Rayner, N.J.W. and Harris, C.J.
Neurofuzzy Mission Management System for Multiple Autonomous Vehicles.
Proc. Conf on The Role of Intelligent Systems in Defence
Full text not available from this repository.
This paper is concerned with the engineering of Mission Management Knowledge Based processes for the command of multiple Intelligent Autonomous Vehicles. The processing performed in this component of the IAV architecture is concerned with the ordering of robot activities over time and their monitoring during execution. In order for a Knowledge Based Planner to build plans it must have a representation of the knowledge associated with performing actions in the real world. These actions are represented by operators which define the causal relationship between an actions conditions of initiation and the resultant effects of its execution. In this work Neuro-fuzzy causal models have been developed using continuous B-spline representations of fuzzy predicates implemented on an extended popular horn clause logic programming language PROLOG. In addition attempts have been made to support adaption of the causal knowledge base through the modification of clause ``degrees of confidence''. This has the effect of emphasising certain rules/clauses over others supporting a tuning of the knowledge base to reflect local context dependent domain properties. This extension is significant in that it brings certain Neural Net qualities to the adaption process potentially supporting a more rigorous understanding of adaption and learning in symbolic Knowledge Based systems. Consideration is also made of the use and maintenance of strategic knowledge as a way of improving system performance, both in encouraging increasing intelligent behaviour but also in the improvement of system computational efficiency.
Actions (login required)