Alternative sensor system and MLP neural network for vehicle pedal activity estimation

Area of Science:

• Automotive Engineering
• Machine Learning
• Environmental Science

Background:

• Driver behavior significantly influences vehicle fuel consumption and pollutant emissions.
• Directly measuring pedal activity (throttle, brake, clutch) is vehicle-specific and complex.
• Estimating driver activity indirectly offers a universal approach.

Purpose of the Study:

• To develop a universal method for estimating driver activity.
• To enable accurate analysis of driver behavior's impact on fuel economy and emissions.
• To create a system independent of vehicle type, model, or manufacturing year.

Main Methods:

• Utilized an alternative sensor system measuring engine speed, vehicle speed, frontal inclination, and linear acceleration.
• Employed a multilayer perceptron neural network with a single hidden layer for data analysis.
• Developed an indirect estimation of vehicle pedal activity (throttle, brake, clutch).

Main Results:

• Successfully estimated driver activity indirectly through sensor data.
• The proposed method demonstrated independence from specific vehicle parameters.
• The neural network model effectively processed indirect signals to infer pedal usage.

Conclusions:

• The indirect estimation of driver activity using sensor data and neural networks is feasible and universal.
• This approach provides a valuable tool for analyzing driver behavior's impact on vehicle performance and environmental impact.
• Future research can refine the model for enhanced accuracy in fuel consumption and emissions prediction.