Deep learning based prediction model of recurrent pedal pressing for low speed

Traffic congestion in Malaysia's major cities has become a daily phenomenon, where it is common for a person to be stuck, at least twice daily from their house to the workplace. Being trapped in traffic for hours in a sitting position demands repetitive movements of manually pressing the accelerator and brake pedals excessively, which, if the correct seating posture is not maintained, it can cause significant fatigue and in a long-term, it will negatively impact the driver's health physically and psychologically. In light of this, this paper aims to investigate the relation between the vehicle speed and the recurrent brake pedal pressings pattern at certain leg postures while being trapped in the traffic using Deep learning technique. Several sensors were used for acquisition of input and output data, which are the leg postures and force produced during recurrent braking at low speed. The system utilizes Google Colaboratory to build a model, train and test the model using Python programming language to predict the vehicle speed during the traffic jams. This study begins with an experimental setup and data collection on the brake pedal pressing force and leg posture angle, followed by modeling of the relation using LightGBM deep learning-based model. The model validation was conducted to ensure a good model accuracy, which was found to have more than 80 % accuracy.