Application of GA based neuro - fuzzy automatic generation for teleoperation systems

This paper discusses the design of a controller for teleoperation systems which operate in contact with unknown environments. For this purpose, the controller in the slave part is designed to use a classical controller and a neuro-fuzzy controller. To have a haptic teleoperation, the force feedback from the contact point to the master side is needed. In this study virtual force is fed back to the master instead of the measured contact force and the feedback force is the position error i.e. the difference between the position of master and slave robots. Primary, classical controller was designed for this object, the gains of this controller is obtained by standard genetic algorithm (GA), in offline and optimization method at nominal operating point. In master-slave robot with classical controller, when, robot works in soft or hard environment and the operating point was changed, the obtained controller are not suitable. For this reason, online neuro - fuzzy gain scheduling controller are designed for this system to tune the controller´s parameters. The NF trained in offline mode based on GA and in online application tuned the controller based on monitoring of operating point of system. This method makes the system robust to uncertainties such as modeling errors, parameter mismatches and disturbances. The simulation results show the ability of proposed method in compare of other controllers such as Kalman filter.