Uniform Error Bounds for Gaussian Process Regression with Application to Safe Control

Data-driven models are subject to model errors due to limited and noisy training data. Key to the application of such models in safety-critical domains is the quantification of their model error. Gaussian processes provide such a measure and uniform error bounds have been derived, which allow safe control based on these models. However, existing error bounds require restrictive assumptions. In this paper, we employ the Gaussian process distribution and continuity arguments to derive a novel uniform error bound under weaker assumptions. Furthermore, we demonstrate how this distribution can be used to derive probabilistic Lipschitz constants and analyze the asymptotic behavior of our bound. Finally, we derive safety conditions for the control of unknown dynamical systems based on Gaussian process models and evaluate them in simulations of a robotic manipulator.

[1]  Andreas Krause,et al.  Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting , 2009, IEEE Transactions on Information Theory.

[2]  Aditya Gopalan,et al.  On Kernelized Multi-armed Bandits , 2017, ICML.

[3]  P. Olver.  Nonlinear Systems , 2013 .

[4]  R. Dudley.  The Sizes of Compact Subsets of Hilbert Space and Continuity of Gaussian Processes , 1967 .

[5]  Neil D. Lawrence,et al.  Batch Bayesian Optimization via Local Penalization , 2015, AISTATS.

[6]  Sandra Hirche,et al.  Posterior Variance Analysis of Gaussian Processes with Application to Average Learning Curves , 2019, ArXiv.

[7]  Sandra Hirche,et al.  Gaussian processes for dynamic movement primitives with application in knowledge-based cooperation , 2016, 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[8]  R. Adler,et al.  Random Fields and Geometry , 2007 .

[9]  Jan Peters,et al.  A Survey on Policy Search for Robotics , 2013, Found. Trends Robotics.

[10]  Sandra Hirche,et al.  Bayesian uncertainty modeling for programming by demonstration , 2017, 2017 IEEE International Conference on Robotics and Automation (ICRA).

[11]  Sandra Hirche,et al.  Scenario-based Optimal Control for Gaussian Process State Space Models , 2018, 2018 European Control Conference (ECC).

[12]  Andreas Krause,et al.  Safe learning of regions of attraction for uncertain, nonlinear systems with Gaussian processes , 2016, 2016 IEEE 55th Conference on Decision and Control (CDC).

[13]  Jeremy Levesley,et al.  Error bounds for anisotropic RBF interpolation , 2010, J. Approx. Theory.

[14]  Sandra Hirche,et al.  Feedback linearization using Gaussian processes , 2017, 2017 IEEE 56th Annual Conference on Decision and Control (CDC).

[15]  S. Ghosal,et al.  Posterior consistency of Gaussian process prior for nonparametric binary regression , 2006, math/0702686.

[16]  Ingo Steinwart,et al.  On the Influence of the Kernel on the Consistency of Support Vector Machines , 2002, J. Mach. Learn. Res..

[17]  Shahar Mendelson,et al.  Improving the sample complexity using global data , 2002, IEEE Trans. Inf. Theory.

[18]  Andreas Krause,et al.  Safe Model-based Reinforcement Learning with Stability Guarantees , 2017, NIPS.

[19]  P. Massart,et al.  Adaptive estimation of a quadratic functional by model selection , 2000 .

[20]  C. F. Jeff Wu,et al.  On Prediction Properties of Kriging: Uniform Error Bounds and Robustness , 2017, Journal of the American Statistical Association.

[21]  J. Mercer.  Functions of positive and negative type, and their connection with the theory of integral equations , 1909 .

[22]  Simon Hubbert,et al.  Lp-error estimates for radial basis function interpolation on the sphere , 2004, J. Approx. Theory.

[23]  Zongmin Wu,et al.  Local error estimates for radial basis function interpolation of scattered data , 1993 .

[24]  Holger Wendland,et al.  Scattered Data Approximation: Conditionally positive definite functions , 2004 .

[25]  Dino Sejdinovic,et al.  Gaussian Processes and Kernel Methods: A Review on Connections and Equivalences , 2018, ArXiv.

[26]  Sandra Hirche,et al.  An Uncertainty-Based Control Lyapunov Approach for Control-Affine Systems Modeled by Gaussian Process , 2018, IEEE Control Systems Letters.

[27]  Lei Shi.  Learning theory estimates for coefficient-based regularized regression , 2013 .

[28]  Fernando A. Mujica,et al.  An Empirical Evaluation of Deep Learning on Highway Driving , 2015, ArXiv.

[29]  Robert Schaback,et al.  Improved error bounds for scattered data interpolation by radial basis functions , 1999, Math. Comput..

[30]  Dana Kulic,et al.  Stable Gaussian Process based Tracking Control of Euler-Lagrange Systems , 2018, Autom..

[31]  Andrew M. Stuart,et al.  Posterior consistency for Gaussian process approximations of Bayesian posterior distributions , 2016, Math. Comput..

[32]  Harry van Zanten,et al.  Information Rates of Nonparametric Gaussian Process Methods , 2011, J. Mach. Learn. Res..

[33]  Niels Kjølstad Poulsen,et al.  Neural Networks for Modelling and Control of Dynamic Systems: A Practitioner’s Handbook , 2000 .

[34]  Tong Zhang,et al.  Learning Bounds for Kernel Regression Using Effective Data Dimensionality , 2005, Neural Computation.

[35]  F. J. Narcowich,et al.  Sobolev Error Estimates and a Bernstein Inequality for Scattered Data Interpolation via Radial Basis Functions , 2006 .

[36]  John Shawe-Taylor,et al.  Regret Bounds for Gaussian Process Bandit Problems , 2010, AISTATS 2010.

[37]  Daniel J. Hsu,et al.  Kernel ridge vs. principal component regression: Minimax bounds and the qualification of regularization operators , 2017 .

[38]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[39]  Ameet Talwalkar,et al.  On the Impact of Kernel Approximation on Learning Accuracy , 2010, AISTATS.

[40]  M. Talagrand.  Sharper Bounds for Gaussian and Empirical Processes , 1994 .

[41]  Richard M. Murray,et al.  A Mathematical Introduction to Robotic Manipulation , 1994 .

[42]  Sandra Hirche,et al.  Learning stable Gaussian process state space models , 2017, 2017 American Control Conference (ACC).