Gain-scheduled bicycle balance controller based on system identification

Paper i proceeding, 2024

Balancing a bicycle through steering is similar to balancing an inverted pendulum, with a travel-speed-dependent pivot point. This paper derives a speed-dependent balancing controller for a self-balancing bicycle. This controller is based on an identified gray box model. The identification procedure is formulated as a weighted least squares problem with the time-varying parameter of the model. Identification data was generated on a controlled bicycle robot. Excitation experiments were designed to account for the unstable nature of the problem. Based on this identified model, a gain-scheduled controller is derived for a speed-independent closed-loop performance for a speed range. The controller is further implemented on the bicycle and tested for a set of speeds. Tests performed on the bicycle illustrate the gain-scheduled controller’s performance gain.