view thesis/abstract.org @ 440:b01c070b03d4

save for tonight.
author Robert McIntyre <rlm@mit.edu>
date Sun, 23 Mar 2014 23:43:20 -0400
parents c1e6b7221b2f
children c20de2267d39
line wrap: on
line source
1 Here I demonstrate the power of using embodied artificial intelligence
2 to attack the /action recognition/ problem, which is the challenge of
3 recognizing actions performed by a creature given limited data about
4 the creature's actions, such as a video recording. I solve this
5 problem in the case of a worm-like creature performing actions such as
6 curling and wiggling.
8 To attack the action recognition problem, I developed a computational
9 model of empathy (=EMPATH=) which allows me to use simple, embodied
10 representations of actions (which require rich sensory data), even
11 when that sensory data is not actually available. The missing sense
12 data is ``imagined'' by the system by combining previous experiences
13 gained from unsupervised free play.
15 In order to build this empathic, action-recognizing system, I created
16 a program called =CORTEX=, which is a complete platform for embodied
17 AI research. It provides multiple senses for simulated creatures,
18 including vision, touch, proprioception, muscle tension, and
19 hearing. Each of these senses provides a wealth of parameters that are
20 biologically inspired. =CORTEX= is able to simulate any number of
21 creatures and senses, and provides facilities for easily modeling and
22 creating new creatures. As a research platform it is more complete
23 than any other system currently available.