The primary objective of the REAKT and REAKT II project is to develop a set of tools and the associated methodology for applying knowledge-based systems in real-time domains.
The primary aim of the project was to develop tools and techniques for applying knowledge based systems technology in real time domains.
The project's major achievement so far is a software prototype in LISP of an environment for building real time knowledge based systems.
A multitasking architecture has been designed, based on the blackboard paradigm, where a control component, a communication manager, a timer and a number of user defined domain agents run as different processes. Agents are created dynamically to react to certain events and communicate through a shared data structure (the blackboard). Agents may have deadlines which are considered when scheduling the execution of the agent processes, in order to guarantee response times. When more than 1 agent is required to solve a particular problem, the application developer can predefine sequences or even plans of agents activations. A time map manager has been developed to handle temporal constraints and relationships between future variables values (predictions). These predictions are used by domain agents to do some 'advance work' when time is available. Validation scenarios for the new functionalities have been developed and are currently evolving into the final demonstrator, an oil refinery process control application. A runtime version of the architecture in C/C++ is under implementation. This C runtime will be faster than its LISP counterpart in the development environment, and it will include its own memory management algorithms in order to comply with the real time constraints of the demonstrator.
The project has also produced methodological guidelines, collected into an application methodology handbook.
The project was to produce definitions, specifications and prototypes of various techniques, to be eventually integrated into a toolkit for the efficient development, deployment and maintenance of knowledge-based modules embeddable into real-time applications. Research areas of particular interest include progressive reasoning techniques, which are identified as a key element to provide guaranteed response time, as well as temporal reasoning and coherence management mechanisms. The methodological work is mainly focussed on the implications of real-time issues on the modelling and development of knowledge-based systems.
A demonstrator system in the area of process control has been chosen to validate both the toolkit and the methodology. Its components exhibit various real-time characteristics in order to demonstrate that the technology is flexible enough to span a large domain of operations. By the end of a follow-on project, 7805, the set of tools and the methodology it will provide should increase the range of applications in which real-time knowledge-based systems could be integrated, thus increasing the market size for both this kind of products and the applications built on top of them.
Funding Schemeundefined - undefined
LE3 1UF Leicester