 |
Morphology, Evolution & Cognition Laboratory |
329 Votey Hall |
|
Burlington, VT 05405 |
|
Tel: (802) 656-4665; Cell: (802) 578-4445 |
|
Fax: (802) 656-0696 |
|
| Home | Research | Outreach | Publications | People |
| Behavior Chaining |
| One of the long-standing challenges in intelligent robotics is how to get a robot to autonomously exhibit multiple dynamic behaviors. For example, how does one create a robot capable of finding objects of interest, picking them up once it has found one, and transporting it somewhere else? |
| Robot Introspection |
| Without recourse to internal models, robots and animals both must resort to trial-and-error to learn sophisticated, and potentially dangerous behaviors. In one line of research we are investigating how robots can create approximate, yet physically-realistic simulations of their own bodies, and use them to predict the physical repercussions of as-of-yet untried actions. |
| Topological Modeling |
| Most automated modeling engines fix the underlying structure of the model, and then attempt to tune the parameters of the model such that it reproduces observed behavior. In topological modeling we attempt to optimize both the parameters as well as the underlying mathematical form of the model. |
| Evolving Brains and Bodies Simultaneously |
| The majority of work in robotics has focused on optimizing the robot's controller such that it exhibits some behavior of interest. In one line of research we are investigating how evolutionary algorithms can be used to design the robot's mechanical structure and controller simultaneously. |