Cosyne 2007 Workshops
February 26-27, 2007
The Canyons, Utah
Reducing the complexity of sensorimotor control
Abstract
Acting in a complex and uncertain environment is a hard problem which the sensorimotor system solves with apparent ease. Can we understand how it does that, to the point of being able to build synthetic systems with comparable performance? Many hypothetical strategies have been inferred from analyses of simple movements. However some of them are too simplistic to solve complex control problems while others call for computations that are generally intractable. Is there a middle ground, which can enable brains and robots to achieve high performance without a prohibitive amount of computation? To address this question, the workshop will bring together a diverse group of researchers focusing on motor behavior, neurophysiology, computational modeling, and robotics. The talks will cover a range of topics including neural representations that facilitate computation, construction of complex skills from motor primitives, combining reward-based and imitation-based learning, adaptation on multiple spatial and temporal scales.
Program
| 8:30 | Opening remarks | |||
| 8:40 | Dana Ballard | Equilibrium Points: A Unifying Abstraction for Sensory-Motor Control | ||
| 9:10 | Andrea D'Avella | Muscle synergies as building blocks for sensorimotor control | ||
| 9:40 | Emo Todorov | Compositionality in optimal control | ||
| 10:10 | --- break --- | |||
| 10:30 | Andrew Ng | Reinforcement learning and apprenticeship learning for robotic control | ||
| 11:00 | Dinesh Pai | Biomechanisms for sensorimotor control | ||
| 11:30 | --- skiing --- | |||
| 4:00 | Konrad Koerding | Inverse optimal control | ||
| 4:30 | Mark Churchland | Neural dynamics during movement planning and execution | ||
| 5:00 | Maurice Smith | A multi-timescale model of learning, retention and generalization | ||
| 5:30 | --- break --- | |||
| 5:50 | Stefan Schaal | Reinforcement learning for task-level control | ||
| 6:20 | Philip Sabes | The eyes don't have it (all): Optimal integration of reach planning across "coordinate frames" | ||
| 6:50 | General discussion |
Organizer
Emo Todorov (UCSD): todorov@cogsci.ucsd.edu
