Construction and Research of Machine Learning-based Pedal Misoperation Recognition Method

Abstract

Background: A large number of traffic accidents are caused by drivers mistakenly stepping on the pedal in emergencies, highlighting an urgent concern about reducing pedal misoperation of drivers. Nowadays, machine learning has been widely applied in the fields of automatic driving, vehicle-circuit coordination, automatic obstacle avoidance, etc. However, the detection technology that focuses on the driver’s driving behavior during driving has not yet been able to greatly reduce the occurrence of traffic accidents. Purpose: Data collected by Chang’an University’s comprehensive driving behavior test platform were used to recognize whether a car driver has the behavior of pedal misoperation by developing and comparing a variety of machine learning algorithms. This paper proposes a method based on machine learning algorithms that takes into account the visual characteristics of the driver to identify pedal misoperation. Research methods: Five different machine learning algorithms were compared for the behavioral judgment of pedal misoperation of drivers through multiple evaluation indexes. Then, the performance of each algorithm was evaluated. As verified, the RandomForest algorithm outperforms all other algorithms with an accuracy rate of 98.4%.Conclusion: According to the research results, a method for recognizing the pedal misoperation behavior of drivers based on the RandomForest algorithm considering visual characteristics can more accurately recognize whether there is a pedal misoperation behavior.