Tire lateral force estimation and grip potential identification using Neural Networks, Extended Kalman Filter, and Recursive Least Squares

This paper presents a novel hybrid observer structure to estimate the lateral tire forces and road grip potential without using any tire-road friction model. The observer consists of an Extended Kalman Filter structure, which incorporates the available prior knowledge about the vehicle dynamics, a feedforward Neural Network structure, which is used to estimate the highly nonlinear tire behavior, and a Recursive Least Squares block, which predicts the road grip potential. The proposed observer was evaluated under a wide range of aggressive maneuvers and different road grip conditions using a validated vehicle model, validated tire model, and sensor models in the simulation environment IPG CarMaker((R)). The results confirm its good and robust performance.