Applying Embedded Hybrid ANFIS/Quantum-Tuned BPNN Prediction to Collision Warning System for Motor Vehicle Safety

This study is to explore how to realize high-performance collision warning system (CWS), providing the precaution against traffic crash in transit. An embedded hybrid adaptive network-based fuzzy inference system (ANFIS) plus quantum-tuned back-propagation neural network (QT-BPNN) built in the platform with DaVinci+XScale-NAV270 was employed to realize collision warning system and we also installed motor vehicle event data recorder (MVEDR). Finally, experiments and verification of the proposed approach were done successfully to achieve better accuracy and more effectiveness on warning level issuing and event data record to motor vehicle.

1. Introduction

Vehicle safety is a very important issue, but people did not pay much attention on it in early days. The statistics of the injuries and fatalities in traffic accidents become more serious year by year. Therefore, someone began to notice the relevant problems of vehicle safety and make more studies and solution to this issue, for example, Adaptive Cruise Control (ACC) [1], Antilock Brake System (ABS)[2] [3], Collision Warning System (CWS), Event Data Recorder (EDR) [4], On-Board Diagnostics (OBD) [5], and Emergency Automatic Brake (EAB), etc. Generally speaking, the causes of the traffic crash can be categorized into three parts, “people”, “vehicle” and “environment” [6]. In United States, National Highway Traffic Safety Administration (NHTSA) also points out that there are about 80% − 90% traffic crashes caused by the “people” factor [7].