As autonomous robots become increasingly multifunctional and adaptive, it becomes difficult to determine the extent of their capabilities, i.e. the tasks they can perform and their strengths and limitations at these tasks. A robot’s appearance can provide cues to its physical as well as cognitive capabilities. We present an algorithm that builds on these cues and learns models of a robot’s ability to perform different tasks through active experimentation. These models not only capture the robot’s inherent abilities but also incorporate the effect of relevant extrinsic factors on a robot’s performance. Our algorithm would find use as a tool for humans in determining “What can this robot do?”.

We applied our algorithm in modeling a NAO and a Pepper robot at two different tasks. We first illustrate the advantages of our active experimentation approach over building models through passive observations. Next, we show the utility of such models in identifying scenarios a robot is well suited for in performing a task. Finally, we demonstrate the use of such models in a collaborative human-robot task.

Committee:

Professor Manuela Veloso (chair)

Professor Jean Oh

Devin Schwab