Embedded decision support platform based on multi-agent systems

There has been an outstanding use of memory storage of processors as current applications: Artificial Intelligence-based applications, 3D-reconstruction or Blockchain ones, take advantage of their large computing effort, as well as their ability to support greedy treatments, and it grows the concern about providing reasonable resources for an efficient performance. This increase in demand requires optimized hardware configurations. In this context, the multi agent approach is a suitable solution to control the required resources for a multi-technology application. This paper investigates the scenario in which an embedded architecture implemented on FPGA is using a dynamic reconfigurable system that involves a multi-agent-based control part allowing optimizing and scheduling resources for technology processing requests. Hence, according to the required technology, the minimum of resources is selected. In a first time, performance is evaluated in terms of number of slices’ resources and task execution time for both a fixed and a dynamic reconfigurable architecture. A Dynamic Partial Reconfiguration is used to minimize efficiently the number of agents and consequently to minimize the allocated resources. The proposed dynamic reconfigurable architecture allows to save efficiently resources and execution time constraints.