vincent
Mon, 07 Jul 2008 12:41:24 +0000
Another paper at the Distributed Event Based Systems conference was on “Event Based Constraints for SensorNet Programming”. I had high hopes for an interesting paper on goal-driven CEP, but in fact the authors (from Brown University, USA) seemed to define “constraint” as a synonym of “filter” (as in “temporal, spatial, attribute-based or existential constraints on some event”). So there was no CLP-type relaxing of constraints during event processing to try and achieve some goal.
The paper did describe some an interesting approach to distributed CEP, using spatial and temporal event processing elements to divide up a high-datarate environment (detecting people being chased in a crowd). The main problem was, as always in distributed event processing (without a shared memory bus), ensuring that all possible event correlations can take place (with the right event being in the right processor at the right time). Probably for some sensor processing, though, a 100% success rate is not a requirement (i.e. if someone runs across an “event process boundary”, they will be presumably be picked up by the next event processor at some point).


Source...