Evaluation of a Multi-Goal Solver for Use in a Blackboard Architecture

Abstract

This article presents a multi-goal solver for problems that can be modeled using a Blackboard Architecture. The Blackboard Architecture can be used for data fusion, robotic control and other applications. It combines the rule-based problem analysis of an expert system with a mechanism for interacting with its operating environment. In this context, numerous control or domain (system-subject) problems may exist which can be solved through reaching one of multiple outcomes. For these problems which have multiple solutions, any of which constitutes an end-goal, a solving mechanism which is solution-choice-agnostic and finds the lowest-cost path to the lowest-cost solution is required. Such a solver mechanism is presented and characterized herein. The performance of the solver (including both the computational time required to ascertain a solution and execute it) is compared to the naïve Blackboard approach. This performance characterization is performed across multiple levels of rule counts and rule connectivity. The naïve approach is shown to generate a solution faster, but the solutions generated by this approach, in most cases, are inferior to those generated by the solver.