An optimal fuzzy logic controller for an underactuated unicycle

In this paper, we propose a hybrid design method for fuzzy logic controller (FLC), where the control objective for unicycle is to achieve velocity control of the wheel while keep the pendulum at the balanced position that is an unstable equilibrium. The hybrid design consists of three phases. First, FLC structure including the number of rules, membership function, inference, and parametric relations, are chosen based on heuristic knowledge about the unicycle. Then, based on a linearized model and linear feedback, the output parameters of FLC are determined quantitatively for the stabilization of the unicycle. Next, fine tuning of FLC output parameters are carried out using an iterative learning tuning (ILT) algorithm, where ILT iteratively minimizes an objective function that specifies the desired unicycle performance. The rationale of introducing the hybrid FLC design is to fully utilize available information, which is achieved by combining model-based and model-free designs, hence improve FLC performance. Through intensive simulations and comparisons, the effectiveness of the proposed hybrid FLC is validated.