Path Generation Algorithm Based on Crash Point Prediction for Lane Changing of Autonomous Vehicles

Abstract

To reduce the calculation time needed to determine the optimal path, the form of the road and the path of an autonomous vehicle were linearized; additionally, among multiple obstacles, only those that were potentially dangerous were chosen. By considering the movement of moving obstacles, the cost was calculated. The calculation time was shortened by reducing the number of design variables of the optimal path, when changing lanes to avoid obstacles, to two. Limiting conditions, such as the lateral and longitudinal acceleration, were excluded from the cost calculation by restricting the search region of the design variable. The final result was calculated using a relatively free search of the golden-section search regarding the initial value setting. For the golden-section search, the number of final design variables was reduced to one; this was done by optimizing the search direction. The search direction was determined based on the final position of the vehicles and the calculated optimal points. By including a collision avoidance algorithm and moving in a short period of time, the calculated optimal path prevented accidents due to path errors caused by simplification. The path could be found easily, even for complex road shapes and with multiple vehicles nearby.