Neuro-dynamics and neuro-fuzzy based approaches to real-time motion planning and control of mobile robot systems

Motion planning and control are fundamentally important in robotics. In this thesis, neuro-dynamics and neuro-fuzzy based approaches are proposed for real-time motion planning and control of mobile robot systems, including task assignment of multi-robots, robot motion planning with obstacle avoidance, robot path tracking, and robot navigation in unknown environments using onboard sensors. For multirobots, a self-organizing map based approach is developed. It is capable of dynamically controlling a group of mobile robots to achieve multiple tasks at different locations in dynamic environments subject to uncertainties. The robot motion can be dynamically adjusted to uncertain environments. For a mobile robot, a gated dipole neuro-dynamics approach is proposed for real-time motion planning with obstacle avoidance in non-stationary environments. The real-time robot motion is planned through the dynamic activity landscape of a neural network. Based on the backstepping model and the gated dipole model, a neuro-dynamics based approach is proposed for real-time path tracking of a mobile robot. It is capable of generating bounded acceleration control signals to smoothly track a reference path without the velocity jump problem, and without the "perfect velocity tracking" assumption. For a mobile robot using only onboard sensors, a neuro-fuzzy based method is developed for real-time navigation in unknown environments. A simple fuzzy rule base is designed. Two learning algorithms are developed to tune the parameters of the membership functions and automatically suppress redundant fuzzy rules. The "dead cycle" problem is resolved by a state memory strategy.