Under Review
OUTLINES
Abstract
Real world experiment setting
Video Demo 1 (Zero-shot cooperative pursuit with unseen homogeneous partners)
Video Demo 2 (Zero-shot cooperative pursuit with unseen random heterogeneous partners)
Video Demo 3 (Zero-shot cooperative pursuit with unseen random heterogeneous partners)
Zero-shot coordination (ZSC) is a significant challenge in multi-agent collaboration, aiming to develop agents that can coordinate with unseen partners they have not encountered before. Recent cutting-edge ZSC methods have primarily focused on two-player video games such as Overcooked and Hanabi. In this paper, we extend the scope of ZSC research to the multi-drone cooperative pursuit scenario, exploring how to construct a drone agent capable of coordinating with multiple unseen partners to capture multiple evaders. We propose a novel Hypergraphic Open-ended Learning Algorithm (HOLA-Drone) that continuously adapts the learning objective based on our hypergraphic-form game modeling, aiming to improve cooperative abilities with multiple unknown drone teammates. To empirically verify the effectiveness of HOLA-Drone, we build two different unseen drone teammate pools to evaluate their performance in coordination with various unseen partners. The experimental results demonstrate that HOLA-Drone outperforms the baseline methods in coordination with unseen drone teammates. Moreover, HOLA-Drone demonstrates strong generalization to unseen environments when collaborating with previously unseen teammates. Furthermore, real-world experiments validate the feasibility of HOLA-Drone in physical systems. Videos can be found on the project homepage~\url{https://sites.google.com/view/hola-drone}.
In our experimental setup, we utilize Crazyflie drones whose positions are meticulously tracked using a network of 12 motion capture cameras operating at a frequency of 120 Hz. These positional data points are collected by a motion capture server connected via USB and subsequently transmitted to our pursuit planner over an Ethernet connection. The pursuit planner processes this data and forwards the positional information to the drones through a wireless communication module. Control commands are issued to the drones at a reduced frequency of approximately 10 Hz to ensure stable operation. Our multi-quadrotor communication framework is built upon the robust CrazySwarm system https://github.com/USC-ACTLab/crazyswarm. To ensure precise tracking of velocity commands generated by the pursuit planner, each drone is equipped with an onboard inner-loop PID controller.Â