Dynamic Application Model for Scheduling with Uncertainty on Reconfigurable Architectures

Applications executed on embedded systems require dynamicity and flexibility according to user and environment needs. Dynamically reconfigurable architecture could satisfy these requirements but needs efficient mechanisms to be managed efficiently. In this paper, we propose a dedicated application modeling technique that helps to establish a predictive scheduling approach to manage a dynamically reconfigurable architecture named OLLAF. OLLAF is designed to support an operating system that deals with complex embedded applications. This model will be used for a predictive scheduling based on an early estimation of our application dynamicity. A vision system of a mobile robot application has been used to validate the presented model and scheduling approach. We have demonstrated that with our modeling we can realize an efficient predictive scheduling on a robot vision application with a mean error of 6.5%.