Adaptive and Evolvable Hardware - A Multifaceted Analysis

The paper explores adaptive and evolvable hardware (AEH) solutions from three essential perspectives: (a) technology, (b) economics and (c) system architecture. After setting the basis for the AEH terminology and taxonomy, the paper takes a look at the "market" and what adaptation is expected to provide to satisfy user needs and to solve real-world problems. Technologies that offer adaptation are explored, seeking common principles and techniques beneath the wide diversity of application areas. The focus here is on hardware that derives its flexibility from reconfiguration. The paper continues with an economic perspective and how adaptability offers better solutions not only for end users, but also for product providers. Finally, it explores system architecture ideas, to which biology offers continuous inspiration and may drive future architectural developments. Directions with maximal impact in advancing AEH appear to focus towards (1) reducing cost per function, (2) reducing reconfiguration overhead, (3) increasing re-programming speed, (4) improving algorithmic efficiency, (5) embedding self-reconfiguration algorithms into reconfigurable matter (or, more general, diffusing intelligence to finest HW levels), (6) ensuring that the system goes through safe states during adaptation and evolution, (7) and creating a specific development language and dedicated development tools.