Two wheel self-balanced mobile robot identification based on experimental data

Many researches have been accomplished about modeling of two wheel robots which stabilize themselves autonomously and various models have been proposed to explain the dynamic model of them. This paper considers two phases of system identification process to achieve dynamic equilibrium of the robot. In the first phase, it investigates a part of system as multi input multi output (MIMO) system. Considering the current of motors as inputs, and the angular velocity of wheels as outputs. In the second phase, this study investigates another part of the system as multi input single output (MISO) system which its inputs are the angular velocity of the wheels (the output of the first phase) and output is the vertical angle of the robot axis. This study compares ARX, ARMAX, BJ and OE as candid as linear models and Wiener and Hammerstein as nonlinear models. The results shows that the Weiner nonlinear model with saturated output has the best performance for the first phase and the Wiener nonlinear model with Wavelet network is suggested for the second phase of identification process.