Task-space feedback information such as visual feedback is used in many modern robot control systems as it improves robustness to model uncertainty. However, existing task-space feedback control schemes are only valid locally in a finite task space within a limited sensing zone where the singularity of the Jacobian matrix can be avoided. The global stability problem of task-space control system...

We propose a representation for the set of forces robot can counteract using full system dynamics: residual force polytope. Given nominal torques required by dynamic motion, this models which be sustained without interfering with that motion. The polytope used to analyze and compare admissible different trajectories, but it also define metrics solving optimization problems, such as in trajector...

Authors' current addresses: A.C. Sanderson, Department of Electrical, Computer and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180; M.A. Peshkin, Department of Mechanical Engineering, The Technological Institute, Northwestern University, 2145 Sheridan Rd., Evanston, IL 60201; L.S. Homem de Mello, Robotics Institute, Department of Electrical and Computer Engineering, Camegi...

Several kernel based methods for multi-task learning have been proposed, which leverage relations among tasks as regularization to enhance the overall learning accuracies. These methods assume that the tasks share the same kernel, which could limit their applications because in practice different tasks may need different kernels. The main challenge of introducing multiple kernels into multiple ...

Learning complex skills by repeating and generalizing expert behavior is a fundamental problem in robotics. A common approach is learning from demonstration: given examples of correct motions, learn a policy mapping state to action consistent with the training data. However, the usual approaches do not answer the question of what are appropriate representations to generate motions for specific ...