Agent-based reconfigurable systems

This dissertation extends the concepts of multi-agent technology that are so prevalent in today's distributed artificial intelligence software systems to reconfigurable systems—systems in which the functionality of both the associated hardware and software can be altered some time after the system has been constructed. Partitioning the hardware design space into entities called agents, which are autonomous units of execution that have the capability of interacting with the environment and each other, has been made much more attractive by recent advances in the capabilities of reconfigurable hardware. In a reconfigurable embedded processing environment, one possible benefit of the multi-agent approach is the use of a common design methodology for both the hardware and software components of the system. The utilization of this new paradigm has the potential to greatly increase the flexibility, efficiency, expandability, and maintainability of reconfigurable systems and to provide an attractive alternative to the current set of disjointed approaches that are currently applied to this problem domain. The focus of this dissertation is on the suitability of reconfigurable hardware to support hardware agents. Hardware agents can be used in conjunction with traditional software agents in hybrid architectures to support many of today's more complex applications. The dissertation introduces a general agent model that can be applied to hardware. In this dissertation the basic concepts of multi-agent technology is introduced and relate how these concepts can be incorporated in systems that contain reconfigurable logic elements such as field programmable gate arrays.; It is also shown how hardware agents can be created using a conventional hardware description language that when synthesized produces synchronous and asynchronous hardware constructs. Several examples that implement finite state and data flow machines using the hardware agent are introduced. The potential use of hardware agents in future image-processing and information gathering/monitoring systems is then examined. Then it is presented how hardware multi-agent technology can be applied to real-time sensor fusion applications.; The incorporation of hardware agents makes it possible to better utilize system resources and develop a unified system model that spans the hardware/software continuum.