Adapt and Survive
Using evolution to design real things -- nuclear reactors and financial trading systems -- may not be a surprise to every Dr. Dobb's reader. Genetic algorithms in their modern form have been around since the 1970s. Their sweet spot has always been design in a large search space with a well-defined and easily measured fitness function.
By contrast, the idea of incorporating evolution and adaptability into the operational fabric of engineering systems is a whole new departure. Instead of leading to the construction of a high-precision machine that can handle every possibility, such designs adapt to possibilities the designers cannot anticipate -- what Jet Propulsion Laboratory's Glenn Reeves calls "unknown unknowns." If devices can combine evolutionary software with flexible engineering substrates (like programmable logic arrays) to circumvent errors, they become self-healing and adaptive organisms. If design principles combine evolutionary adapability with a hierarchical feedback-control framework to ensure safety to life, property, and the environment, the resulting systems can be safer, cheaper, and far longer-lived.
For More Information
- What Really Happened on Mars?", Glenn Reeves, The Risks Digest Volume 19: Issue 49.
- A Conversation with Glenn Reeves, Jack Woehr, Dr. Dobb's Journal
- Reconfigurable VLSI Architectures for Evolvable Hardware: From Experimental Field Programmable Transistor Arrays to Evolution-Oriented Chips, Adrian Stoica, Ricardo Zebulum, Didier Keymeulen, Raoul Tawel, Taher Daud, and Anil Thakoor. IEEE Transactions on Very Large Scale Integration Systems 9, no. 1 (2001): 227.32.
- A New Accident Model for Engineering Safer Systems, Nancy Leveson. Safety Science, Vol. 42, No. 4, April 2004.