Robotic Tasks Using Soar

Soar's ability to modulate perception and action through the use of encoding and decoding productions makes it applicable to robotic domains. An advantage of using Soar in these domains is that its learning mechanism, chunking, can be used to cache the results of previous problem solving and planning into more reactive behavior. This is similar to the generation of stimulus-response rules in Theo and exhibits a corresponding speedup as problematic behavior becomes routine.

Robo-Soar (Laird and Rosenbloom, 1990) uses a camera vision system and a Puma robot arm to manipulate -- stack, unstack, move -- blocks in a small but real blocksworld domain. Similarly, Hero-Soar is a mobile robot implementation which navigates in an office environment and uses an extendable arm to pick up empty cups. These systems demonstrate Soar's applicability to domains with uncertainty, where perception can be incomplete, and unpredictable, since dynamic, unexpected events may occur to interrupt the current processing and behavior.

To go to the Soar capabilities list, press UP

To return, press HOME. To go to the next document, press NEXT.