
Haptic Robotic
Companion
Robot design & development
Role
Prototyper &
Designer
collaborators
Dániel György Szittya
Haris Arfan
João Pedro Sousa
How do you make a robot feel alive without a face, voice, or screen? This project explored whether a small companion robot could communicate emotional states: joy, anger, calm, distress, entirely through the way it moves its head when you touch it.
Research context

The project sits at the intersection of physical prototyping, affective computing, and interaction design research. It raises questions directly relevant to the next generation of wearables, AR/VR haptic controllers, and social robots: what is the minimum physical vocabulary needed to communicate emotional state, and how do you test for it rigorously?
Choosing the right motor


Microspora
5010 BLDC
The motor gives the head precise, backdrivable torque control, meaning the robot can push back against your hand while expressing an emotion. A fundamentally different quality from a simple servo. The entire system fits in a 140×190mm handheld form.
CAD & Technical details



CAD Sketch and rendering
We built the physical robot from scratch: 3D-printed in PLA, with a BLDC drone motor embedded in the neck joint, driven by a custom SimpleFOC firmware written in C++ on PlatformIO.



The design challenge was translating emotion into motion parameters: speed, amplitude, rhythm, direction change, and timing. We designed 8 haptic patterns and refined them down to four through two user studies with 12 participants total.
User Studies
Study 1: Mapping haptic patterns to emotions
We built 8 haptic patterns with varying speed, amplitude, and rhythm, and tested them with 6 participants. First, participants placed each pattern on a valence-arousal grid without any labels, just by feel. Then they experienced each pattern again and assigned it one of four emotions: Angry, Unhappy, Relaxed, or Delighted, and described what physical qualities led to their choice.


Key findings: Speed and rhythm were the strongest differentiators. Fast patterns clustered in high arousal, slow ones in low arousal, consistently across participants. Two patterns worked well: Relaxed (6/6 agreement) and a Delighted candidate. Angry was almost never recognised, with participants defaulting to Unhappy for all negative patterns. This told us we needed a fundamentally different movement vocabulary for Angry, and that the Unhappy/Relaxed boundary needed sharper physical contrast.
Study 2: Validating refined patterns in a full experience
We refined to 4 patterns and tested with 6 new participants across two phases. Phase 2 was a free interaction with the full robot in a Tamagochi inspired game environment, with sound and a visual interface active alongside haptics. Key findings: Relaxed hit 100%, Angry improved to 67%. Delighted failed at 33%, with the bounce mechanic consistently read as agitation rather than joy. Unhappy reached 50%, with remaining confusions mostly with Relaxed due to both occupying low-arousal space.

In Phase 2, haptics and sound scored equally (4.7/5) as contributors to emotional understanding, well above the visual interface (3.5/5). All 6 participants noticed the movement changing and attributed a personality to the robot without prompting.
This project shows that a robotic companion can communicate emotions through touch alone, with users accurately recognizing several emotional states and attributing a personality to the robot without prompting. The results suggest that responsive haptic movement can create a sense of aliveness and companionship, while future work should explore how different physical forms influence emotional interpretation.
