Control Strategies for Reducing Ground Tear in Agricultural Vehicles with Overactuated Steering

Abstract

This thesis investigates how two different steering configurations, tire steering and pivot steering, as well as control strategies can be utilized to reduce ground impact caused by shear forces in different scenarios. The vehicle dynamics for the two steering configurations and their ground interaction on deformable soil is modelled and simulated using Matlab. A state space formulation for each steering setup is derived and used for the development of an LQR and MPC controller. The control objective is reference tracking of the longitudinal velocity, and front and rear steering angles with the ultimate goal of reducing excessive shear forces. The developed modelbased controllers utilize input rate penalties to promote smooth control actions, as well as weighting terms on slip ratio and slip angle, which are directly related to the generated shear forces. In addition, the MPC formulation incorporates explicit constraints, allowing tire slip to be directly limited. A PI controller is also developed as a benchmark for comparison with the model-based controllers. Both the steering configurations and developed controllers are evaluated in simulation under both uniform and asymmetric terrain conditions. The pivot steering model combined with the implemented controller shows a reduction in total shear force compared to the tire steering model. However, it shows slightly reduced capabilities in reference tracking performance compared to tire steering. The model based controllers were successfully implemented and shows improved performance compared with the PI controller in terms of ground impact reduction as a result of the input rate limitations. Furthermore, the model based controllers could successfully incorporate slip ratio weights for reduction of longitudinal shear forces. The MPC also provided promising results in incorporating weights on slip angle and direct constraints on tire slip. Both steering configurations successfully incorporates an extended Kalman filter for state estimation, although with reduced performance