Control in the reliable region of a statistical model with Gaussian process regression

We present a novel statistical model-based control algorithm, called Control in the Reliable Region of a Statistical Model (CRROS). A statistical model is unreliable when its state passes into a region where training data is sparse. CRROS drives the state away from such an unreliable region while pursuing the desired output by taking advantage of the redundancy in the input-output relationships. We validated the performance of CRROS by a simulation with a redundant manipulator and experiments with a robot. In the experiments, a manipulator called the Flex-finger, for which it is challenging to build an analytical model, is controlled to demonstrate the practical effectiveness of the proposed method.