Source-linked AI summary
Feelium: A Touchable Blimp Body for Aerial Telepresence
George Xi Wang, Henghao Li, Shan Lin, Yunge Wen, Jiaqian Hu, Yuhua Jin
TL;DR
Feelium addresses the gap between touchable aerial platforms and remote telepresence by making a blimp’s envelope a shared touch surface. It combines first-person VR embodiment, onboard touch sensing, and social touch interactions; preliminary tests show contact-hover separation, with validation limited to dim-room lighting and one envelope patch.
Problem
Touchable drones have not passed contact back to their operators, and existing blimp envelopes and touchable inflatables do not connect a touched surface to a remote person.
Source: §1 · p. 2Method
Results
AUC 0.99 separated contact from hover on continuous clips, and 0.90 balanced accuracy held on the unseen remainder after threshold fitting on each clip’s first 60%.
Source: §4 · p. 3Takeaways & Limitations
Feelium offers a touchable body for aerial telepresence and early technical evidence of feasibility, motivating dyadic evaluation of social presence and whole-envelope sensing.
Source: §5 · p. 3Validation used dim-room lighting and one rectangular sensitive-skin patch of the envelope.
Source: §4 · p. 3
Abstract
from arXiv · showhide
Floating things invite touch. We present Feelium, a blimp-based telepresence platform that enables visual embodiment and touch interaction through its inflatable skin. Through a VR headset, a remote person inhabits the blimp, looking out of it first-person, appearing on its skin as a face or avatar, and steering it through the room. Partners in the room pat it, press a palm against it, draw on it, or lean into it; the skin senses each contact, renders it into the wearer's view in VR spaces. Touch thus provides a physical interaction channel for remote presence, turning the skin into a shared surface between remote and co-located partners.
1 Introduction
Prior touchable drones invite contact but do not return touch to their operators, while blimp envelopes have largely remained output-only. Feelium explores envelope touch in aerial telepresence, combining an inhabitable blimp with onboard touch sensing.
1 Introduction
- Earlier safe-to-touch drone wrapping raised touch from 1.4% to 38.9% of interactions, but the system could not pass touch back to its operator.Later haptic and touchable drones retained fingertip-scale, transient contact, external tracking, and loud, dangerous rotors.Source: §1 · p. 2
- Blimp envelopes have served as carriers or screens, while existing touchable inflatables do not connect their touched surface to a remote person.The cited blimp work includes tele-embodiment, projected faces, flying signage, and other forms; the ground-based examples include multi-touch displays, Emoballoon, and tangible balloons.Source: §1 · p. 2
- Feelium integrates a flying-display blimp, first-person VR inhabitation, and an onboard touch-sensing envelope, alongside a vocabulary of two-party touch interactions and preliminary technical validation.The authors describe it as, to their knowledge, the first exploration of envelope touch in aerial telepresence.Source: §1 · p. 2
2 The Feelium System
Feelium combines a projected, steerable indoor blimp with first-person VR embodiment and onboard sensing that detects, localizes, and renders touch. Its RGB camera decides touch; a thermal camera separately senses nearby warm bodies.
2 The Feelium System
- The 64 cm indoor blimp uses a rear-projection front face, four motors for quiet, stable hover, and weighs 323.7 g against approximately 360 g of buoyant lift.Most of its envelope is blackened to retain helium and reject ambient light; a 24.9 g laser projector lights the front face from inside.Source: §2 · p. 2
- A remote wearer sees through the blimp’s front-facing fisheye camera, appears on its envelope as a face or avatar, and steers it with six motion commands.The wide field keeps a reaching hand in view through contact; control, video, and speech travel over HTTP.Source: §2 · p. 2
- The RGB camera detects touch from back-light shadows, using instant-depth or sustained-anchoring gates, then localizes the shadow core and segments contact into tap, hold, or swipe events.Relative shadow depth is measured against an adapting reference image to compensate for uneven projector brightness; sustained anchoring uses the rolling minimum over the past second.Source: §2 · p. 2
- A thermal camera senses warm nearby bodies but makes no touch decisions; detected touches appear as located markers in the wearer’s view, and the projector responds at the touched spot.Thermal sensing compares frames with an adapting ambient baseline; conducted heat through the membrane is too weak and slow for touch detection.Source: §2 · p. 2
3 Interacting with an Inhabited Blimp
Feelium turns contact primitives composed over time and area into exchanges between a co-located partner and the remote wearer. The interactions range from taps and drawing to sustained, large-area contact.
3 Interacting with an Inhabited Blimp
- Everyday gestures become two-party exchanges: head taps ripple and return a warm pulse, palms meet as merged prints, cheek taps bloom a heart, and sustained lean-in contact creates a breathing glow.The wearer can also invite touch with a projected shimmer and respond through small body motions.Source: §3 · p. 2
- The sustained-anchoring score on labelled clips keeps contact almost entirely above the fitted threshold while hover repeatedly dips below it.The interactions are framed for presence-focused settings such as a goodnight palm-press, a drawing game, or a relative joining a family gathering.Source: §3 · p. 2, Figure 2 · p. 3
4 Preliminary Results
The preliminary validation found that sustained anchoring separated contact from hover well on continuous clips, while touch localization clustered around prompted targets. The pipeline ran onboard, but validation was limited to dim-room lighting and one envelope patch.
4 Preliminary Results
- On three continuous ~75 s clips, the sustained-anchoring score separated contact from hover at AUC 0.99; fitting a threshold on each clip’s first 60% yielded 0.90 balanced accuracy on the unseen remainder.Source: §4 · p. 3
- 95% recall but only 77% precision resulted from thresholding darkened area alone, as a hovering hand darkens the membrane nearly as much as touch.With a hand present, the detector registered activity in 171 of 198 trials and never fired in 30 empty scenes.Source: §4 · p. 3
- RMS spread was 4–21 px, with a 9 px median, for 69 detected tap and hold landing points around prompted targets; coordinates remained in camera space.Screen-to-camera calibration was future work, and the onboard pipeline ran at approximately 15 fps.Source: §4 · p. 3
- Validation covered 228 labelled trials over three sessions plus three continuous clips, but results were limited to dim-room lighting and one rectangular sensitive-skin patch.The labelled trials included 138 prompted touches and 90 negatives; the negative set included hovers, no-contact hand disturbances, and empty scenes.Source: §4 · p. 3
5 Conclusion and Future Work
Feelium integrates an onboard touch-sensing envelope with first-person VR embodiment for two-party social touch. Preliminary technical validation offers early evidence of feasibility and potential, motivating further evaluation and system extensions.
5 Conclusion and Future Work
- Preliminary technical validation provides early evidence of Feelium’s feasibility and potential.Source: §5 · p. 3
- A dyadic evaluation of social presence with the full touch loop remains future work.Source: §5 · p. 3
- Future extensions include whole-envelope sensing, haptic feedback to the wearer, and gesture-piloted flight.Source: §5 · p. 3
- The work aims to bring separated partners and families within arm’s reach through a quiet indoor blimp running fully onboard.Source: §5 · p. 3