Intelligent collision risk assessment based on Neural Network Ensemble

In this paper, we propose the collision risk assessment system. When pedestrian is detected by radar or another sensor, system could know the pedestrian's position and velocity. Using this information, system can compute the collision risk. If system does not concerned about the simulation time, Monte Carlo Simulation is simple and powerful method. But in dynamic circumstance, the position and velocity of pedestrian is changed rapidly. So I propose to apply Neural Network Ensemble in this problem. Neural Network train the network using training data, this process take a long time. But by using trained network, system can compute the collision risk quickly. However, wide range of input data can cause huge memory use, and lengthy simulation time. So we propose apply Neural Network Ensemble to this problem. Neural Network Ensemble separate the input data and training each network with different data set. This method will reduce the computation load with small error.