Beyond Sensor Streams: HALO-Agent as a Body-State Representation Layer for Wearable Agents
WearAgent 2026 · Shanghai
A body-state representation layer that connects continuous wearable sensing with context-aware agent reasoning.
Program detailsWenbo Zhang · 张文博
PhD Candidate at SCUT · Visiting Researcher at UCL (Mar-Aug 2026)
Wearable sensing has become good at recognising movement, but health and interaction often depend on states we cannot observe directly - muscle activation, loading, and force. I develop multimodal systems that recover these states from practical sensors and turn them into interpretable feedback and context for human-centered wearable agents.

Upcoming
From body-state representation for wearable agents to real-time feedback for muscle-effort regulation.
WearAgent 2026 · Shanghai
A body-state representation layer that connects continuous wearable sensing with context-aware agent reasoning.
Program detailsSession D5: Sports Injury and Performance · Yangtze Hall
A real-time auditory biofeedback system for perceiving and regulating lower-limb muscle effort during exercise.
Program detailsResearch
I study how everyday sensors can recover consequential human states and support timely, reliable interaction.
How can sparse wearables capture movement, physiology, and physical interaction outside controlled laboratories?
How can multimodal models recover muscle activation, loading, pressure, and force that cannot be observed directly?
How can inferred body states become low-burden feedback and useful context for personalized wearable agents?
The central challenge is to learn from expensive lab-grade ground truth while keeping sensing sparse, calibration practical, and feedback useful in everyday settings.
Selected systems
Four systems that define my current research direction across sensing, state inference, and closed-loop interaction.

Making muscle effort audible for eyes-free, in-the-loop regulation during lower-limb exercise.

Organising muscle activation and biomechanical loading into interpretable body states.

Estimating lower-limb muscle activation and movement kinetics from sparse wearables.

Research outputs
* Equal contribution.
News
Presenting MuscleSense in the Sports Injury and Performance session at ACM UbiComp/ISWC 2026 in Shanghai. Event details
Presenting HALO-Agent at the WearAgent 2026 workshop, co-located with ACM UbiComp/ISWC 2026. Event details
HALO was accepted by ACM IMWUT 2026.
Completed a six-month visiting research stay at University College London with Prof. Jagmohan Chauhan.
Gave an invited seminar at UCLIC, University College London: "Beyond Motion: Body-State-Aware Wearable Intelligence for Human Movement and Interaction." Event details
MuscleSense was accepted by ACM IMWUT 2026.
WatchForce was published in Computer Methods and Programs in Biomedicine.
Presented MuscleSense at MobiUK 2026, University of Cambridge.
PPGSpeech was published in IEEE Internet of Things Journal.
Invited as a panelist at ACM UbiComp/ISWC 2025 to discuss impact, risks, security, and privacy in ubiquitous computing research.
Served as a Student Volunteer at ACM UbiComp/ISWC 2025.
KineticsSense and Motion2Press were accepted by ACM IMWUT.
Received the China Scholarship Council Joint Ph.D. Scholarship.
PressInPose was accepted by ACM IMWUT.
Contact
I welcome research discussions and collaborations on multimodal human sensing, wearable intelligence, mobile health, and human-centered AI.