Path Planning for Intelligent Vehicle Collision Avoidance of Dynamic Pedestrian Using Att-LSTM, MSFM, and MPC at Unsignalized Crosswalk

In this article, path planning for intelligent vehicle collision avoidance of dynamic pedestrian using attention mechanism-long short-term memory network (Att-LSTM), modified social force model (MSFM), and model predictive control (MPC) is systematically investigated, and pedestrian-dynamic vehicle conflict scene at an unsignalized crosswalk is covered. First, a data-driven stacking fusion model based on the Att-LSTM and MSFM is proposed for pedestrian path prediction. Pedestrian heterogeneity (age and gender) is taken into account for the first time. The data-driven stacking fusion model is verified with the existing methods. Then an MPC-based path planning-tracking system considering pedestrian path prediction is developed. The predicted path of pedestrian is treated as a pedestrian-safety region with the combination of a semiellipse and semicircle, and the front wheel steering angle is calculated to prevent the intelligent vehicle from colliding with the dynamic pedestrians. Simulink–Carsim simulations are presented to simulate the real dynamic environment where crossing pedestrians exist. The results illustrate that the developed MPC system considering pedestrian path prediction can provide dynamic path planning performance acceptably and effectively, and make it possible for the intelligent vehicle to present the great feasibility for pedestrian safety protection and traffic efficiency improvement.