Impedance-reflecting teleoperation with a real-time evolving neural network controller

A real-time evolving neural network controller (RTENNC) for controlling teleoperation systems with uncertainties is presented. Two different slave position controllers, the RTENNC and a proportional-integral-derivative (PID) controller, were experimentally compared using it one-degree-of-freedom test bed. With both controllers, force feedback to the master is obtained using measurements from a force sensor mounted on the slave manipulator. When various environments are explored with this system, the RTENNC provides better impedance reflection than the PID controller. For each environment, position tracking, force reflection, and the impedance ratio (the ratio of the sensed environment impedance to the impedance displayed to the human operator) were recorded in order to determine a measure of system transparency. "Adaptive transparency," the ability of simultaneous position following and force tracking between the master and slave for changing environment impedances, was higher for the RTENNC. The RTENNC resulted in an impedance ratio of approximately 1.0, whereas the PID controller had errors of about 20%.