Please use this identifier to cite or link to this item: http://dx.doi.org/10.18419/opus-11302
Authors: Raghavan, Priyanga
Title: Learning planning domains for office buildings using IoT data
Issue Date: 2020
metadata.ubs.publikation.typ: Abschlussarbeit (Master)
metadata.ubs.publikation.seiten: 50
URI: http://elib.uni-stuttgart.de/handle/11682/11319
http://nbn-resolving.de/urn:nbn:de:bsz:93-opus-ds-113190
http://dx.doi.org/10.18419/opus-11302
Abstract: Commercial buildings contribute to more energy consumption worldwide. Buildings are equipped with devices like sensors and actuators to know the current state of the building. The intelligence lies in making use of that collected data to make appropriate decisions that reduce energy consumption without compromising the comfort of occupants. For that we require an approach that can effectively plan and execute the operations. Energy consumption in commercial buildings by automatic control of HVAC and lighting systems using Artificial Intelligence (AI) planning ensures energy savings without any compromise in the comfort of its occupants. But AI planning requires an accurate definition of planning tasks. The generation of a planning domain model requires time, planning expertise and knowledge of the domain to be modelled. Alternately, planning domain can be generated automatically, which can save time and compensate for the domain expert’s incomplete knowledge. In this work, we use inductive rule learning technique to learn planning domains from Internet of Things (IoT) data. The planning domain for office buildings to control the operation of heating, ventilation and lighting system is defined. As the planning domain is learnt from the sensor data, an approach that can handle continuous numerical information is necessary. The problem of learning planning domain model is modelled as the classification problem and the rules are extracted from the IoT data. The appropriate post-processing techniques are used to convert the learned rules into the Planning Domain Definition Language (PDDL) code. The generated planning domain is tested using the handcrafted domain and the results show that the planning domain learner can learn the continuous numerical information and the relation between them along with propositional fluents.
Appears in Collections:05 Fakultät Informatik, Elektrotechnik und Informationstechnik

Files in This Item:
File Description SizeFormat 
MasterThesis_Priyanga_Raghavan.pdf456,26 kBAdobe PDFView/Open


Items in OPUS are protected by copyright, with all rights reserved, unless otherwise indicated.