Real-time riding simulation model for motorcycle including path-following controller with nonlinear model predictive control

Abstract

This study constructs a path-following controller (rider model) based on nonlinear model predictive control (NMPC) with steering torque as the control input. By combining this with the dynamics model derived by Hatakeyama et al. (2024), we propose a simulation model that considers the main dynamic characteristics of a motorcycle, such as weave and wobble modes. In particular, the internal model used in NMPC has been simplified to ensure a reasonable response when the simulation time is shorter than the real time. In the path-following controller, the parts that calculate the reference roll angle from the reference path and those that calculate the steering torque input from the reference roll angle are handled separately. These components are constructed as cascade structures of outer and inner loops, respectively. This allows the two loops to appropriately use different sampling periods and prediction horizon settings for NMPC, enabling fast computation. In addition, an optimal NMPC solver can be used for both the outer and inner loops, further enhancing fast computation. In riding simulations involving high speeds and large roll angles, where nonlinearities are stronger, stable responses that are roughly consistent with actual riding were obtained in a simulation time shorter than the assumed time for one lap in actual circuit riding. The developed simulation model can be used as a reliable and practical support tool for motorcycle designing.