The ability of an aircraft crew to maintain situation awareness is recognized in military operations as crucial to mission success and survivability. Aircraft crews are normally confronted with numerous displays and complex control tasks in their cockpit, which may lead to circumstances of heavy workload, and hence non-optimal decisions are made. These circumstances can be avoided by employing computerized aids, with a view to reducing information overload, which efficiently process and filter out irrelevant information before concise information is presented to the crew. Thus, computerized aids are critical in assisting the crew to be effectively aware of real-time dynamic situations of the environment. Automated decision aids are also useful to help the crew improving efficiency and flight safety. An on-board automated system that performs these decision aids is referred to crew assistant system.
The design of a crew assistant system must therefore be guided by the goal of maintaining and enhancing crew situation awareness. There exist a number of crew assistant systems such as Pilot’s Associate, SRI/Grumman decision support system. These systems followed a deterministic approach where the problem domain is carefully articulated and a matching solution domain is anticipated. In other words, the automation of routine tasks or goal/plan pairs is pre-programmed with a view to providing aids to the crew in solving problems and managing their tasks successfully. These systems are deterministically static, and hence they cannot cope with a dynamic environment where unforeseen situations often occur such as on a battlefield. Furthermore, these systems are implemented in the well-known framework of Blackboard Architecture where knowledge sources which interact with each other via a common backboard are distributed. The major benefit of the Backboard Architecture is the ease of controlling the interaction among knowledge sources. On the other hand, the central control will pay penalties on the computational cost and on input-output transmission which are bottlenecks for real-time performance.
The system for situation awareness proposed in this paper follows a different approach where the multiagent technology is applied as a framework for implementation. The architecture is based on multiple cooperative agents; each performs one defined function. Rather than being static with pre-defined goals/plans, the agent has its own knowledge and inference: rules as resources for resolving variously dynamic circumstances.