EmoMime: Augmenting Social Behavior and Self-Expression via Wearable Robotic Limbs

TL;DR

EmoMime is a neck-mounted wearable robotic-limb device from the Inami–Kadomon Lab (UTokyo), accepted to the SIGGRAPH 2026 Emerging Technologies program. The robotic limbs mimic the wearer’s own gestures to amplify social signals (nodding, waving, expressive body language) and to enable “comforting self-touch” — the wearable can pat, hug, or steady the wearer as a form of self-directed affection. It is an HCI / human-augmentation artifact, not an ML paper: no learned policy is described in the public materials, and the artifact filed here is the demo video plus SIGGRAPH announcement rather than a preprint.

Key claims

EmoMime is a neck-mounted wearable robotic-limb system that both augments outward social signals and enables comforting self-touch. [SIGGRAPH 2026 program summary]
The system is a “new type of wearable robot” positioned as a proposal / research demo (提案) rather than a productized system. [tweet]
It is being shown at SIGGRAPH 2026 Emerging Technologies (E-Tech), the venue for novel interactive artifacts. [SIGGRAPH 2026 program]
Prior work from the same lab lineage (Sasaki, Saraiji, Fernando, Minamizawa, Inami — MetaLimbs, MetaArms) established feet-controlled and foot-mapped supernumerary arms; EmoMime extends this thread toward social/affective use cases. [author cross-reference; not a claim of the EmoMime work itself]

Method

Public materials do not disclose the control stack. From the video and press summaries the artifact is a shoulder/neck-worn mechatronic assembly with articulated limbs that mirror or amplify the wearer’s motion, with a mode for self-directed touch (the arms reaching back to pat or embrace the wearer). No details of sensing, actuation, or any learned components are given in the announcement or tweet; the SIGGRAPH E-Tech track typically documents such details in a 2-page abstract published closer to the conference.

Results

No quantitative results are reported in the announcement materials. Acceptance to SIGGRAPH 2026 E-Tech is the headline outcome. The demo video shows the limbs executing social gestures in sync with the wearer and performing self-touch actions; no user study or benchmark is cited in the public post.

Why it’s interesting

For Luma’s research feed this is a stylistic outlier — the wiki is heavily weighted toward LLMs, video/world models, and manipulation VLAs, and there is essentially no prior filing on wearable HCI hardware or supernumerary limbs. The closest thematic neighbor is Human-to-Robot Retargeting, which studies how to map human body motion onto robot embodiments; EmoMime inverts that direction (robot limbs mirror the human) but shares the underlying question of what a good human↔robot motion mapping looks like when the robot is worn. It is also worth watching as a source of social-behavior demonstrations that could feed the kind of egocentric human-behavior datasets increasingly used to bootstrap VLAs — the same lab’s MetaLimbs / MetaArms lineage produced widely-cited motion mappings in prior SIGGRAPH years.