What is Soft Computing?
Recent years have seen major advances in actuator and sensor technology, computing technology, and the emergence of a collection of new tools that can solve problems in an unconventional yet effective way. Constructed from identical data processing elements that are arranged in some regular pattern, Artificial Neural Networks (ANN) exhibit surprising abilities to capture non-linear relationship among variables, perform pattern classification, feature extraction, encode associate memory, among others. Fuzzy Logic (FL) can emulate human-like rule-based operations using linguistic terms such as "if it starts to become hot, turn the temperature down a little bit." Genetic Algorithms (GA) gives us a new way to perform optimization without actually solving equations in the traditional sense. These new tools are very good at what they do when applied to problems that can exploit their strengths. They are not "soft" in the sense that they are not rigorous, but because they do not lend themselves to easy analysis if one tries to cast them in the form of standard mathematics. To say it in a different way, it is not so easy to prove theorems about them. Most of the time, one resorts to intuition to develop a sense why they actually work, and yet it is equally intuitive to convince oneself that they should not work. For example, it is very hard to believe that the genetic algorithm would produce anything useful the first time one hears its explanation. It is much easier to appreciate it once one actually tries it out for oneself. Among the three mentioned methods, ANN may be considered to be the "hardest" and GA the "softest". Of course, other soft computing techniques such as DNA Computing and Simulated Annealing are also very intriguing. What makes these tools stand apart is the high dose of creativity in the way they were created. One just does not derive these tools in the conventional sense. They require rather wild inspiration.
Recent Highlight:
Please click on the following item for additional details:
Griffith, C., and Phan, M.Q., "Self-Learning of Target Interception and Obstacle Avoidance Rules for an Autonomous Vehicle," to appear.
Phan, M.Q., Juang, J.-N., and Hyland, D.C., "On Neural Networks in Identification and Control of Dynamic Systems," Wave Motion, Intelligent Structures, and Nonlinear Mechanics, Guran, A. and Inman, D. (eds.), World Scientific, Singapore, 1995, pp. 194-225.
For additional references on learning control and identification, please refer to List of Publications.