Kinematic and dynamic adaptive control of a nonholonomic mobile robot using a RNN

In this paper, an adaptive neurocontrol system with two levels is proposed for the motion control of a nonholonomic mobile robot. In the first level, a recurrent network improves the robustness of a kinematic controller and generates linear and angular velocities, necessary to track a reference trajectory. In the second level, another network converts the desired velocities, provided by the first level, into a torque control. The advantage of the control approach is that, no knowledge about the dynamic model is required, and no synaptic weight changing is needed in presence of robot's parameters variation. This capability is acquired through prior 'meta-learning'. Simulation results are demonstrated to validate the robustness of the proposed approach.