A Sensor Fusion Method Based on an Integrated Neural Network and Kalman Filter for Vehicle Roll Angle Estimation

Area of Science:

• Automotive Engineering
• Robotics
• Control Systems

Background:

• Accurate vehicle roll angle estimation is crucial for active safety systems.
• Inertial Measurement Unit (IMU) sensors offer a cost-effective solution for motion detection.
• Existing methods may struggle with vehicle non-linearities and sensor noise.

Purpose of the Study:

• To develop a novel sensor fusion estimator for precise vehicle roll angle estimation.
• To integrate a Neural Network (NN) with a Kalman filter for enhanced accuracy.
• To validate the estimator's performance across various dynamic maneuvers and conditions.

Main Methods:

• A Neural Network (NN) estimates a 'pseudo-roll angle' using easily measurable IMU sensor data.
• The estimated 'pseudo-roll angle' is incorporated into a Kalman filter for noise reduction and error minimization.
• The NN was trained on diverse maneuvers (J-turn, double lane change, lane change) under varying speeds and friction coefficients.

Main Results:

• The proposed NN and Kalman filter-based estimator effectively accounts for vehicle non-linearities.
• The estimator demonstrates reduced estimation errors (variance of norm and maximum errors).
• Experimental comparisons show superior performance over methods relying solely on suspension deflections.

Conclusions:

• The novel sensor fusion approach provides a robust and accurate method for vehicle roll angle estimation.
• This technique offers significant improvements in estimating vehicle dynamics, particularly under challenging conditions.
• The integration of NN and Kalman filtering presents a promising direction for advanced automotive safety systems.