[论文解读] Towards Large-Scale Scalable MAV Swarms with ROS2 and UWB-based Situated Communication

Jorge Peña Queralta, Xianjia Yu+2 more|arXiv (Cornell University)|Mar 11, 2021

UAV Applications and Optimization被引用 4

一句话总结

该论文提出了一种可扩展、低成本的微型飞行器（MAV）蜂群平台，采用ROS2、Wi-Fi网状网络和基于UWB的定位通信技术，实现分布式定位与控制。通过将UWB测距与ROS2话题及batman-adv网状路由相结合，系统实现了去中心化、可扩展的蜂群操作，具备实时相对定位与通信能力，支持在无GNSS环境下的复杂协同行为。

ABSTRACT

The design and development of swarms of micro-aerial vehicles (MAVs) has recently gained significant traction. Collaborative aerial swarms have potential applications in areas as diverse as surveillance and monitoring, inventory management, search and rescue, or in the entertainment industry. Swarm intelligence has, by definition, a distributed nature. Yet performing experiments in truly distributed systems is not always possible, as much of the underlying ecosystem employed requires some sort of central control. Indeed, in experimental proofs of concept, most research relies on more traditional connectivity solutions and centralized approaches. External localization solutions, such as motion capture (MOCAP) systems, visual markers, or ultra-wideband (UWB) anchors are often used. Alternatively, intra-swarm solutions are often limited in terms of, e.g., range or field-of-view. Research and development has been supported by platforms such as the e-puck, the kilobot, or the crazyflie quadrotors. We believe there is a need for inexpensive platforms such as the Crazyflie with more advanced onboard processing capabilities and sensors, while offering scalability and robust communication and localization solutions. In the following, we present a platform for research and development in aerial swarms currently under development, where we leverage Wi-Fi mesh connectivity and the distributed ROS2 middleware together with UWB ranging and communication for situated communication. We present a platform for building towards large-scale swarms of autonomous MAVs leveraging the ROS2 middleware, Wi-Fi mesh connectivity, and UWB ranging and communication. The platform is based on the Ryze Tello Drone, a Raspberry Pi Zero W as a companion computer together with a camera module, and a Decawave DWM1001 UWB module for ranging and basic communication.

研究动机与目标

• 实现大规模、去中心化的MAV蜂群，最大限度减少中心化控制。
• 解决集中式控制和传统Wi-Fi在蜂群机器人中的局限性。
• 开发一种基于市售组件的低成本、可复现平台。
• 将基于UWB的相对定位与ROS2集成，实现可扩展的通信与协调。
• 支持复杂协同行为，如队形控制和分布式感知。

提出的方法

• 利用Ryze Tello无人机搭载Raspberry Pi Zero W和Decawave DWM1001 UWB模块，实现机载处理与UWB通信。
• 采用batman-adv实现Wi-Fi网状网络，支持自愈、可扩展的节点发现以及互联网网关功能。
• 使用一对一UWB测距，并结合可扩展系统如SnapLoc，估算蜂群内各节点的相对位置。
• 将ROS2消息嵌入UWB测距数据包中，实现在极低带宽开销下并发的定位与通信。
• 利用Wi-Fi网状网络实现高带宽数据传输，并通过话题节流机制管理网络负载。
• 集成TFMini微型激光雷达，提升高度估计精度与系统鲁棒性。

实验结果

研究问题

• RQ1 能否仅使用市售组件和去中心化通信，构建低成本、可扩展的MAV蜂群？
• RQ2 如何有效将基于UWB的相对定位与ROS2集成，以实现实时蜂群协调？
• RQ3 Wi-Fi网状网络在大型MAV蜂群中能在多大程度上支持可靠、可扩展的通信？
• RQ4 能否通过UWB消息嵌入实现并发定位与通信，同时不牺牲可扩展性？
• RQ5 在无GNSS环境中，采用分布式网状控制的系统性能如何？

主要发现

• 该平台仅使用市售组件和开源软件，即可实现可扩展、去中心化的MAV蜂群操作。
• 采用batman-adv的Wi-Fi网状网络支持在动态蜂群配置中实现自动节点发现和可靠通信。
• UWB测距可实现精确的相对定位，若已知网关或锚点位置，还可实现全局定位。
• 将ROS2消息嵌入UWB数据包中，可在极低开销下实现并发通信与定位。
• 该系统支持在无GNSS环境中实现队形控制和分布式感知等复杂协同行为。
• ROS2、UWB与Wi-Fi网状网络的集成，为大规模蜂群研究提供了可复现、低成本的基础平台。