Adaptive force control for the ARM manipulator

The development and modelling of a compliant device to enable compliant force control is described. The early stages of the development of the control formalism is also described, A PUMA 560 six degree-of-freedom manipulator, holding a crack detection probe, is required to make contact with a surface in space and then track along it using force feedback information. A novel design of compliant device has been built. For this case, the compliant device is an assembly of eight pre-tensioned springs situated between the manipulator end flange and the tool. A force sensor is mounted between the tool and the compliant device. It is necessary to be able to relate the recorded contact forces and torques to the changes in translations and orientations of the tool relative to the manipulator. Artificial neural networks (ANNs) have been used to model the compliant device and thus map the robot tool position in space. An experiment re-assesses adaptive compliant force control based on an ARMA model and using a simplified force detecting compliance. The next stage of work, using artificial neural networks (ANNs) to model the compliant devices' local stiffness characteristics and thus selecting suitable controller gains is described, It should be noted that these two sections use very different compliant devices and so are not quantitatively comparable.