Mercurial > cortex
view thesis/abstract.org @ 525:25f23cfd56ce
alter pics, enhance text.
author | Robert McIntyre <rlm@mit.edu> |
---|---|
date | Mon, 21 Apr 2014 02:13:23 -0400 |
parents | 8e52a2802821 |
children | d304b2ea7c58 |
line wrap: on
line source
1 Here I demonstrate the power of using embodied artificial intelligence2 to attack the /action recognition/ problem, which is the challenge of3 recognizing actions performed by a creature given limited data about4 the creature's actions, such as a video recording. I solve this5 problem in the case of a worm-like creature performing actions such as6 curling and wiggling.8 To attack the action recognition problem, I developed a computational9 model of empathy (=EMPATH=) which allows me to recognize actions using10 simple, embodied representations of actions (which require rich11 sensory data), even when that sensory data is not actually available.12 The missing sense data is ``imagined'' by the system by combining13 previous experiences gained from unsupervised free play. The worm is a14 five-segment creature equipped with touch, proprioception, and muscle15 tension senses. It recognizes actions using only proprioception data.17 In order to build this empathic, action-recognizing system, I created18 a program called =CORTEX=, which is a complete platform for embodied19 AI research. It provides multiple senses for simulated creatures,20 including vision, touch, proprioception, muscle tension, and hearing.21 Each of these senses provides a wealth of parameters that are22 biologically inspired. =CORTEX= is able to simulate any number of23 creatures and senses, and provides facilities for easily modeling and24 creating new creatures. As a research platform it is more complete25 than any other system currently available.