Automatic Parking System for Multi-vehicle in the Autonomous Driving Platform of the DeepRacer

In the present research on automatic parking, there are few automatic parking on multi-vehicle scheduling, and the simulation of the scene is relatively simple. In this paper, a multi-vehicle scheduling parking management method is proposed, and the scene of a double-deck parking lot is used in the simulation experiment. In the experiment. We preset a parking time for each vehicle so that the algorithm can classify the vehicles based on their parking time. Vehicles with shorter parking times will be arranged in the parking lot on the first floor which is closer; vehicles with longer parking times will be arranged in the parking lot on the second floor which is farther away. First, we studied the motion action and parking process of the vehicle, and a kinematics model is built. Secondly, we built the model of vehicles and a double-deck parking lot with SolidWorks. At the same time, the sensors are configured for the vehicles, and the models are imported into Gazebo. Then, we use line patrol navigation to allow the vehicle to move to the parking lot on different floors according to the patrol lines of different colors, completing the path planning of a single vehicle from the starting point to the parking space through ROS communication. Finally, utilize the parking management node to divide all vehicles into various units, and then schedule parking management for each unit separately, completing the automatic parking of the total vehicles. The results demonstrated that scheduling with 2 vehicles as the minimum unit is about 30% faster than scheduling with 1 vehicle as to the minimum unit.