Parallel processing for real-time rule-based decision aids

The rapid technology development in the past two decades has made today´s combat a complicated task. A large amount of information can be available in a mission from both on-board and off-board sources. Effectively utilizing the information is necessary to achieve successful and optimal results. Decision aids that operate in real-time are an important issue as all DoD components strive to reduce the crew size of their various weapon systems. The decision aids will help reduce the workload of the crew and increase the efficiency and reliability of operations. Since a large number of criteria and rules must be evaluated and checked in a very short time period in combat automation, parallel processing may be needed in order to meet the timing requirement. In this study, we investigate a parallel processing technique for complicated decision aids that employ two-state rule-based systems. The study focuses on evaluation of decision rules, although that is just part of the decision aids problem. The rule base is decomposed into subsets for individual processing units. The rule-checking task is distributed to multiple processors to speed up the response. One merit of the explored technique is the scaleability. The number of processors can be altered based on the processing load and the availability of processors. The Intel Paragon high performance computer (a 2-dimensional mesh processor architecture) is selected for experiments. This paper introduces the data structures for rule bases and the proposed parallel processing technique. Experimental results using different numbers of processors are presented