Integrated Strategy for Automatic Lane-Changing Decision and Trajectory Planning in Dynamic Scenario

Abstract

Abstract Automatic lane change is a necessary part for autonomous driving. This paper proposes an integrated strategy for automatic lane-changing decision and trajectory planning in dynamic scenario. The Back Propagation Neural Network (BPNN) is used in decision-making layer, whose prediction accuracy of the discretionary lane-changing is 94.22%. The planning layer determines the adjustable range of the average vehicle speed based on the size of the “lane-changing demand”, which is obtained based on the data of hidden layer in neural network, and then dynamically optimizes the lane-changing curve according to the vehicle speed and the current scenario. In order to verify the rationality of the proposed lane-changing architecture, simulation experiments based on a driving simulator is performed. The experiments show that the vehicle’s maximum lateral acceleration under the proposed lane-changing trajectory at a speed of 70km/h is about 0.1g, which means the vehicle has better comfort during lane-changing. At the same time, the proposed lane-changing trajectory is more in line with the human driver’s lane-changing trajectory compared with that of other planning strategy. Meanwhile, the planning strategy can also support the lane-changing trajectory planning on a curved road.