Scholarlink

<p>Efficient collaboration within a robot swarm relies on precise relative localization among swarm members. However, in real-world scenarios such as GNSS-denied indoor environments, achieving accurate localization becomes challenging due to the absence of a global reference frame. This study evaluates the relative localization accuracy and operational feasibility of four systems: infrared-based (IR), visual-inertial odometry (VIO), ultra-wideband (UWB), and a decentralized spatial foundation model. We conduct low-fidelity simulations using simplified sensor models with large number of robots and high-fidelity simulations with realistic sensor observations to assess each system’s capability in supporting collective swarm behaviors. The systems are further validated through real-world experiments involving two swarms: one composed of five aerial robots and the other of five ground robots, performing three distinct collaborative behaviors. The results provide a comparative analysis of the systems, highlighting their estimation accuracy, communication overheads, and their impact on behavior performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Swarm robotics localization: comparing methods from infrared to foundation models

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Ali Imran,
  • Vivek Shankar Varadharajan,
  • Rafael Gomes Braga,
  • Giovanni Beltrame,
  • David St-Onge

摘要

Efficient collaboration within a robot swarm relies on precise relative localization among swarm members. However, in real-world scenarios such as GNSS-denied indoor environments, achieving accurate localization becomes challenging due to the absence of a global reference frame. This study evaluates the relative localization accuracy and operational feasibility of four systems: infrared-based (IR), visual-inertial odometry (VIO), ultra-wideband (UWB), and a decentralized spatial foundation model. We conduct low-fidelity simulations using simplified sensor models with large number of robots and high-fidelity simulations with realistic sensor observations to assess each system’s capability in supporting collective swarm behaviors. The systems are further validated through real-world experiments involving two swarms: one composed of five aerial robots and the other of five ground robots, performing three distinct collaborative behaviors. The results provide a comparative analysis of the systems, highlighting their estimation accuracy, communication overheads, and their impact on behavior performance.