Overcoming Data Distribution Shift in AIoT Mobile Sensing

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "Overcoming Data Distribution Shift in AIoT Mobile Sensing"

By

Mr. Tianlang HE


Abstract:

Mobile devices (e.g., smartphones, wearables, and drones) are increasingly
equipped with various sensing modules, including IMUs, magnetometers, and
radio sensors. The field of mobile sensing studies practical systems that
utilize sensing data to infer the physical context of mobile devices (e.g.,
device locations and user activities), which is fundamental to IoT services
spanning mobile health, location-based services, and the low-altitude
economy. While the recent integration of AI technologies has advanced the data
processing of mobile sensing (a paradigm known as AIoT), naively applying AI in
mobile scenarios often leads to unexpected failures. This is primarily due to
distribution shifts of sensing data in between model training and real-world
deployment, owing to factors like user heterogeneity and environmental
variations. Through strategies of adaptation and generalization, this thesis
studies deployable approaches to overcome data distribution shifts in several
tasks within AIoT mobile sensing.

First, we establish adaptation capabilities for regression tasks by introducing
TASFAR, a generic algorithm for source-free unsupervised domain adaptation
that is both agnostic to distribution shifts and storage-efficient for mobile
deployment.

Second, we enhance generalization for two classification tasks by proposing
ELESON and PRID. ELESON achieves single-modal generalization for classifying
user’s conveyor states using phone INS under arbitrary user behaviors. PRID
achieves multi-modal generalization by aligning IMU and BLE data for detecting
proximity between mobile devices under diverse device carriage and multipath
environments.

Finally, we extend generalization ability to sequential decision-making by
proposing LoRaCompass, a robust reinforcement learning model that guides a LoRa
sensor to efficiently search for an IoT tag across diverse, unknown
environments.


Date:                   Thursday, 5 February 2026

Time:                   2:00pm - 4:00pm

Venue:                  Room 2132C
                        Lift 22

Chairman:               Prof. Chi Ying TSUI (ISD)

Committee Members:      Prof. Gary CHAN (Supervisor)
                        Prof. Raymond WONG (Co-supervisor)
                        Dr. Chaojian LI
                        Dr. Wei WANG
                        Prof. Wai Ho MOW (ECE)
                        Dr. Jack Yiu Bun LEE (CUHK)