We explore the feasibility of classifying the emergency braking situations solely from the pattern of lifting motion of the accelerator pedal. The ultimate objective such a classification lies in reducing the time lag between the instant (i) when the driver lifts their foot from the accelerator pedal and (ii) then energetically presses the brake pedal in order to stop the car in hazardous, emergency braking situations. At high speed, this time lag results in several meters of traveled distance. Therefore, minimizing the lag would result in reduced risks of road traffic accidents lead and enhanced road safety. We proposed a machine learning approach of classifying the emergency braking based on gradient boosting. The latter infers the braking situation as either emergency braking or normal braking or deceleration from the training set of time series of braking events recorded from 11 humans (men and women, aged 22–52) while driving a car in a full scale drive simulator. The experimental results show that the trained classifier detects the emergency braking situations with the f-score of about 0.941 on the testing set of the dynamics of the accelerator pedal.
[1]
Gaël Varoquaux,et al.
Scikit-learn: Machine Learning in Python
,
2011,
J. Mach. Learn. Res..
[2]
J. Friedman.
Greedy function approximation: A gradient boosting machine.
,
2001
.
[3]
Ivan Tanev,et al.
Cause and Effect Relationship between the Dynamics of Accelerator and Brake Pedals during Emergency Braking
,
2018
.
[4]
Roger Fletcher,et al.
A Rapidly Convergent Descent Method for Minimization
,
1963,
Comput. J..
[5]
Yoav Freund,et al.
A decision-theoretic generalization of on-line learning and an application to boosting
,
1995,
EuroCOLT.
[6]
J. Friedman.
Stochastic gradient boosting
,
2002
.
[7]
Tom Fawcett,et al.
An introduction to ROC analysis
,
2006,
Pattern Recognit. Lett..