Multi-Objective Intrinsic Evolution of Embedded Systems (MOVES)
The Project:
This project aims at the
investigation of intrinsically evolvable embedded systems. Simulated
evolution provides embedded systems with a means to react properly to
unforeseen changes in the environment and the system state. The vision
behind this project is that emerging reconfigurable hardware
architectures combined with
biologically-inspired optimization methods will enable new approaches
for
achieving self-adaptation, self-optimization, robustness and
fault-tolerance in embedded systems.
In an
intrinsically evolved system the evolutionary process runs together
with the function under evolution on the same target platform. This is
a necessary precondition for autonomous operation. Applying
evolutionary algorithms on the level of reconfigurable hardware allows
to
design systems that are able to reorganize their structure online. Such
a reorganization is needed in two cases: changes in the system's
environment and changes in the system state. Changes in the environment
are typically slow, allowing for adaptation by evolutionary techniques.
Moreover, the evolutionary design techniques are targeted at multiple
objectives, such as functional quality, hardware area, speed, and power
consumption. A
trade-off between these objectives has to be found that is good enough
for the current operation mode. Changes in the system state, ie. the
system's resources, can be rather fast and radical. In such a case, we
may switch between pre-evolved alternative solutions to keep the system
operational. Functional recovery
as a reaction to changes is inherent in the evolutionary design method.
Our Mission:
- Develop new resource-aware models and algorithms for intrinsic evolution of embedded systems
- Investigate multi-objective optimization techniques for reconfigurable hardware
- Examine domain-specific approaches to answer challenges such as scalability
- Develop the basic technology for self-reconfiguration of hardware and software functions
Tools:
Contact:
Dipl.-Inf. / Dipl. Math. Paul Kaufmann
Prof. Dr. Marco Platzner
The project is funded by the German Research Foundation (DFG)
within the
priority program 1183, ”Organic Computing”.