Using associative content-addressable memories to control robots

Abstract

It has been shown that an associative content-addressable memory can be used to model a robot and the world with which the robot interacts. The model can be learned by storing experiences in the memory. To make predictions the memory is searched in parallel for relevant experience. An initial implementation of such a memory-based modeling scheme was made on a parallel computer, the connection machine. The implementation was used to model and control a simulated planar two-joint arm and a simulated running machine. The authors describe the issues and problems that arose in this preliminary work.