Immo Colonius

Activity Recognition on Spatial Data in Logistic Processes

Optimizing logistic processes, like storing goods in warehouses, aims at reduction of maintenance costs and work time. To enhance established procedures, those have to be observed and analyzed.
Observations that have to be done in a logistic scenario at least include the amount and type of commodities used, their locations and movements done.
Once data about logistic processes is gathered, adequate clustering and reasoning has to be applied to extract knowledge about the procedures.
Having this in mind, techniques from the field of plan recognition, from the field of Spatial Cognition, or Symbolic Learning mechanisms can be applied if the gathered data is stored in an adequate abstract framework to extract characteristic activity patterns.
The scientific question therefore is twofold:
First, how to sort and cluster the collected data in a way that uncertainty is minimal, but the data itself is potent enough to allow activity recognition.
Second, how spacial-temporal recognition processes can be modeled, so that it represents a working abstraction from the measured data and can be used by symbolic reasoning.