Comments from the break-out discussion 8/7/2011

"Off-the-shelf platforms for AI education and research:
What's available? What's ideal?

Introductions:

What are platforms you've used?
lab-specific Segway with WAM arm
Aibo
iRobot Create
Lego robotics
small arms and humanoids
Trrstan robots: based on audio output from a phone/mp3 player

What are their positive/negative characteristics?

- difficult to use your own software (especially AI software) on new hardware
- hardware can be a time sink that detracts from the concepts at hand
- lots to learn with hardware and software, e.g., ROS + PR2
- cost can be prohibitive
- battery life can be frustratingly short
- repairs can be prohibitive/expensive
- need standards among platforms to help sharing curriculum and software


Additional prompts:

What features are critical in a platform?

+ need to be easy to install
+ need to _work_ for the task at hand
+ it should be easy/easier to connect AI & robots
+ there is interest in stronger curricular support
+ it's important that the software abstract away the details not
central to your goals, e.g., low-level control within an AI course

Manipulation:

+ fingers are much better than grippers, but this is a
hardware/engineering problem, not an AI problem at the moment

Is there a sweet spot in the price?

U. Kassell:  $50 is the aim, e.g., "textbook price"
HMC: $500 is the goal (a multi-year investment for small teams)