A Soft Robotic Interface for Chick-Robot Affective Interactions

Jue Chen; Alexander Mielke; Kaspar Althoefer; Elisabetta Versace

cs.ROApr 9, 2026

A Soft Robotic Interface for Chick-Robot Affective Interactions

Authors: Jue Chen, Alexander Mielke, Kaspar Althoefer, Elisabetta Versace

Organizations: School of Biological and Behavioural Sciences, Queen Mary University of London, United Kingdom.

Abstract

The potential of Animal-Robot Interaction (ARI) in welfare applications depends on how much an animal perceives a robotic agent as socially relevant, non-threatening and potentially attractive (acceptance). Here, we present an animal-centered soft robotic affective interface for newly hatched chicks (Gallus gallus). The soft interface provides safe and controllable cues, including warmth, breathing-like rhythmic deformation, and face-like visual stimuli. We evaluated chick acceptance of the interface and chick-robot interactions by measuring spontaneous approach and touch responses during video tracking. Overall, chicks approached and spent increasing time on or near the interface, demonstrating acceptance of the device. Across different layouts, chicks showed strong preference for warm thermal stimulation, which increased over time. Face-like visual cues elicited a swift and stable preference, speeding up the initial approach to the tactile interface. Although the breathing cue did not elicit any preference, neither did it trigger avoidance, paving the way for further exploration. These findings translate affective interface concepts to ARI, demonstrating that appropriate soft, thermal and visual stimuli can sustain early chick-robot interactions. This work establishes a reliable evaluation protocol and a safe baseline for designing multimodal robotic devices for animal welfare and neuroscientific research.

Figures & tables

Fig. 1: Examples of chick–soft robot affective interface interactions in two experiments. a. Experiment 2 (horizontal, with faceplate and heating, see 2 a): chick looking toward the faceplate visual cue. b. Experiment 3 (vertical, with faceplate and heating, see 2 b): chick maintaining continuous contact with the heated interface.

Fig. 2: System design. a. Horizontal interface with a large faceplate. b. Vertical interface with a small faceplate. Both interfaces share the same structure: soft furry surface, heating pad, silicone pouch and a base (3D-printed in the horizontal interface; polypropylene in the vertical interface). c. The key steps in silicone pouch fabrication (embedded bladder, sealing, and two-layer silicone casting). d. Pouch surface deformation and internal pressure profile during one actuation cycle. e. Control architecture: an Arduino Uno controls two L298N motor drivers that switch between four air pumps (air in or out for two pouch groups).

Fig. 3: Experimental setups. a–b. Experiments 1a and 1b: two non-heated horizontal interfaces were placed at the midpoints of the short sides of the arena. In Experiment 1b, one big faceplate was attached to either the left or right interface, counterbalanced across sessions. c. Experiment 2: four horizontal interfaces were placed at the four corners, with two diagonally opposite interfaces heated and fixed. All interfaces were paired with big faceplates. d. Experiment 3: two vertical interfaces were placed at the midpoints of the short sides; the right interface was heated and fixed, and both interfaces were paired with small faceplates. Heated components are shown in red lines, and pink dashed regions mark the contact areas (where used to compute preferences). Water and food locations are indicated. All interfaces are honey-colored. During each session, just one side was in the ”breathing state” (as indicated by the wavy blue lines). The two sides alternated every five minutes across the session.

Fig. 4: Preferences in Experiment 1a (horizontal) and Experiment 1b (horizontal, with face). Figures display raw data for visualization, whereas all data statistics were based on the fitted Beta mixed-effects models. The x-axis represents the time of the experimental session, divided into six consecutive 5-minute time bins. The isolated “overall” data points on the right of each panel indicate the estimated marginal means averaged across all six bins for the entire session. Red dashed lines indicate the chance levels. Error bars represent the standard error of the mean (SEM). Asterisks denote statistical significance (* $p<0.05$ , ** $p<0.001$ ). a. Interface preference over time. Experiment 1a (solid line) and Experiment 1b (dashed line). b. Breathing preference over time. Experiment 1a (solid line) and Experiment 1b (dashed line). c. Face preference over time in Experiment 1b.

Fig. 5: Preferences in Experiment 2 (horizontal, with faceplate and heating) and Experiment 3 (vertical, with faceplate and heating). Figures display raw data for visualization, whereas all data statistics were based on the fitted Beta mixed-effects models. The x-axis represents the experimental session divided into six consecutive time bins (5 min each). The isolated “overall” data points on the right of each panel indicate the estimated marginal means averaged across all six bins for the entire session. Red dashed lines indicate theoretical chance levels. Error bars represent the standard error of the mean (SEM). Asterisks denote statistical significance (* $p<0.05$ , ** $p<0.001$ ). a. Interface preference over time in Experiment 2 (horizontal, with faceplate and heating). b. Heating preference over time in Experiment 2. c. Breathing preference over time in Experiment 2. d. Interface preference over time in Experiment 3. e. Heating preference over time in Experiment 3. f. Breathing preference over time in Experiment 3.

Exp.	$N$	Layout	Main Outcomes
1a	10	Horizontal	Behaviorally neutral and no avoidance.
1b	10	Horizontal (with faceplate)	Overall preference for face-like pattern. Overall preference for interface.
2	20	Horizontal (with faceplate and heating)	Overall preference for heating. Overall preference for interface.
3	14	Vertical (with faceplate and heating)	Overall preference for heating. Overall preference for interface.

TABLE I: Summary of Experimental Results (Transfered)

Social Robotics

Social robotics studies robots that interact with people through social communication, assistance, collaboration, and behavior adapted to human contexts.

See all 80 papers

Emotional Expression in Low-Degrees-of-Freedom Robots: Assessing Perception with Reachy Mini
May 12, 2026Amit Rogel, Elmira Yadollahi, Guy LabanHuman-Robot Interaction Emotion Recognition
Robotic Musicianship Lab, Georgia Institute of Technology, Atlanta, GA, USA · School of Computing and Communications, Lancaster University, Lancaster, UK · Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, Beer Sheva, Israel +2
Robotic Affection -- Opportunities of AI-based haptic interactions to improve social robotic touch through a multi-deep-learning approach
May 4, 2026Ali Askari, Jens GerkenHuman-Robot Interaction Robotic Tactile Sensing
TU Dortmund University, Germany
From Pets to Robots: MojiKit as a Data-Informed Toolkit for Affective HRI Design
Mar 12, 2026Liwen He, Pingting Chen, Ziheng Tang +4Human-Robot Interaction Social Robotics
The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, Guangdong, China · Tsinghua University, Beijing, China · Beijing Institute of Technology, Beijing, China +3

May 12, 2026J/K move · Enter open · S save

Emotional Expression in Low-Degrees-of-Freedom Robots: Assessing Perception with Reachy Mini

Amit Rogel, Elmira Yadollahi, Guy Laban

Robotic Musicianship Lab, Georgia Institute of Technology, Atlanta, GA, USA · School of Computing and Communications, Lancaster University, Lancaster, UK · Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, Beer Sheva, Israel+2

Human-Robot Interaction Emotion Recognition Affective Computing

Open paper AskPDF arXiv

Emotion expression is central to human--robot interaction, yet little is known about how people interpret affect on robots with sparse, non-anthropomorphic expressive capabilities. This study examined how people perceive emotional expressions displayed by Reachy Mini (Pollen Robotics and Hugging Face), a low-degree-of-freedom (low-DoF) robot with a constrained and distinctly non-human expressive repertoire. In an online within-subjects study, 100 participants viewed 10 short video clips of Reachy Mini expressing different emotions and, for each clip, identified the perceived emotion, rated its valence and arousal, and evaluated the robot on social-perception traits. Exact emotion recognition was modest overall and varied considerably across expressions, with anger, sadness, and interest recognized more reliably than emotions such as love, pleasure, shame, and disgust. However, participants were generally more successful at recovering broader affective meaning than exact emotion labels, particularly along valence and arousal dimensions. Emotional expressions also shaped social evaluation, as positive expressions were perceived as warmer and more sociable than negative ones, and animacy varied less across conditions. These findings suggest that even constrained robotic expressions can communicate affective meaning and influence social impressions, positioning Reachy Mini as a useful benchmark for studying affective communication in low-DoF robots.

Abstract

Social Robotics

Explore similar work

Emotional Expression in Low-Degrees-of-Freedom Robots: Assessing Perception with Reachy Mini

Robotic Affection -- Opportunities of AI-based haptic interactions to improve social robotic touch through a multi-deep-learning approach

From Pets to Robots: MojiKit as a Data-Informed Toolkit for Affective HRI Design