Computational Intelligence is a fairly new name covering a fairly new field. There is no consensus (yet) on what computational intelligence exactly is, but there is a widely accepted view on which areas belong to it: evolutionary computing, fuzzy computing and neurocomputing. The World Congress on Computational Intelligence held every four years (1994 Orlando, 1998 Anchorage, 2002 Honolulu) consists of three tracks, the IEEE International Conference on Evolutionary Computing, Fuzzy Computing, and Neurocomputing.
Enclosed in the name computational intelligence is a `message', according to scientific folklore it is chosen to indicate the link to and the difference with artificial intelligence. While some techniques within computational intelligence are often counted as artificial intelligence techniques (e.g. genetic algorithms, or neural networks) there is a clear difference between these techniques and traditional, logic based artificial intelligence techniques. In general, typical artificial intelligence techniques are top-to-bottom where, i.e., the structure of models, solutions, etc. is imposed from above. Computational intelligence techniques are generally bottom-up, where order and structure emerges from an unstructured beginning.
The areas covered by the term computational intelligence are also known under the name soft computing. Again, according to scientific folklore, this name was chosen to indicate the difference between soft computing and operations research, also known as hard computing. The two areas are connected by the problem domains they are applied in, but while operations research algorithms usually come with crisp (and strict) conditions on the scope of applicability and proven guarantees for a solution (or even an optimal solution), soft computing puts no conditions on the problem but also provides no guarantees for success, a deficiency which is compensated by the robustness of the methods.
The greater context in which we position our research is set by a number of prominent trends related to information processing devices in our lives (homes, offices, etc).
- The number and the computing power of such devices is growing.
- The communication capabilities and the degree of connectedness of such devices is growing.
- Users / owners want these devices work with minimal effort on setup and maintenance, that is, intelligently.
- Users / owners want these devices work together, that is, collectively.
As a consequence, we expect that the next wave of artificial intelligence will be collective intelligence, based on heterogeneous groups of many connected units. Furthermore, we envision two features becoming essential: adaptivity and autonomy.
We are especially interested in the combination of collectivity, adaptivity, and autonomy. Systems in the intersection of these areas include (future versions of) swarm robotic systems, smart grids, distributed sensor networks, eHealth systems with interactive sensing devices, ambient assisted living, and smart vehicles.
Within this context, we perceive adaptivity as the Grand Challenge in collective intelligent systems of the future. We foresee a pivotal role for adaptive capabilities because these systems must be equipped for scenarios where the operational circumstances are:
- not fully known in advance,
- so complex that behavioural rules cannot be designed & coded by traditional analytical means.
Our research is focused on algorithmic aspects. In particular, we work in evolutionary computing and machine learning, addressing fundamental issues as well as applications in optimization, data mining, artificial life, robotics, and art.