Model predictive for path following with motorcycles: application to the development of the pilot model for virtual prototyping

Controlling a riderless motorcycle is a challenging problem because the dynamics are nonlinear and non-minimum phase. In this paper, an innovative control strategy is proposed for driving a motorcycle along a given path, tracking a speed profile given as a function of the arc-length of the path. The solution is based on model predictive control. Exploiting the possibility given by MPC to work on trajectories, we invert the cause-effect structure of the problem and act as if the roll angle was an input. We then determine, among a polynomial set of roll angle trajectories, the optimal one in terms of the error at preview distance from the target path. By inverting the dynamics, we compute the steering and longitudinal controls needed to track the computed roll trajectory.