VAM-HRI Workshops: VR/AR/MR for Human-Robot Interaction
Sources: VAM-HRI workshop reports, 2026-04-01 Raw: VAM-HRI 2024 Workshop; VAM-HRI 2025 Workshop Updated: 2026-08-12
Overview
VAM-HRI is the flagship workshop at the ACM/IEEE HRI conference, focusing on VR, AR, and mixed reality for human-robot interaction. The 2024 (Boulder, CO, 21 papers) and 2025 (Melbourne, 11 papers) editions together document the field’s direction: LLMs entering XR interfaces, LfD data quality as the central operational concern, and a shift from lab demonstrations toward industrial and clinical deployment. Meta Quest headsets have become the de facto research platform.
2025 Workshop (Melbourne, 11 papers)
Key Papers
ARCap (Paper #3): Comparative study of five LfD data collection strategies — two traditional, two AR-augmented traditional, one pure AR. AR-augmented methods consistently outperformed traditional baselines across four ML models. Pure AR nearly matched the best AR-augmented method with fewer ergonomic constraints. Quantitative evidence that AR augmentation improves LfD data quality. (See also: Simulation Data Collection)
SoftBiT (Paper #4, Waseda/CMU): Addresses the visual occlusion problem in soft robotic teleoperation — when soft fingers grasp an object, they’re often hidden from the operator’s view. Uses Meta Quest 2 to overlay real-time estimated soft finger deformations onto the operator’s view via a sim-to-real shape estimation pipeline. Provides proprioceptive feedback through XR.
LLM Shared Control (Paper #2): Integrates LLMs into a shared-control telepresence pipeline for navigation in dynamic, unstructured environments. The LLM’s scene understanding provides contextual adaptation beyond classical shared-control heuristics.
LLM Safety Annotation (Paper #7, UMBC / US Army Research Lab): Automates safety-relevant annotation of robot perception data in high-risk environments (EOD, industrial hazard zones) using LLMs, reducing costly human labeling effort.
Parkinson’s MR Rehabilitation (Paper #9): MR-assisted tremor rehabilitation integrating ergonomic hand support + interactive MR exercises. Design methodology informed by clinician and patient feedback. Finding: 77% of PD patients express frustration with traditional rehabilitation monotony; prior VR/AI therapy exposure correlated with higher motivation.
2025 Themes
- LLMs entering XR-HRI: Papers #2 and #7 both integrate language models as core components — the VAM-HRI community’s adoption of foundation models.
- LfD data quality as central concern: Papers #3, #4, #6 all address “how to collect better demonstration data with VR/AR,” directly responding to the robot learning data bottleneck.
- Shift toward industrial and clinical deployment: Papers #10, #11 are set in real factory environments, not lab demos.
- Meta Quest as de facto research hardware: Quest 2 and Quest Pro dominate.
2024 Workshop (Boulder, CO, 21 papers)
Key Papers
AR-guided IO setup (Paper #1): Mobile AR app for robot IO port wiring guidance. User testing: 25% lower mental demand, 15% higher usability vs. traditional methods — first quantified AR advantage in robot setup/maintenance.
VR Nuclear Sonification (Paper #2, U. West of England / Sellafield): Sonification of robot sensor data (radiation, temperature, flammable gas) in VR-simulated nuclear facility. Developed with actual Sellafield facility operators. Represents a rare research direction: sound as primary AR modality in safety-critical high-risk environments where visual channels are saturated.
VR Operator Identification (Paper #4): Head, eye, and hand tracking from VR teleoperation achieves near-100% operator identification accuracy (70/30 train/test split). Demonstrates VR biometric authentication potential.
Unity↔ROS Performance (Paper #11, UMass Lowell): Empirical benchmark of three ROS-for-Unity middleware implementations. TCP Connector (binary): ~0.6s image transfer; ROS# (WebSocket/JSON): ~10s. 16× difference from JSON serialization + Python rosbridge overhead. Critical reference for anyone building VR+ROS pipelines.
Legibility optimization (Paper #19, CU Boulder / CAIRO Lab): Algorithm jointly optimizes workspace object placement + AR virtual obstacle projections to improve human motion legibility (how predictably humans move). Better legibility → more accurate robot predictions of human intent → better task fluency.
Assistive robot MR training (Paper #20): Virtual Kinova Jaco arm with multiple AI-assisted control modes, mirrored to physical robot via ROS. MR training environment before operating the real device. Addresses safety barriers to learning assistive robotics.
2024 Themes
- Assistive robotics and accessibility (Papers #5, #7, #20, #21): biggest single theme of 2024, less prominent in 2025
- HRI methodology research (Papers #6, #10, #13, #14, #16): over 25% of papers focused on “how to use VR/AR for HRI research itself” — reflection on VR as a research tool
- Industrial AR practicality (Papers #1, #8, #15): directly deployable industrial and educational AR applications
- Tufts/UMass Lowell network: Cleaver, Sinapov (Tufts) and Allspaw, Yanco (UMass Lowell) appear across multiple papers — a visible cross-year research collaboration
Trends Across Both Years
- Meta Quest dominates as research hardware: Quest Pro (#1-2025), Quest 2 (#4-2025, 9-2024), HoloLens 2 (#5-2025, 15-2024) are the platform landscape
- LLM integration is accelerating: absent in 2024, two papers in 2025
- LfD data quality concerns are growing: from a minority theme in 2024 to a central concern in 2025
- ROS+Unity bridge is a consistent pain point: the performance disparity documented in 2024 Paper #11 affects most VR+robot research setups