Adaptive Constraint-Based Agents in Artificial Environments

[PARTIAL KNOWLEDGE]   [Unordered Domains]   [Ordered Domains]   [Probabilities]

[ Please note: The project has been discontinued as of May 31, 2005 and is superseded by the projects of the ii Labs. There won't be further updates to these pages. ]

Ordered Domains

(Related publications: [PUBLink] [PUBLink])

In contrast to the value sets of the previous section, the elements of a state resource's domain can also be in a specific ordering relation. For example, the agent wants to fill a bucket with water, but he does not know how much water is in the bucket to start with. The amount of water in the bucket can be modeled by an integer state resource.

Of course, it is possible to apply the Unknown-Known-Not representation from the previous section to each domain value. But this would be a rather costly approach. It is more efficient to subsume consecutive values of the same knowledge level by intervals. The figure below shows an example.

It is even more efficient to consider only the convex (vertical) hull of intervals (trapezoids) of the same knowledge level. Actually, this method is not precise, because already excluded intermediate values may not be accessible any more. Consequently, the intervals (trapezoids) of different knowledge levels might overlap. For example, in the 65th second of the figure above, the convex hull of the Not knowledge level includes the Known values.

[PARTIAL KNOWLEDGE]   [Unordered Domains]   [Ordered Domains]   [Probabilities]

For questions, comments or suggestions, please contact us.

Last update:
May 20, 2001 by Alexander Nareyek