Emergent Co-Adaptive Strategies in Heterogeneous Multi-Robot Systems Via Meta-Learning
Hexiang Zheng, Xvchun He, Yuan Gao
AI summary
Problem
Existing heterogeneous multi-robot systems lack social adaptability, which causes misalignment with human movement patterns, disrupts activity rhythms, and undermines user trust in dynamic public environments.
Approach
The framework fuses large language models for real-time crowd urgency estimation with model-agnostic meta-learning to enable robots to rapidly adjust their behaviors and dialogue strategies without retraining.
Key results
- Dynamic shift between egoistic and altruistic strategies based on crowd urgency
- 21% improvement in crowd guidance success within three minutes
- 39% reduction in human physical and temporal cognitive load
- 16% increase in trust and 21% increase in perceived anthropomorphism
Why it matters
Demonstrates scalable human-robot co-adaptation, providing a practical foundation for socially intelligent robotic systems in complex public environments.
Abstract
As teamed robots increasingly share public spaces with humans, the ability to co-adapt—to mutually adjust behavior in response to one another—becomes essential for safe, efficient, and socially acceptable operation. This paper introduces a socially co-adaptive framework for heterogeneous multi-robot systems (HMRS) that enables real-time adaptation to human behavior while preserving cooperative task execution. Our approach fuses large language models for natural language understanding with model-agnostic meta-learning to allow robots to rapidly generalize across diverse social contexts. We implement and validate the system using a real-world HMRS composed of robots with different roles—workers, a station, and a social robot—interacting with 44 human participants under induced behavioral states (relaxed vs. nervous). Results reveal significant behavioral adaptation: the system dynamically shifts between egoistic and altruistic strategies, improving crowd guidance success by 21%. It also reduces human cognitive load—specifically, physical demands by 39% and temporal demands by 39%—while increasing trust by 16% and per- ceived anthropomorphism by 21%. This work demonstrates the feasibility of human-robot co-adaptation at scale, laying the groundwork for socially intelligent robotic systems capable of thriving in complex, human-centered environments.