학위논문 (박사)-- 서울대학교 대학원 : 기계항공공학부, 2015. 2. 박종우. ; This thesis is concerned with motion learning for complex, high-dimensional robot systems operating in unstructured environments, and subject to various task constraints whose analytical characterization may not be a priori available. We first present a Gaussian process algorithm for learning the configuration space of a robot subject to holonomic task constraints. Given an observed data set of points that lie on this task constrained configuration space, or constraint manifold, a point-to- manifold distance function is constructed that measures the distance of any given point from the constraint manifold. The observed data are first encoded using a Gaussian mixture model, and the distance function is learned via Gaussian process regression. The constructed distance function admits an explicit representation that can be differentiated to obtain analytic gradients. We apply this distance function and its gradient to a sampling-based path planning problem for a robot performing a constrained task. We also propose an efficient method for generating suboptimal motions for multibody systems using Gaussian process dynamical models. Given a dynamical model for a multibody system, and a trial motion, a lower-dimensional Gaussian process dynamical model is fitted to the trial motion. New motions are then generated by performing a dynamic optimization in the lower-dimensional space. We introduce the notion of variance tubes as an intuitive and efficient means of restricting the optimization search space. The performance of our algorithm is evaluated through detailed case studies of raising motions for an arm, swing, pitching and jumping motions for a humanoid and lifting motions for a mobile manipulator. ; Contents Abstract List of Tables List of Figures 1 Introduction 1 1.1 Contributions of This Thesis 1.1.1 Configuration Space Learning for Constrained Robot Tasks 1.1.2 Robot Motion Optimization via Gaussian Process Dynamical Model 1.2 Organization 2 Preliminaries 2.1 Introduction 2.2 Robot Modeling 2.2.1 Geometric Dynamics of a Rigid Body 2.2.2 Robot Dynamics Algorithm 2.3 Gaussian Model Learning 2.3.1 Gaussian Mixture Model 2.3.2 Gaussian Process Models 3 Configuration Space Learning for Constrained Robot Tasks 3.1 Introduction 3.2 Point-to-Manifold Distance Function 3.2.1 Determining the Weight Function ω k (q) 3.2.2 Determining the Distance Function d k (q) 3.2.3 Properties 3.3 Experiments and Discussion 3.3.1 Distance Function Learning 3.3.2 Comparison with Other Learning Methods 3.3.3 Constraint Learning for a Robot Arm 4 Robot Motion Optimization via Gaussian Process Dynamical Model 4.1 Introduction 4.2 General motion optimization algorithms 4.3 GPDM-based Motion Optimization Algorithms 4.3.1 Formulation 4.3.2 Search Space in Latent Space 4.3.3 Collision Avoidance 4.3.4 Algorithm 4.4 Experiments and Discussion 4.4.1 Arm Motions 4.4.2 Humanoid Simulations 4.5 Mobile Manipulator Experiments 4.5.1 Lifting Motion Optimization 4.5.2 Collision Avoidance Simulation 5 Conclusion A Appendix A.1 Dynamics of a Rigid Body A.2 Recursive Dynamics Algorithms A.2.1 Recursive Forward Dynamics Algorithm A.2.2 Recursive Inverse Dynamics Algorithm A.2.3 Recursive Hybrid Dynamics Algorithm A.2.4 Inverse and Hybrid Dynamics with Contact Model Bibliography Abstract ; Doctor