hArtes aims at laying the foundation for a new holistic (end-to-end) approach for complex real-time embedded system design, with the latest algorithm exploration tools and reconfigurable hardware technologies. The proposed approach addresses, for the first time, optimal and rapid design of embedded systems from high-level descriptions, targeting a combination of embedded processors, digital signal processing and reconfigurable hardware. We will develop modular and scalable hardware platforms that can be reused and re-targeted by the tool chain to produce optimized real-time embedded products.
The results will be evaluated using advanced audio and video systems that support next-generation communication and entertainment facilities, such as immersive audio and mobile video processing.
Innovations of our approach include:

• support for both diagrammatic and textual formats in algorithm description and exploration;
• a framework that allows novel algorithms for design space exploration, which aims at automating design partitioning, task transformation, choice of data representation, and metric evaluation for both hardware and software components;
• a system synthesis tool producing near-optimal implementations that best exploits the capability of each type of processing element.

An important step in the hArtes design process developed at the DEI is the identification of a good partitioning and mapping of the given application. The partitioning phase addresses the identification of parallel code fragments on which transformation and metric analysis will be performed. This means that the input system description is decomposed into a set of tasks, trying to expose as much parallelism as possible. The allocation of the modules onto the different components of the hardware platform is performed as a step-by-step process, exploiting advanced reinforcement learning techniques. The project started in September 1st, 2006 and has a duration of three years.