LEARNING CONTROLLERS FOR INDUSTRIAL ROBOTS

One of the most significant cost factors in robotics applications is the design and development of real-time robot control software. Control theory helps when linear controllers have to be developed, but it doesn't sufficiently support the generation of non-linear controllers, although in many cases (such as in compliance control), nonlinear control is essential for achieving high performance. This paper discusses how Machine Learning has been applied to the design of (non-)linear controllers. Several alternative function approximators, including Multilayer Perceptrons (MLP), Radial Basis Function Networks (RBFNs), and Fuzzy Controllers are analyzed and compared, leading to the definition of two major families: Open Field Function Function Approximators and Locally Receptive Field Function Approximators. It is shown that RBFNs and Fuzzy Controllers bear strong similarities, and that both have a symbolic interpretation. This characteristics allows for applying both symbolic and statistic learning algorithms to synthesize the network layout from a set of examples and, possibly, some background knowledge. Three integrated learning algorithms, two of which are original, are described and evaluated on experimental test cases. The first test case is provided by a robot KUKA IR-361 engaged into the "peg-into-hole" task, whereas the second is represented by a classical prediction task on the Mackey-Glass time series. From the experimental comparison, it appears that both Fuzzy Controllers and RBFNs synthesised from examples are excellent approximators, and that, in practice, they can be even more accurate than MLPs.