Prospect for Indoor Localization Kaifei Chen UC Berkeley

Why indoor localization is important Applications Indoor Parking Lots Localization Services Shopping Mall Other Services Building Operating System (BOSS) Conference Rooms 3/2/2021 Indoor Emergency Rescue Software-Defined Building 2

Existing Indoor Localization Systems Fingerprint-Based Localization 3/2/2021 Beacon-Based Localization Motion Sensor-Based Localization (Dead Reckoning) Software-Defined Building 3

Why Still Work On Indoor Localization • They have assumptions and limitations – Fingerprints : ubiquitous, unique, and consistent – Beacons : extra infrastructures, overhead – Dead reckoning : normal movement of phones • Disconnection between services and applications – Over-emphasis on physical accuracy 3/2/2021 Software-Defined Building 4

Why Still Work On Indoor Localization Shopping Mall Indoor Parking Lots Conference Rooms 3/2/2021 Indoor Emergency Rescue Software-Defined Building 5

Why Still Work On Indoor Localization • They have assumptions and limitations – Fingerprints : ubiquitous, unique, and consistent – Beacons : extra infrastructures – Dead reckoning : normal movement of phones • Disconnection between services and applications – Over-emphasis on physical accuracy – Applications mostly need semantic information 3/2/2021 Software-Defined Building 6

Why Still Work On Indoor Localization Shopping Mall Indoor Parking Lots Conference Rooms 3/2/2021 Indoor Emergency Rescue Software-Defined Building 7

Why Still Work On Indoor Localization • They have assumptions and limitations – Fingerprints : ubiquitous, unique, and consistent – Beacons : extra infrastructures – Dead reckoning : normal movement of phones • Disconnection between services and applications – Over-emphasis on physical accuracy – Applications mostly need semantic information • Too much administrator and user efforts 3/2/2021 Software-Defined Building 8

Our Solutions • Combine Indoor Localization Systems – Limitations are mutually independent – Context awareness and localization confidence – Take advantages of data fusion techniques • Semantic Localization • Automatic Data Collection and Map Generation 3/2/2021 Software-Defined Building 9

Indoor Localization System Framework 3/2/2021 Software-Defined Building 10

Prototype Evaluation • Implemented – Wi. Fi signal strength fingerprint localization – Acoustic Background Spectrum (ABS) fingerprint localization – Kalman Filter for combination • Executed experiments (3 paths) in RADLab in Soda Hall 3/2/2021 Software-Defined Building 11

Evaluation : Wi. Fi Confidence When Confidence is High, Error is Low = When Error is High, Confidence is Low 3/2/2021 Software-Defined Building 12

Evaluation : ABS Confidence When Confidence is High, Error is Low = When Error is High, Confidence is Low Exceptions exist 3/2/2021 Software-Defined Building 13

Evaluation : Kalman Filter Combination Framework leans to choose better estimations and avoids spikes Average performance is the same as the better one 3/2/2021 Software-Defined Building 14

Future Work • Semantic Localization Compared to physical localization • Automatic Data Collection and Map Generation – Crowdsourcing 3/2/2021 Software-Defined Building 15

Prospect • Plug and play indoor localization service in any building • Robust indoor localization services • Provide semantic location information • Generic APIs for application development – Personalized energy usage tracking – Energy saving: HVAC and light control 3/2/2021 Software-Defined Building 16

Thank you! Q&A Kaifei Chen http: //kaifei. me kaifei@cs. berkeley. edu 3/2/2021 Software-Defined Building 17