Adaptive Reinforcement Learning: Increasing the applicability for large and time varying systems using parallel Gaussian Process regression and adaptive nonlinear control