2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops) : PerCom Workshops 2022
Page range:
316-321
Conference title:
CoMoRea '22 : 18th Workshop on Context and Activity Modeling and Recognition
Location of the conference venue:
Pisa, Italy
Date of the conference:
25.03.2022
Publisher:
Passarella, Andrea
;
Vallati, Carlo
;
Di Marco, Antinisca
;
Dürr, Frank
Plan and goal recognition are important tasks in constructing smart assistants and can be used to support activity recognition. However, all current approaches have the disadvantage of either being slow and inefficient regarding both online computation time and prior manual modelling effort tackled by human experts due to using domains which makes them infeasible for real-life settings; or requiring large amounts of annotated observation sequences which is not realistic as data annotation is expensive. This paper introduces a new approach requiring neither annotated observation data nor domains but only exemplary plans for each possible goal. It is based on plan-similarity based heuristics which makes it fast, yet is still able to achieve good results. This leads to new possibilities regarding the applicability in real-life settings and increases the usefulness for supporting activity recognition.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Search Authors in
