Comparative Analysis of Classifiers for Classification of Emergency Braking of Road Motor Vehicles

Albert Podusenko,Katsunori Shimohara,Vsevolod Nikulin,Ivan Tanev

doi:10.3390/a10040129

Abstract

We investigate the feasibility of classifying (inferring) the emergency braking situations in road vehicles from the motion pattern of the accelerator pedal. We trained and compared several classifiers and employed genetic algorithms to tune their associated hyperparameters. Using offline time series data of the dynamics of the accelerator pedal as the test set, the experimental results suggest that the evolved classifiers detect the emergency braking situation with at least 93% accuracy. The best performing classifier could be integrated into the agent that perceives the dynamics of the accelerator pedal in real time and—if emergency braking is detected—acts by applying full brakes well before the driver would have been able to apply them.

Highlights

In recent years, there has been noticeable progress in technologies devoted to driving aids in road motor vehicles [1]
The main objective of our research is to resolve the problem of emergency braking classification (EBC)The solely from the motion ofisthe by building an intelligent
In a real-world driving situation, the most critical error might be in false positive cases, i.e., when automated braking is activated in an incorrectly classified emergency braking situation

Summary

Introduction

There has been noticeable progress in technologies devoted to driving aids in road motor vehicles [1]. The interest in this field is driven by the continuous desire to increase road safety. Examples of active driving aids include anti-locking brake systems, traction control, an electronic stability program, and automated brake assistance. The latter can be implemented either as a fully automated braking system [2,3] or as an assistant to the human driver [4,5]. With the intention of building a reliable and applicable helpful system, we decided to leave fully automated braking aids out of the scope of our current research and focus on brake assistance instead

Objectives

Methods

Results

Discussion

Conclusion