Authors: Martin Palkovic Jeroen Declerck Prabhat Avasare Miguel Glassee Andy Dewilde Praveen Raghavan Antoine Dejonghe Liesbet Van der Perre
Publish Date: 2012/03/25
Volume: 69, Issue: 3, Pages: 317-327
Abstract
Novel cognitive radio platforms such as IMEC’s COgnitive Baseband RAdio COBRA should ensure the feasibility of multiple streams and their reconfigurability and scalability during runtime The control over these tasks should be dedicated to a runtime controller that reallocates the resources on the platform Eg when the channel conditions change requiring a switching to different modulation and coding scheme or a user starts a new stream Current transaction level models are too detailed for rapid exploration of all runtime options and the highlevel dataflow frameworks such as Kahn process networks lack the dynamism and reconfigurability that is essential for the exploration In this paper we propose the DAtaflow for RunTime DART the highlevel dynamic dataflow platform model framework suited for rapid runtime control development We sketch how to use this framework to develop such a controller in the reactive and more challenging proactive way We derive the component timing based on Instruction Set Simulator ISS simulation and the reconfiguration timing based on Transaction Level Modeling TLM simulation Finally we verify results of our DART approach with full TLM simulation of our platform
Keywords: