Intelligent Control of Omnidirectional Robots using Recurrent Neural Networks
Nonlinear Control, Intelligent Control, Recurrent Neural Networks, Feedback Linearization, Sliding Modes, Mobile Robots.
Due to their great efficiency, security and flexibility, mobile robots are being increasingly used in industry. However, their positioning control is a great challenge due to the non-linear nature of this plant and the difficulty of estimating certain parameters, for example, the friction effects. In this work, non-linear controllers are applied to the trajectory control of an omnidirectional robot under the effect of unmodeled dynamics. The control approaches used in this work were both non-linear control strategies, Feedback Linearization (FBL) and Sliding Modes (SMC) both incorporated with an intelligent compensator utilizing Recurrent Neural Networks in order to assist the control by estimating uncertainties. The chosen architecture of the neural network was based in the need to compensate more complex dynamics and at the same time the restriction of computational complexity so that it could be embedded in the hardware of a mobile robot. The stability properties were proven by the principle of assintotic stability proposed by Lyapunov and the performance of the strategies were verifed through both simulations and experiments using Robotino, an omnidirectional mobile robot produced by Festo Didatics and a performance gain was observed when compared with the neural network without the recurrence.