Confirmed Special Sessions

Topic: New Frontiers in the Unsupervised Evolution of FPGA-based Circuits

Chairs: Marco D. Santambrogio (MIT) and Fabio Cancare (Politecnico di Milano)

Evolvable Hardware is a highly multidisciplinary eld concerning the creation or the adaption of physical circuits through evolutionary strategies aiming at improving the circuits behavior with respect to a given specication. The key elements in Evolvable Hardware are the Evolutionary Algorithms used (e.g. canonic genetic algorithm) and the programmable devices over which the circuits are deployed (e.g. FPGAs). In the last years, researchers were able to develop Evolvable Hardware techniques based on either FPGAs, FPTAs, FPAAs or custom devices; even if important results were obtained using all these devices, FPGA-based Evolvable Hardware techniques are the most explored and promising ones.

FPGA-based Evolvable Hardware systems have been successfully used in the past to create simple controllers, lters and to deal with fault recovery and system adaption. Most of the experiments reported, however, are based on old FPGA devices (e.g. Xilinx Virtex-2 Pro) and are frequently limited by the amount of available resources and on the capabilities of the targeted devices. Novel works should exploit last-generation FPGAs in order to scale to real-world applications.

The topics of interest include, but are not limited to:

- FPGA-based evolvable hardware systems and architectures
- Gate-level evolvable hardware techniques
- Function-level evolvable hardware techniques
- Multi-FPGA systems for evolvable hardware
- Real-world Evolvable Hardware applications

Topic: Dynamically Reconfigurable Embedded Systems

Chair: Giovanni Beltrame, Ecole Polytechnique de Montreal, Canada

Rationale of the need of the proposed special session: Due to the increasing complexity of modern embedded systems, companies and research institutes alike are trying to obtain more general solutions to be applicable to larger ranges of products. These new requirements in flexibility and adaptivity, particularly strong in small and medium volume markets like aerospace and biomedical hardware, create new constraints for the software and hardware designers. A popular solution to these problems consists in the exploitation of reconfigurable logic: this introduces more flexibility in the system without excessively sacrificing performance (as characteristic of pure software solutions). The fact that these systems can adapt to different environmental conditions and user demands opens the relatively new field of self-optimization, in which devices constantly change their configuration to attain certain performance goals. This session aims at collecting some of the most recent works in this new field, concerning both software and hardware approaches.