Impedance control of a manipulator using a fuzzy model reference learning controller

Many of the tasks robots are called upon to perform require mechanical interaction with an environment. As such, success depends on the capability to control this interaction. Commanding a force or motion trajectory, in conjunction with regulating and modulating the relationship between force and motion is one approach to controlling the interaction. This approach is called impedance control and the relationship between forces and motion is the desired manipulator impedance that the controller seeks to achieve.Simulation results are presented comparing the proposed control scheme with an impedance control using a traditional PID controller, showing that impedance control using a FMRLC can offer better performance in providing desired motion and desired force in repetitive tasks. The presented experimental results confirm that the FMRLC exhibits learning abilities as well as the ability to adapt to varying system parameters.The study shows that a well designed FMRLC used for position based impedance control, during repetitive tasks involving the interaction between a manipulator and an environment, can offer excellent performance where force and motion trajectory tracking is required.Many conventional and adaptive controllers depend on knowledge of the structure and parameters of the system being controlled. In some cases it is difficult to develop an appropriate model of the system in order to design a control scheme. In others, changes to the system's parameters adversely affect the performance of the controller to the extent that the desired manipulator impedance is not achieved. The aim of this study is to develop and evaluate a position based impedance control scheme using a Fuzzy Model Reference Learning Controller (FMRLC). In the proposed control scheme, the static relationship between displacement and force is quantified through a desired stiffness. The desired dynamic behaviour of the system is quantified through the use of a reference model. Through ensuring that the closed loop behaviour of the system is the same as the behaviour of the reference model, the dynamic relationship between motion and force can be regulated. When properly developed, the FMRLC has the ability to improve performance during repetitive tasks by learning through continued interaction with its environment in order to maintain desired and predictable behaviour.