Real time adaptive fuzzy control of inverted pendulum

In this paper, we provide a real-time adaptive fuzzy controller utilized for balancing a self-erecting inverted pendulum with time-varying mechanical parameters. One learning algorithm is based upon truncated temporal backpropagation and weight perturbation suitable for realization in hardware. Main focus was the ability to learn from trials and the possibility to implement the functions in hardware for on-chip learning. The simulations and testing were done using hardware-in-the-loop, the robustness and adaptivity of the controller are shown from simulation traces.