Planning for the Future

The KTP Project has been a huge success already, but what's happening next? Below you'll find an (incomplete) list of events which the project is hoping to incorporate. Be aware that the plan is always adapting to changing times, as any good research and development project should.

If you think you have something to contribute to the future of this project, please contact Chris with details.

Clustering algorithms have been used by data analysts in many fields. "Clustering is the classification of objects into different groups, or more precisely, the partitioning of a data set into subsets (clusters), so that the data in each subset (ideally) share some common trait - often proximity according to some defined distance measure." Ref: Wikipedia Data Clustering

For example, one can create a vector from a selection of flight parameters (such as altitude, vertical acceleration, airspeed etc) at regular intervals during the flight. In its most basic form, clustering could be used simply to identify similar phases of flight by grouping the above vectors into clusters. This is performed using a distance algorithm, which groups those nodes that are closest together.

There are two important factors when working with clustering. The number of clusters used and the starting position of the cluster nodes both play a big part in outcome of the cluster results. For example, creating more clusters would split the data into more groups - this might help you to identify more situations (such as hard landings, go-arounds etc), but it might also disrupt other clusters (e.g. splitting a sensible single cluster into two). Cluster node starting positions have a big influence on which items of data belong to which cluster, and therefore the sizes of each cluster.

Work is being conducting on expanding the initial clustering algorithms already created in order to analyse entire flights rather than just particular sections of interest. If this is successful, work can then be carried out on identifying these clusters using external sources of information and then implimenting statistical analysis on the cluster changes over time. There are so many routes to be explored here, all of which have some value, but given time constraints it is important that time is set aside for researching other AI techniques.

(top)

If anyone knows of any cases where fuzzy logic has been used already in similar situations, please contact me as I'd be interested in researching this further.

"Degrees of truth are often confused with probabilities. However, they are conceptually distinct; fuzzy truth represents membership in vaguely defined sets, not likelihood of some event or condition. For example, if a 100-ml glass contains 30 ml of water, then, for two fuzzy sets, Empty and Full, one might define the glass as being 0.7 empty and 0.3 full. Note that the concept of emptiness would be subjective and thus would depend on the observer or designer. Another designer might equally well design a set membership function where the glass would be considered full for all values down to 50 ml. A probabilistic setting would first define a scalar variable for the fullness of the glass, and second, conditional distributions describing the probability that someone would call the glass full given a specific fullness level. Note that the conditioning can be achieved by having a specific observer that randomly selects the label for the glass, a distribution over deterministic observers, or both. While fuzzy logic avoids talking about randomness in this context, this simplification at the same time obscures what is exactly meant by the statement the 'glass is 0.3 full'." Ref: Wikipedia Fuzzy Logic

(top)

Following the release of the Flight Operations paper, we are looking forward to producing a technical paper on the research of applying AI techniques to Flight Data. The next big conference on the agenda is the IEEE World Congress on Computational Intelligence, www.wcci2008.org.

(top)

• Home
• AI Clustering Algorithms
• Fuzzy Logic Algorithms
• Technical Paper