Safety benefit evaluation of intelligent driving systems based on multisource data mining

Different manufacturers achieve intelligent driving system function diversely, which imposes higher impartiality requirements for the evaluation method of the third party. To this end, this article presents a safety benefit evaluation method of intelligent driving systems based on multi-source data mining. On the basis of the discussion over the nature of general system identification, this approach uses neural network to learn the behavior of the evaluated object using the running vehicle data collected and provided by the manufacturer. Combined with the trained network controller, the test scenario model and the car following model extracted from the field operational tests data, and with the occupant injury model obtained from the accident data, Monte Carlo random simulation is used to calculate the injury risk with or without the evaluated system, then the safety benefit by comparison is estimated. In this article, the adaptive cruise system and the automatic emergency braking system are evaluated. The results show that the neural network can accurately imitate the behavior of the object to be evaluated. There is only 0.01 error between the evaluation results using this network and the real object.