismatch between the number of vehicles and the parking area has led to the increasingly difficult parking problem,which has a serious impact on urban congestion.At the same time,the rapid development of electric vehicles brings new parking scenarios in the future.The integration of parking facilities and charging facilities has become a future development trend.The limitation of charging time brings higher requirements to future parking scenarios.In response to these problems,reasonable parking guidance strategies are essential.With the rapid development of intelligent transportation,Internet and communication technologies,intelligent outside-area parking guidance is also possible.Aiming at the parking and charging problem of electric vehicles in future scenarios,this paper analyzes the factors that affect the user’s choice of parking area,and optimizes outside-area parking induction from the user’s vehicle side and the parking area side respectively.Real-time information exchange is realized by using the communication architecture between vehicles and roadside units under vehicle-road coordination,and a reservation mechanism is introduced to build a future parking information prediction model.In order to reduce parking costs,allocate resources reasonably,and improve parking space utilization,this paper proposes a The learned intelligent off-site parking induction strategy is set up with two groups of reinforcement learning problem models,and a dual-model off-site parking induction strategy based on reinforcement learning is further proposed.This paper mainly does the following work:(1)Introduce a reservation mechanism to solve the uncertainty of future parking availability information in parking induction.In order to build the electric vehicle parking guidance scenario model,it is necessary to predict the future parking information on the premise of obtaining real-time information.When building the parking information prediction model,the prediction of the parking availability probability is abandoned,and the parking time and charging time are calculated by parking reservation.Presetting the travel destination,predicting the available information of future parking through the model,and solving the uncertainty of future parking information to a certain extent is the premise of realizing the parking induction strategy.(2)Combined with reinforcement learning method,the electric vehicle parking induction problem is modeled on the basis of Markov decision process.Considering the user’s requirements for walking distance,queue waiting time,and the goal of balancing the parking occupancy rate of each parking lot,the problem is modeled from two aspects,and different state spaces and reward functions are designed.The strategy of the two models is integrated by the ranking voting mechanism,and a dual-model outside-area parking induction strategy based on reinforcement learning is proposed,and the optimal induction strategy considering the global optimization is obtained.(3)Build a simulation environment on the ONE simulation platform,model the dual-model reinforcement learning induction model,implement the parking induction algorithm,and compare it with several known parking induction algorithms.By comparing the average parking time of vehicles,the total distance spent on parking,and the cumulative number of stops,it is concluded that the intelligent off-site parking guidance strategy can significantly reduce parking costs,allocate resources reasonably,and improve the utilization of parking spaces in electric vehicle charging scenarios.