Drone in Love: Emotional Perception of Facial Expressions on Flying Robots

Title of the Paper

Drone in Love: Emotional Perception of Facial Expressions on Flying Robots

Paper Information

• Research Area: Human-Computer Interaction, Affective Computing, Design and Evaluation of Flying Robots (Drones)
• Keywords: Human-Computer Interaction, Affective Computing, Emotion Recognition, Facial Expressions, Anthropomorphism, Robots, Drones

Research Background and Problem

• Research Problem: As drones increasingly enter human living spaces, their social acceptability and comprehensibility have become critical concerns. Existing studies suggest that social robots can enhance interaction quality with humans by exhibiting emotional expressions. However, compared to ground robots, the social design of flying drones is more challenging, and the effectiveness of their emotional expression designs remains to be validated.
• Research Significance: Facial expressions are a core channel of nonverbal communication, playing a crucial role in both robotics research and interpersonal communication. Designing and evaluating drones' facial expressions to convey emotions can improve their acceptability in scenarios such as collaborative tasks and companion interactions.
• Motivation and Related Work:
  • Drone design needs to consider social factors such as emotional expression, friendliness, and perceived intent.
  • Humans typically infer emotions through facial expressions, but the non-anthropomorphic nature of drones may impact the emotion recognition process.
  • Findings from ground robot studies, such as facial expression design and recognition patterns, cannot be directly applied to drones.

Proposed Solution

• Proposed Approach:
  • The authors designed a set of drone "emotional facial expression" renderings, showcasing five basic emotions: Joy, Sadness, Fear, Anger, and Surprise, each with three intensity levels (low, medium, high).
  • Two user studies (N=98, N=98) were conducted to evaluate the effectiveness and impact of emotion recognition through facial expressions under static and dynamic presentation conditions.
• Innovations:
  • Integration of the existing Facial Action Coding System (FACS) to optimize drone-specific facial designs, including eyes, eyebrows, pupils, and mouth.
  • Exploration of the relationship between drone emotional expressions and human emotional responses.
• Implementation Steps:
  1. Construct facial expression images corresponding to basic emotions step by step based on the FACS system.
  2. Conduct two online experiments using static images and dynamic videos, respectively.
  3. Collect participant data, including choice responses, confidence ratings, and explanatory descriptions, to analyze human recognition accuracy and interpretive patterns for different emotions.

Research Findings

• Specific Findings:
  • Participants could accurately recognize the five basic emotions (Joy, Sadness, Fear, Anger, Surprise) displayed by drones. However, "Disgust" was less effectively recognized.
  • Recognition rates varied across emotions: for instance, "Joy" had the highest recognition rate in static presentations, while "Sadness" performed best in dynamic presentations.
  • Humans not only understood the drones' emotions but also created narratives for their emotional states, incorporating environmental or human factors, and even exhibited sympathy and empathy toward the drones.
• Comparison with Existing Solutions and Advantages:
  • Compared to relying solely on drone flight paths to express emotions, facial expression design provided more precise emotion recognition capabilities.
  • Dynamic presentations better conveyed overall emotional states, eliciting stronger emotional responses and interaction intentions from participants.
• Experimental or Evaluation Results:
  • Overall recognition rates for the five emotions ranged from 43%-95% in static conditions and 43%-99% in dynamic conditions.
  • Emotional responses from participants included emotional impact, sympathy, and positive social behavioral intentions (e.g., comforting the flying robot).
• Limitations and Future Directions:
  • The experiments were conducted in an online simulated environment, without considering complex factors such as environmental noise and flight behavior during real-world drone interactions.
  • Future research should integrate drone facial expression design with flight paths to explore the effects of multimodal emotional expression.
  • Further studies are needed to investigate the design and effectiveness of emotional expressions in specific application scenarios (e.g., delivery, law enforcement, companionship).

Design and Methodological Recommendations

• Design Recommendations:
  • Use basic facial features (e.g., eyes, eyebrows, mouth) to construct emotional expressions on non-anthropomorphic drones.
  • Focus on designing the five basic emotions to enhance social acceptability and recognizability.
  • Consider the reciprocity of human emotional responses and design drone emotional expressions that can elicit positive social behaviors.
• Methodological Recommendations:
  • Combine static and dynamic presentation methods in research to comprehensively evaluate the effectiveness of drone emotional expressions.
  • Dynamic presentations are more suitable for holistic perception studies, while static images are better for refining specific facial features.

This study provides valuable insights for improving the social acceptability and application potential of drones in social environments, while also highlighting the theoretical significance of drone facial expression design for affective computing and human-computer interaction research.