A robust observer based on energy-to-peak filtering in combination with neural networks for parameter varying systems and its application to vehicle roll angle estimation

Abstract

This paper presents a robust observer based on energy-to-peak filtering in combination with a neural network for vehicle roll angle estimation. Energy-to-peak filtering estimates the minimised error for any bounded energy disturbance. The neural network acts as a ‘pseudo-sensor’ to estimate a vehicle ‘pseudo-roll angle’, which is used as the input for the energy-to-peak-based observer. The advantages of the proposed observer are as follows. 1) It does not require GPS information to be utilised in various environments. 2) It uses information obtained from sensors that are installed in current vehicles, such as accelerometers and rate sensors. 3) It reduces computation time by avoiding the calculation of observer gain at each time sample and utilising a simplified vehicle model. 4) It considers the uncertainties in parameters of the vehicle model. 5) It reduces the effect of disturbances. Both simulation and experimental results demonstrate the effectiveness of the proposed observer.