Performance Analysis of Sensor Fusion Models for Brake Pedal in a Brake-by-Wire System

This paper focuses on analyzing the performance of sensor fusion models for a brake pedal in electromechanical braking (EMB) systems. To properly control a brake, a signal obtained from a pedal is important and must be stable. Sensor fusion models can be used to make the pedal signal more stable and resilient against the abnormalities. In this work, a pedal sensor and a pedal effort sensor are exploited for the sensor fusion to control the brake. Then, we analyze the performance of each fusion model (e.g., average method, moving average method, median filter, iterative filter, and Marzullo's algorithm) in various fault scenarios. We propose the hybrid method using multiple models to have the better resiliency and to detect the faulty sensor as well. For the practical experiments, we employ the real measurement data set obtained from a vehicle at different speeds. Then, its fused result is validated using EMB system test bench to confirm how the fused result influences on the motor of the brake.