This paper experimentally implements a surface following algorithm for unknown surfaces under a hybrid position and force scheme. The mechanization uses a low friction roller at the end of a two-link manipulator so that primarily the normal force is measured. With the normal force information and joint angle information, the surface normal direction is calculated and a position reference is calculated along the surface tangent. Experimental evaluation shows the algorithm works for a line, a concave arc, and a convex arc. Also, when different force levels are commanded or different controllers are used, the algorithm still performs well. Thus, the algorithm allows a constant force to be applied while following a surface whose shape is not known a priori.

This paper compares the experimental evaluation of robust reduced-order hybrid position and force controllers to standard full-order and reduced-order linear-quadratic Gaussian (LQG) controllers. A robust reduced-order position controller and a reduced-order force controller are designed using the SANDY direct optimization program. Full-order LQG position and force controllers are designed and then reduced through balanced truncation. Experimental results show that the reduced-order controllers provide comparable performance to the full-order controllers. Thus, by use of direct-optimization, the benefits of smaller controllers (smaller memory requirements and faster sampling rates) can be achieved while accounting for varying plant conditions (robustness).

An iterative constrained inversion technique is used to find the control inputs to the plant. That is, rather than training a controller network and placing this network directly in the feedback or feedforward paths, the forward model of the plant is learned, and iterative inversion is performed on line to generate control commands. The control approach allows the controllers to respond online to changes in the plant dynamics. This approach also attempts to avoid the difficulty of analysis introduced by most current neural network controllers, which place the highly nonlinear neural network directly in the feedback path. A neural network-based model reference adaptive controller is also proposed for systems having significant dynamics between the control inputs and the observed (or desired) outputs and is demonstrated on a simple linear control system. These results are interpreted in terms of the need for a dither signal for on-line identification of dynamic systems.