Embodiment tiers
The Unitree R1, two NAO units, Furhat and Temi provide multiple robot form factors, plus a flat-screen chatbot baseline.
A rare combination
The current stack reaches from controlled lab experiments to mobile embodiment and multimodal physiology—without needing a new major platform.
Physiology layers
16-channel EEG for central activity; ring-based PPG, EDA, temperature and movement for autonomic response.
Clean experimental principle
Use the same dialogue model and content across bodies so embodiment—not intelligence quality—is the treatment.
Best first program
Build a linked three-paper stream: embodiment → trust violation and repair → physiological mechanism. One validated logging pipeline can support all three.
Current research stack
Unitree R1 EDU ↗Mobile humanoid · ROS 2 · voice pipelineLab configuration: no dexterous hands / Orin
FurhatOn-hand social-robot platformCapabilities / SDK require verification
2× NAO V6On-hand robot unitsSDK setup in progress
TemiOn-hand robot unitModel / controls require verification
BIOPAC Ring AT01 ↗PPG · EDA · temperature · acceleration
OpenBCI + 2 caps ↗Cyton + Daisy · 16-channel EEG
Compute1 powerful desktop · 2 laptops
Dedicated Wi‑FiRobot control · event logging · LSL
Eight publishable study families
Filter by the primary contribution, then open a study for design, measures, equipment and published precedents.
Concrete design
A 3 × 2 experiment: NAO vs Furhat vs text chatbot, crossed with mental-state language (“I decided”) vs mechanistic language (“my program executed”). Keep dialogue content identical.
Measures
- Agency and experience; IDAQ; Godspeed anthropomorphism/likability
- Intentional-language coding in free-text explanations
- Optional ring EDA as an arousal mediator
NAO V6FurhatLaptopBIOPAC ring
Best fit: HRI · ICSR · RO-MAN · Frontiers in Robotics & AI
Concrete design
Repeated advice task: staged 90% vs 60% reliability, followed by 2–3 consequential errors. Randomize apology, denial, explanation, promise, or justification as the repair.
Measures
- Ability, benevolence and integrity; MDMT
- Behavioral reliance, delegation, and later advice-following
- EDA/HR reactivity at each violation
NAO or FurhatBIOPAC ringLaptop
Best fit: HRI · CHI · ACM THRI · CHB · MISQ/JMIS · ICIS
Concrete design
Room search or inventory task with three authority conditions: human allocates, robot proposes, or robot allocates autonomously. Focus R1 on navigation, escort and guidance—not manipulation.
Measures
- Completion time, errors, NASA-TLX and satisfaction
- Voluntary delegation rate and trust
- Optional ring-based workload/arousal
Unitree R1 EDUNAO alternativeLaptopWi‑Fi / ROS 2
Best fit: HRI · CHI · ICRA/IROS · MISQ · CSCW
Concrete design
Compare empathetic vs neutral dialogue while recording 16-channel EEG and the ring, synchronized with robot event markers through LSL. A robot breathing-rate manipulation can test physiological contagion.
Measures
- Frontal alpha asymmetry and EEG engagement indices
- SCR frequency/amplitude, HR and RMSSD
- Cross-correlation with speech and gesture onsets
OpenBCI 16-chBIOPAC ringFurhat or NAODesktop / LSL
Cyton + Daisy specification ↗: 125 Hz in 16-channel mode, appropriate for band power and “ERP-lite” analyses, not high-frequency work.
Best fit: NeuroIS · JAIS/MISQ · Emotion · HRI · Frontiers
Concrete design
A remote expert guides a local participant through assembly or negotiation. Compare Temi-mediated contact with laptop video and face-to-face interaction; confirm which navigation and event logs the lab unit exposes before preregistration.
Measures
- Networked Minds social presence
- Approach distance from video; add navigation logs only if the lab unit exposes them
- Turn-taking, task performance and inclusion
TemiLaptopsDesktop logging
Best fit: HRI · CHI · CSCW · Frontiers
Concrete design
Deliver identical career-advice or self-disclosure dialogue through Furhat, NAO and a text chatbot. Lock model, prompt and content across conditions; verify each robot’s installed SDK path before protocol lock.
Measures
- Disclosure depth and word count
- Trust, social presence, engagement and advice acceptance
- Ring arousal; gaze allocation after eye-tracking is added
FurhatNAO V6Laptop chatbotBIOPAC ring
Best fit: HRI · CHI · CUI · ICSR
Concrete design
A timed decision task with AI support: cross black-box vs explainable recommendations with low vs high error rate. Pair the experiment with a separate survey study for a two-study paper.
Measures
- Techno-overload, complexity and uncertainty
- EDA/HR arousal and frontal EEG workload markers
- Performance and discontinuance intention
Furhat / NAO / laptopBIOPAC ringOpenBCI
Best fit: MISQ · ISR · JMIS · JAIS · ICIS · CHB
Three paper directions
- Ring validation: compare HR and RMSSD against a Polar H10 or MP160; validate SCL and SCR rate only against an EDA reference such as an MP160 or Shimmer3 GSR+.
- Sync methods: quantify latency/jitter across OpenBCI, ring and robot markers; release an LSL harness.
- Scale methods: compare mind-perception instruments across embodiments, or test R1 as a scripted “robot researcher.”
BIOPAC ringOpenBCIRobot event logsReference sensor needed
Best fit: Behavior Research Methods · Sensors · NeuroIS · Frontiers methods · CHI/HRI alt proceedings
Equipment roadmap
Ranked by research unlock per dollar for this lab—not by technical prestige.
Best measurement upgrade
Pupil Labs Neon closes the largest gap: no current instrument can measure gaze during free robot interaction.
Best affordable condition
A Meta Quest headset adds the missing virtual embodiment and telepresence comparison at the lowest major-hardware cost.
Meta Quest 3S 128GB
VR avatar, telepresence and stress-induction conditions.
Adds an entire experimental condition family; the same 19 April 2026 source listed Quest 3 512GB at $599.99. Re-check both before purchase.
Pupil Labs Neon
Wearable, calibration-free gaze for mobile HRI.
Unlocks face-vs-task attention, joint attention and mutual gaze across A1, A2 and A6.
1–2× Polar H10
Cardiac reference and field HR/HRV.
~$90 each estimate
Low-cost reference needed for a credible ring-validation paper.
Emotiv Epoc X
14-channel quick-don saline EEG with LSL support.
Adds a mobile/field EEG lane when gel caps are impractical; raw-data software listed at $9.99/month.
OpenBCI EmotiBit bundle
Ambulatory EDA, PPG, temperature and IMU.
Open, LSL-compatible field sensing and a backup physiology stream.
2× 4K webcams
Frontal and room-view behavioral recording.
~$150–200 each estimate
Enables affect/behavior corpora with free OpenFace or Py-Feat; defer FaceReader.
Second BIOPAC Ring AT01
Dyadic autonomic synchrony.
Quote-based unverified
High value for multi-person HRI, but confirm price and data-export workflow first.
Second Cyton + Daisy
True 16+16-channel EEG hyperscanning.
Powerful but narrower; buy only after dyadic demand is proven with rings.
Mobile EEG
Current limit: 16-channel gel setup, 125 Hz in full mode, vulnerable to cap shift and cable sway. Cheapest partial fix: OpenBCI dry comb electrodes, $49.99 per 30-pack ↗.
Behavioral coding
Current limit: robot cameras are low-angle and needed for autonomy. Start with 2–3 fixed cameras and free OpenFace/Py-Feat + ELAN.
Touch & haptics
Current limit: R1 has no dexterous hands and NAO hands are mitten-like. A bHaptics TactSuit X40 is estimated at $330–500; pressure/force sensors are estimated at $50–200.
Longitudinal field sensing
Current limit: the ring uses an iOS-only app; whether raw export requires a cloud round-trip remains unverified. EmotiBit is the most stack-compatible open option; Empatica research pricing remains quote-based.
Build the measurement backbone once
A rigorous shared protocol turns eight disconnected ideas into a cumulative research program.
Validate the ring
Concurrent rest, Stroop/TSST-lite and robot interaction with a borrowed reference device.
- Bland–Altman agreement
- ICC for HR and RMSSD
- SCL and SCR-rate agreement
Measure synchronization
Characterize end-to-end delay and jitter across OpenBCI, ring and robot event markers.
- LSL event spine
- Offset distributions
- Open synchronization harness
Test the instruments
Compare Godspeed with newer mind-perception measures across NAO, Furhat and text.
- Measurement invariance
- Embodiment sensitivity
- Behavioral convergence
Recommended sequence
- Verify local raw export from the ring and the Furhat SDK/LLM route.
- Build one timestamped logging format for task events, robot actions and physiology.
- Pilot the ring against a cardiac/EDA reference before making mechanism claims.
- Run the embodiment study first; reuse its platform for trust repair and technostress.
- Add gaze only after the dialogue and event pipeline is stable.
Open questions before protocol lock
- BIOPAC does not publish the Research Ring AT01 or ACK100W-RNG software price.
- Confirm whether Senstream exports raw ring data locally without a cloud round-trip; this determines synchronization design.
- Verify the Furhat unit’s installed SDK version and LLM integration path.
- Noldus FaceReader, iMotions, Shimmer3 GSR+ and Empatica research pricing are quote-only.
- All prices were gathered through indexed web pages on 23 September 2026; re-check before purchase. “Estimate” labels are not vendor-confirmed.