Control of a pneumatic gantry robot for grinding: a neuro-fuzzy approach to PID tuning

This paper addresses an application that involves the grinding of the edges of steel blanks with a pneumatic gantry robot. It presents a PID tuning method that uses an adaptive neuro-fuzzy inference system (ANFIS) to model the relationship between the controller gains and the target output response, with the response specification set by desired percent overshoot and settling time. The ANFIS based input-output model is then used to tune on-line the PID gains for different response specifications. Experimental results demonstrate that better performance can be achieved with ANFIS tuning relative to 1) Ziegler-Nichols tuning and 2) trial and error tuning. The actual response specifications with ANFIS matched the desired response specifications with an accuracy of 95%