Context Estimation and Learning Control through Latent Variable Extraction: From discrete to continuous contexts Export

@inproceedings{citeulike:2629260, abstract = {Recent advances in machine learning and adaptive motor control have enabled efficient techniques for online learning of stationary plant dynamics and it's use for robust predictive control. However, in realistic domains, system dynamics often change based on unobserved external contexts such as work load or contact conditions with other objects. Previous multiple model approaches to solving this problem are restricted to finite, discrete contexts without any generalization and have been tested only on linear systems. We present a framework for estimation of context through hidden latent variable extraction - solely from experienced (non-linear) dynamics. This work refines the multiple model formalism to bootstrap context separation from context-unlabeled data and enables simultaneous online context estimation, dynamics learning and control based on a consistent probabilistic formulation. Most importantly, it extends the framework to a continuous latent model representation of context under specific assumptions of load distribution}, author = {Petkos, G. and Vijayakumar, S.}, booktitle = {Robotics and Automation, 2007 IEEE International Conference on}, citeulike-article-id = {2629260}, citeulike-linkout-0 = {http://dx.doi.org/10.1109/ROBOT.2007.363634}, citeulike-linkout-1 = {http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4209398}, doi = {10.1109/ROBOT.2007.363634}, journal = {Robotics and Automation, 2007 IEEE International Conference on}, keywords = {context, function\_approximation, high\_dimensions, nonlinear}, pages = {2117--2123}, posted-at = {2008-04-04 11:36:26}, priority = {0}, title = {Context Estimation and Learning Control through Latent Variable Extraction: From discrete to continuous contexts}, url = {http://dx.doi.org/10.1109/ROBOT.2007.363634}, year = {2007} }

TY - CONF ID - citeulike:2629260 L3 - citeulike-article-id:2629260 N2 - Recent advances in machine learning and adaptive motor control have enabled efficient techniques for online learning of stationary plant dynamics and it's use for robust predictive control. However, in realistic domains, system dynamics often change based on unobserved external contexts such as work load or contact conditions with other objects. Previous multiple model approaches to solving this problem are restricted to finite, discrete contexts without any generalization and have been tested only on linear systems. We present a framework for estimation of context through hidden latent variable extraction - solely from experienced (non-linear) dynamics. This work refines the multiple model formalism to bootstrap context separation from context-unlabeled data and enables simultaneous online context estimation, dynamics learning and control based on a consistent probabilistic formulation. Most importantly, it extends the framework to a continuous latent model representation of context under specific assumptions of load distribution JF - Robotics and Automation, 2007 IEEE International Conference on EP - 2123 TI - Context Estimation and Learning Control through Latent Variable Extraction: From discrete to continuous contexts T2 - Robotics and Automation, 2007 IEEE International Conference on SP - 2117 KW - context KW - function_approximation KW - high_dimensions KW - nonlinear AU - Petkos, G AU - Vijayakumar, S PY - 2007/// UR - http://dx.doi.org/10.1109/ROBOT.2007.363634 ER -