Robust Control of a Low-Cost Mobile Robot Using a Neural Network Uncertainty Compensator

This paper presents a robust control design for a low-cost mobile robot under modeling uncertainties and external disturbances. We use a radial basis function neural network (RBFNN) to estimate and compensate for the model uncertainties and external disturbances. The proposed control design is model-free with guaranteed stability and good path-following performance. The RBFNN weight regulation and adaptive gains are designed based on the Lypanov method. Simulation and experimental results illustrate the design and demonstrate the strength of the proposed control applied to a nonholonomic wheeled mobile robot driven by low-cost permanent magnet dc motors without shaft encoders. The comparison results between proposed control and feedback linearization control confirm the effective role of the compensator in terms of precision, simplicity of design and computations.