Automotive Crash Detection Using Multi-sensor Data Fusion

The focus of this paper is the early detection of the frontal crash in automobiles, for the purpose of effective Airbag deployment decision, using information from multiple sensors deployed on-board, specifically accelerometer sensors located on the vehicle engine and on the driver’s seat-belt. Measured acceleration signal data streams from the sensors are fused, based on principles of Multi-Sensor Data Fusion (MSDF), for faster detection of a crash. Adaptive Kalman filtering is employed for simultaneous estimation of individual sensor data and their signal level fusion. The proposed crash detection system is simulated in MATLAB and validated using US-NHSTA (National Highway Safety Traffic Administration) automotive crash datasets. For comparative evaluation, crash detection algorithms for individual sensor data are also simulated and tested on the same datasets. The MSDF based system resulted in faster crash detection when compared to single sensor systems.