Take the Dog to the Park: Quadruped Robot for Joint Attention Training with Autistic Children in Naturalistic Settings

Paper Title

Take the Dog to the Park: Quadruped Robot for Joint Attention Training with Autistic Children in Naturalistic Settings

Publication Info

• Topic area: Robot-assisted interventions for autistic children in naturalistic settings
• Keywords: Autism Spectrum Disorder, joint attention, social robots, quadruped robot, naturalistic intervention, outdoor settings, child-robot interaction, embodied co-exploration, Wizard-of-Oz control, developmental therapy

Background and Problem

• Problem / challenge: Existing robot-assisted joint attention (JA) interventions for autistic children are predominantly stationary and confined to indoor settings, limiting skill generalization and ecological relevance.
• Significance: Joint attention is a foundational skill for social communication and language development, and its improvement is critical for broader developmental gains in autistic children.
• Motivation and related work: Prior research has shown that social robots can improve JA skills due to their predictable and controllable behaviors. However, most interventions have been limited to static environments, failing to leverage the dynamic, sensory-rich contexts of daily life. This paper addresses the gap by introducing a mobile, quadruped robot capable of operating in naturalistic settings.

Solution

• Proposed approach: A quadruped robot dog intervention designed for joint attention training in both indoor and outdoor environments, emphasizing embodied co-exploration.
• Novelty:
  1. Integration of a mobile quadruped robot into outdoor settings for JA training, moving beyond stationary paradigms.
  2. Development of a structured protocol combining ABA and NDBI principles, embedded in playful, naturalistic routines.
  3. Insights into the interaction dynamics and design implications for mobile robotic interventions in real-world contexts.
• Procedure and key techniques:
  • A four-week intervention with six autistic children, transitioning from indoor to outdoor settings.
  • Use of a Wizard-of-Oz (WoZ) control system for responsive robot operation.
  • Progressive environmental scaffolding to adapt children to increasing sensory and attentional demands.
  • Structured joint attention trials with adaptive prompting hierarchies and parent involvement.

Results

• Concrete findings:
  • Independent success rates in robot-led JA trials increased from below 20% to over 90% for some participants.
  • Success rates in human-led JA assessments rose from 32% at pretest to 91% at immediate posttest, with sustained gains (82%) at delayed posttest.
  • Parent-reported improvements in daily social communication, including eye contact, initiating communication, and responding to others.
• Advantage over baselines:
  • Demonstrated transfer of JA skills from robot-led sessions to human interactions.
  • Effective adaptation to outdoor environments, supporting ecological validity and skill generalization.
• Experiments / evaluation:
  • Participants: Six autistic children (ages 5-8) with moderate to severe social impairments.
  • Measures: In-session JA performance, Unstructured Joint Attention Assessment (UJAA), parent-reported social communication, and qualitative field notes.
  • Settings: Three indoor sessions followed by three outdoor sessions in semi-natural environments.
• Limitations and future work:
  • Small sample size (n=6) limits statistical generalizability.
  • Controlled outdoor settings may not fully represent real-world environments.
  • Reliance on Wizard-of-Oz control limits scalability; future work should focus on autonomous systems.
  • Need for extended longitudinal studies to assess long-term maintenance of gains.

Summary

This study introduced a quadruped robot dog intervention for joint attention training in autistic children, transitioning from indoor to outdoor settings. Over four weeks, six participants showed significant improvements in joint attention skills, with gains transferring to human interactions and daily social communication. The mobile, animal-like robot enabled embodied co-exploration, fostering engagement and ecological relevance. The findings highlight the potential of mobile robots for naturalistic interventions and provide design insights, including robust cueing, child self-regulation mechanisms, and caregiver integration. Future work should address scalability, autonomous operation, and long-term outcomes.