Machine Conscious Architecture for State Exploitation and Decision Making

Abstract : This research addressed a critical limitation in the area of computational intelligence by developing a general purpose architecture for information processing and decision making. Traditional computational intelligence methods are best suited for well-defined problems with extensive, long-term knowledge of the environmental and operational conditions the system will encounter during operation. These traditional approaches typically generate quick answers (i.e., reflexive responses) using pattern recognition methods. Most pattern recognition techniques are static processes which consist of a predefined series of computations. For these pattern recognition approaches to be effective, training data is required from all anticipated environments and operating conditions. The proposed framework, Conscious Architecture for State Exploitation (CASE), is a general purpose architecture designed to mimic key characteristics of human information processing. CASE combines low- and high-level cognitive processes into a common framework to enable goal-based decision making. The CASE approach is to generate artificial phenomenal states (i.e., generate qualia = consciousness) into a shared computational process to enhance goal-based decision making and adaptation. That is, this approach allows for the appropriate decision and corresponding adaptive behavior as the goals and environmental factors change. To demonstrate the engineering advantages of CASE, it was used in an airframe application to autonomously monitor the integrity of a flight critical structural component. In this demonstration, CASE automatically generated a timely maintenance recommendation when unacceptable cracking was detected. Over the lifetime of the investigated component, operational availability increased by a minimum of 10.7%, operational cost decreased by 79%, and maintenance intervals (i.e., MTBM) increased by a minimum of 900%.