Learn Fast, Forget Slow: Safe Predictive Learning Control for Systems with Unknown and Changing Dynamics Performing Repetitive Tasks

We present a control method for improved repetitive path following for a\nground vehicle that is geared towards long-term operation where the operating\nconditions can change over time and are initially unknown. We use weighted\nBayesian Linear Regression (wBLR) to model the unknown dynamics, and show how\nthis simple model is more accurate in both its estimate of the mean behaviour\nand model uncertainty than Gaussian Process Regression and generalizes to novel\noperating conditions with little or no tuning. In addition, wBLR allows us to\nuse fast adaptation and long-term learning in one, unified framework, to adapt\nquickly to new operating conditions and learn repetitive model errors over\ntime. This comes with the added benefit of lower computational cost, longer\nlook-ahead, and easier optimization when the model is used in a stochastic\nModel Predictive Controller (MPC). In order to fully capitalize on the long\nprediction horizons that are possible with this new approach, we use Tube MPC\nto reduce the growth of predicted uncertainty. We demonstrate the effectiveness\nof our approach in experiment on a 900\\,kg ground robot showing results over\n3.0\\,km of driving with both physical and artificial changes to the robot's\ndynamics. All of our experiments are conducted using a stereo camera for\nlocalization.\n