Core Concepts
Deploying robotic platforms to map hazardous glacier moulins is a promising yet challenging endeavor, requiring meticulous preparation, resilient equipment, and robust safety protocols.
Abstract
The paper presents a field deployment of a custom-made lidar-inertial mapping platform on the Athabasca Glacier in the Canadian Rockies. Glacier environments, particularly the ice chasms known as moulins, pose significant risks and logistical challenges for human operators. The authors leveraged an automated platform to collect data within these hazardous settings, aiming to unlock new applications and deepen the understanding of glacial environments.
The key challenges encountered during the deployment included:
- Unpredictable weather conditions, requiring rugged and resilient equipment
- Vulnerability of non-rugged components, such as the Raspberry Pi 4B computer on the platform
- Adherence to stringent safety protocols, necessitating thorough preparation and testing
- Environmental impact considerations, such as the use of a metal cage to prevent material loss
- Mapping difficulties due to the lack of distinct features in the moulin environment, leading to under-constrained and degraded solutions
The authors emphasize the paramount importance of meticulous preparation, comprehensive testing, and robust verification procedures to ensure the success of such deployments in extreme environments. Future work will focus on designing a new version of the measurement platform and exploring sensor fusion techniques to improve the robustness of the mapping process.
Stats
The platform used to record data was equipped with a lidar (Robosense RS-16), two Inertial Measurement Units (IMUs) (Xsens MTi-10, Vectornav vn100), and a barometric pressure sensor (DPS310).
Quotes
"Weather conditions were very erratic. Within a few hours, operators sustained sudden shifts ranging from hail, rain, and fog to snow and freezing rain, highlighting the unpredictable nature of the environment."
"Deploying rugged platforms emerged as imperative, with our experience revealing a vulnerability in the form of a non-rugged computer (Raspberry Pi 4B) on the platform."
"Ultimately, these conditions led to the necessity to compute the map offline at a much lower rate, since the lack of features, and therefore constraints, and the high-velocity movements prevented the ICP registration from converging."