×

Knowledge granularity spectrum, action pyramid, and the scaling problem. (English) Zbl 1007.91049

Liu, Jiming (ed.) et al., Agent engineering. Singapore: World Scientific. Ser. Mach. Percept. Artif. Intell. 43, 29-58 (2001).
A robotical agent must be designed to perform a task in some environment (e.g.finding an object in a room). The knowledge about the environment may be specified more or less finely (in gergon, with finer/coarser knowledge granularity). The finer the granularity, the better the action selected, but the longer the time needed to select it. The paper explores this tade-off.
The authors conclude that it is not possible to determine an optimal knowledge granularity in general. In the examined cases, a good performance is attained with a hierarchical, multi-layer granularity structure whereby a first set of actions is selected by a coarse layer, and then within that set a finer layer selects the final action.
For the entire collection see [Zbl 0966.00020].

MSC:

91C15 One- and multidimensional scaling in the social and behavioral sciences
91E40 Memory and learning in psychology
PDFBibTeX XMLCite