Investigation of Neural ODE LSTM for RSSI Indoor Localization

dc.contributor.authorChang, Charles
dc.contributor.supervisorDong, Xiaodai
dc.date.accessioned2022-09-01T01:52:48Z
dc.date.available2022-09-01T01:52:48Z
dc.date.copyright2022en_US
dc.date.issued2022-08-31
dc.degree.departmentDepartment of Electrical and Computer Engineeringen_US
dc.degree.levelMaster of Engineering M.Eng.en_US
dc.description.abstractTraditionally, indoor localization using received signal strength indicator (RSSI) measurements ignores the elapsed time between positions. This project investigates the use of elapsed time between positions as a feature with RSSI measurements for indoor localization. We use the recently proposed neural ordinary differential equations long short-term memory (ODE-LSTM) to handle the varying time gaps between each position. Our experiment compares the performance of the bidirectional LSTM (BiLSTM) model and the bidirectional ODE-LSTM (Bi-ODE-LSTM) model. The result shows that using time as a training feature for RSSI indoor localization can improve the accuracy. However, the benefit of the ODE-based model decreases as the number of RSSI features increases. Finally, the benefit brought by the Bi-ODELSTM model should be taken into account for the extra model complexity in practical applications.en_US
dc.description.scholarlevelGraduateen_US
dc.identifier.urihttp://hdl.handle.net/1828/14171
dc.language.isoenen_US
dc.rightsAvailable to the World Wide Weben_US
dc.subjectIndoor localizationen_US
dc.subjectNeural Ordinary Differential Equationsen_US
dc.subjectRecurrent Neural Networken_US
dc.titleInvestigation of Neural ODE LSTM for RSSI Indoor Localizationen_US
dc.typeprojecten_US

Files

Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Chang_Charles_MEng_2022.pdf
Size:
2.32 MB
Format:
Adobe Portable Document Format
Description:
Main article
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
2 KB
Format:
Item-specific license agreed upon to submission
Description: