Pedestrian collision risk assessment based on motion prediction considering false alarm reduction

Pedestrian collision risk assessment is crucial for the protection of vulnerable road users (VRUs) in the Automatic Emergency Braking Pedestrians System (AEB-P). The challenge of pedestrian collision risk assessment is to accurately determine whether there is a collision risk between pedestrians and vehicles due to the uncertainty of pedestrian motion. Consequently, a collision risk assessment method based on motion prediction is proposed in this paper. Firstly, the extended Kalman filter (EKF) and Markov are used to estimate and predict the pedestrian state, respectively. According to the states, collision risk is evaluated through the collision sensitive boundary (CSB). Besides, a false alarm reduction algorithm, considering the impact of the past risk states, is proposed for reducing the false alarm rate, and it is suitable for some binary calculation methods for collision risk. Finally the Prescan and Simulink co-simulation model of the AEB-P system is established to evaluate the proposed architecture on the ETH/UCY dataset. The simulation results demonstrate that the proposed collision risk assessment method is safer and more effective than the current algorithms that do not consider motion uncertainty.