Machine intelligence for robotics applications

Abstract

The purpose of this paper is to review research in machine intelligence ongoing at the Center for Engineering Systems Advanced Research (CESAR). As a result of initial experimentation with our HERMIES-I mobile robot, hardware and software upgrades were implemented which enable fully asynchronous sonar operation, improved stepper motor control for the sensory platform, and more reliable wheel drive control. The current system, designated as HERMIES-II, is discussed. Successful demonstration of dead-reckoning navigation and the development of a sensor-based exploration and discovery algorithm which can now handle typical maze problems are reported. The development of HERMIES ''brain'' as a hypercube ensemble machine with concurrent computation and associated message passing is described. Algorithms for mapping precedence-constrained task graphs onto a hypercube yield results with high efficiency and proper load balance. A framework for a hybrid uncertainty analysis theory for decision making is described.