Task-Improving Robots through Observation

and Learning

 

 

Manuela Veloso

Carnegie Mellon University

 

 

 

Robots are physical artifacts with a seamless integration of perception, cognition, and action. My presentation will be focused on learning robots performing tasks in non-stationary domains. Robots need to jointly assess the state of their environment, communicate with each other, make decisions, execute actions towards the achievement of team objectives, and learn from observation and feedback based on the outcome of their actions.