Real-time robot deliberation by compilation and monitoring of anytime algorithms

Anytime algorithms are algorithms whose quality of results improves gradually as computation time increases. Certainty, accuracy, and specificity are metrics useful in anytime algorighm construction. It is widely accepted that a successful robotic system must trade off between decision quality and the computational resources used to produce it. Anytime algorithms were designed to offer such a trade off. A model of compilation and monitoring mechanisms needed to build robots that can efficiently control their deliberation time is presented. This approach simplifies the design and implementation of complex intelligent robots, mechanizes the composition and monitoring processes, and provides independent real time robotic systems that automatically adjust resource allocation to yield optimum performance.