TY  - THES
T1  - Resource-aware plan recognition in instrumented environments
A1  - Schneider,Michael
Y1  - 2010/05/10
N2  - This thesis addresses the problem of plan recognition in instrumented environments, which is to infer an agent';s plans by observing its behavior. In instrumented environments such observations are made by physical sensors. This introduces specific challenges, of which the following two are considered in this thesis: 
- Physical sensors often observe state information instead of actions. As classical plan recognition approaches usually can only deal with action observations, this requires a cumbersome and error-prone inference of executed actions from observed states.
- Due to limited physical resources of the environment it is often not possible to run all sensors at the same time, thus sensor selection techniques have to be applied. Current plan recognition approaches are not able to support the environment in selecting relevant subsets of sensors.
This thesis proposes a two-stage approach to solve the problems described above. Firstly, a DBN-based plan recognition approach is presented which allows for the explicit representation and consideration of state knowledge. Secondly, a POMDP-based utility model for observation sources is presented which can be used with generic utility-based sensor selection algorithms. Further contributions include the presentation of a software toolkit that realizes plan recognition and sensor selection in instrumented environments, and an empirical evaluation of the validity and performance of the proposed models. 
KW  - Planerkennung
KW  - Sensor
KW  - Benutzerverhalten
CY  - Saarbrücken
PB  - Universitäts- und Landesbibliothek
AD  - Postfach 151141, 66041 Saarbrücken
UR  - http://scidok.sulb.uni-saarland.de/volltexte/2010/2953
ER  -