Intelligent networked vehicles face parking problems in the process of urbanization, and closed residential areas put forward higher requirements for autonomous parking systems due to the characteristics of narrow space and complex obstacles. Aiming at the autonomous parking problem of intelligent networked vehicles in the narrow parking space of closed residential areas, this paper proposes a path planning algorithm based on multi-sensor fusion localization. The algorithm constructs an environmental data acquisition system containing 12 ultrasonic sensors and 4 high-definition cameras, and establishes a multi-sensor fusion framework with a camera model, an IMU measurement model, and a kinematic model of a wheel tachometer. An improved inverse extension Hybrid A★ path planning algorithm is designed, which improves the planning efficiency by interchanging the start point and the target point, so that the algorithm expands the nodes from inside the narrow space to the open space. The simulation experiment results show that the path planning time of the algorithm in different scenarios is within 1.4s, of which the fastest planning time is 0.75s. In the test of different parking space sizes, the minimum required parking space size is 6.821m×2.164m when the vehicle speed is 3km/h, and increases to 7.058m×2.205m at 6km/h. The algorithm successfully realizes safe path planning for vertical and parallel parking scenarios, and the error of the vehicle’s intersecting position with the parking space line is controlled within 12 cm. This study provides an effective technical solution for autonomous parking of intelligent connected vehicles in complex residential environments.
