A Key Challenge In Complex Visuomotor Control Is Learning Abstract Representations That Are Effective For Specifying Goals, Planning, And Generalization. To This End, We Introduce Universal Planning Networks (upn). Upns Embed Differentiable Planning Within A Goal-directed Policy. This Planning Computation Unrolls A Forward Model In A Latent Space And Infers An Optimal Action Plan Through Gradient Descent Trajectory Optimization. The Plan-by-gradient-descent Process And Its Underlying Representations Are Learned End-to-end To Directly Optimize A Supervised Imitation Learning Objective. We Find That The Representations Learned Are Not Only Effective For Goal-directed Visual Imitation Via Gradient-based Trajectory Optimization, But Can Also Provide A Metric For Specifying Goals Using Images. The Learned Representations Can Be Leveraged To Specify Distance-based Rewards To Reach New Target States For Model-free Reinforcement Learning, Resulting In Substantially More Effective Learning When Solving New Tasks Described Via Image-based Goals. We Were Able To Achieve Successful Transfer Of Visuomotor Planning Strategies Across Robots With Significantly Different Morphologies And Actuation Capabilities.
In order to provide you with the most up-to-date information possible in the Computer Vision Homepage, we have created a special page for all the submissions that have yet to be filed in their appropriate sub-pages. From time to time, our filing process g
G. Schreiber, A. Stemmer, и R. Bischoff. IEEE Workshop on Innovative Robot Control Architectures for Demanding (Research) Applications How to Modify and Enhance Commercial Controllers (ICRA 2010), стр. 15--21. Citeseer, (2010)
J. Hegenberg, L. Cramar, и L. Schmidt. Datenbrillen - Aktueller Stand von Forschung und Umsetzung sowie zukünftiger Entwicklungsrichtungen (Dortmund 2011), стр. 29-38. Dortmund, BAuA, (2012)
K. Balakrishnan, и V. Honavar. Genetic Programming 1997: Proceedings of the Second
Annual Conference, стр. 389--397. Stanford University, CA, USA, Morgan Kaufmann, (13-16 July 1997)