# Heterogeneous Ground-Air Autonomous Vehicle Networking in Austere Environments: Practical Implementation of a Mesh Network in the DARPA Subterranean Challenge

**Authors:** Harel Biggie, Steve McGuire  
**arXiv:** [2203.12832](https://arxiv.org/abs/2203.12832)  
**Date:** 2022-03-24  
**Summary page:** https://www.summarizepaper.com/en/paper/2203.12832v1/

## Key points

- The paper presents an evaluation of a wireless mesh network system used in the DARPA Subterranean Challenge Final Event
- The system included air and ground robots, deployable mesh extender nodes, and a human operator base station
- The solution is highly customizable with form factor options suited for UAV operations and provides insight into network operations at all levels
- Performance metrics are discussed based on their performance in the Final Event, demonstrating practical successes and limitations of their approach
- Lessons learned and suggestions for future improvements are provided, including porting to arbitrary publish/subscribe architectures and scaling to larger team sizes
- A case study is presented on adding point-of-view (POV) imagery using front cameras from each robot to send highly compressed imagery back to the human supervisor for review
- Grainy images with visible compression artifacts were key to helping navigate through fog and identify artifacts that their autonomous artifact detection system missed
- Implementing a wireless mesh network in a real-life scenario requires significant systems engineering effort but can be successful with proper optimization and prioritization techniques.

## Summary

This paper presents an evaluation of a wireless mesh network system implemented in a real-life scenario during the DARPA Subterranean (SubT) Challenge Final Event. The system included air and ground robots, deployable mesh extender nodes, and a human operator base station. The authors present a real-world evaluation of their stack optimized for air and ground robotic exploration in an RF-limited environment under practical system design limitations. The highly customizable solution utilizes a minimum of non-free components with form factor options suited for UAV operations and provides insight into network operations at all levels. The authors discuss the performance metrics based on their performance in the Final Event of the DARPA Subterranean Challenge, demonstrating the practical successes and limitations of their approach, as well as a set of lessons learned and suggestions for future improvements. They analyze several key ideas that emerged during the development of their system, including porting to arbitrary publish/subscribe architectures and scaling to larger team sizes. In addition, they present a case study on adding point-of-view (POV) imagery using front cameras from each robot to send highly compressed imagery back to the human supervisor for review. This modification required a single line configuration change to instruct the UDP mesh layer to transmit the FPV video topic at a priority level below that of all other traffic (telemetry, artifacts, and map data). These grainy images with visible compression artifacts were key to helping navigate through fog and identify artifacts that their autonomous artifact detection system missed. Overall, this paper highlights that implementing a wireless mesh network in a real-life scenario requires significant systems engineering effort but can be successful with proper optimization and prioritization techniques.