Learning References with Gaussian Processes in Model Predictive Control applied to Robot Assisted Surgery

One of the key benefits of model predictive control is the capability of controlling a system proactively in the sense of taking the future system evolution into account. However, often external disturbances or references are not a priori known, which renders the predictive controllers shortsighted or uninformed. Adaptive and learning based prediction models can provide suitable predictions to the controller and therefor can be applied to overcome this issue. We propose to learn references for model predictive controllers via Gaussian processes. To illustrate the approach, we consider robot assisted surgery, where a robotic manipulator must follow a learned reference position based on optical tracking measurements.