A hybrid strategy for parallel parking by combining geometric and optimization planners

Path planning is a key function of automatic parking systems, especially for parallel parking in tight spaces. A qualified planner should compute quickly, ensure optimality, and avoid extra halfway stops. This paper proposes a hybrid strategy to make up the shortcomings of geometric path planners by optimizing the pose and heading angle at the first entry point into the parking space. Considering the number of maneuvers in the limited parking space, the computation of the entry point is described as a nonlinear programming problem, and solved by the genetic algorithm offline. Based on these, feasible paths for different parallel parking spaces can be directly provided online using the parametric form of circular arcs. The effectiveness of the hybrid strategy is validated by comparative tests in several parallel parking spaces of varying sizes, and the details about adjusting maneuvers in the parking space are discussed. Additionally, the traceability of the planned paths is verified by experimental tests.