Refinements of the Dynamic Inversion Part of Hierarchical 4WIS/4WID Trajectory Tracking Controllers 2023-01-0907

To tackle the over-actuated and highly nonlinear characteristics that four-wheel-independent-steering and four-wheel-independent -driving (4WIS/4WID) vehicles exhibit when tracking aggressive trajectory, a hierarchical controller with layers of computation-intensive modules is commonly adopted. The high-level linear motion controller commands the desired state derivatives of the vehicle to meet the overall trajectory tracking objectives. Then the system dynamic is inversed by the mid-level control allocation layer and the low-level wheel control layer to map the target state derivatives to steering angle and motor torque commands. However, this type of controller is difficult to implement on the embedded hardware onboard since the nonlinear dynamic inversion is typically solved by nonlinear programming. This article refines the dynamic inversion part of current hierarchical trajectory tracking controller for 4WIS/4WID vehicles with consideration of the nonlinear tyre/vehicle dynamics and of computational complexity. First, the mid-level control allocation layer distributes target resultant forces to each tyre via the less computationally troublesome direct allocation method. The current method to determine the attainable generalised force subset (AFS) in direct allocation cannot incorporate the tyre friction circle constraints, leading to an inaccurate AFS. The inaccuracy of AFS will either introduce inversion error or fail to return controls for theoretically-attainable generalised force. Hence, a means of determining the refined AFS, which considers the actual tyre force limits and uses polygons to approximate tyre friction circles, is designed. Second, the low-level wheel control layer steers and drives the wheel to achieve the target tyre forces. An iterative method with proofed convergence is proposed to inverse the nonlinear tyre model. The effectiveness of the refinements is validated by Carsim and MATLAB/Simulink co-simulation. Simulation results demonstrate that the vehicle can well follow an obstacle avoidance trajectory. Comparisons with the direct allocation using the AFS determined by the current method are made. The refined AFS we proposed can not only enlarge the envelope of vehicle controllability, but also preserve the desired direction of the generalised force vector if it is beyond the vehicle’s capability, highlighting the importance of accounting for the nonlinearities we considered during the design.