Tire normal force estimation using artificial neural networks and fuzzy classifiers: Experimental validation

Abstract

Tire normal forces play a crucial role in vehicle dynamic control systems (VDCs), and an accurate estimation of them could substantially improve vehicle handling and safety. Total normal forces on tires are the combination of the static and dynamic values. These values change by varying the vehicle static parameters (vehicle mass and CG position), the road grade, and vehicle dynamic states. All of the mentioned forces and parameters are used in the design of vehicle dynamics controllers. Therefore, to have an efficient estimation solution to access all different target controllers, the proposed estimation algorithm is divided into two parts. The first part is an ANN-based vehicle mass and CG position estimation algorithm and the second one is a deep-learning-based algorithm and an integrated hardware–software method to estimate dynamic normal forces. The first part includes two neural network (NN) blocks related to vehicle roll and pitch dynamics. The outputs of the NN blocks are load distributions for the axes associated with each dynamics. Based on the measurement unit data, the vehicle’s maneuvers are classified into one of many categories used for downstream tasks. The rules of the fuzzy classification logic determine which of the mentioned blocks to be activated. At the same time, an experimentally validated 9-DOF vehicle model instantaneously observes the estimated values. In the second part, long short-term memory (LSTM), gated recurrent unit (GRU), and an integrated hardware–software method were developed to assess the tire’s dynamic normal forces during maneuvers. The performed simulations and the results of field tests show that the proposed algorithm can accurately estimate the considered parameters in the presence of noise and disturbances. It is shown that the proposed algorithm is more accurate and faster than the extended Kalman filter and recursive least square estimation methods to estimate static values.