CATEC has recently published a new scientific article on drone inspection, entitled ‘Autonomous UAV Pipeline Landing via Depth–LiDAR Fusion in GNSS-Denied Refineries’, which has appeared in the journal IEEE Robotics and Automation Letters (RA-L), one of the most prestigious publications in the field of international robotics.
The paper, authored by David Tejero Ruiz, Marco Antonio Montes Grova, Francisco Javier Pérez Grau, Antidio Viguria and Anibal Ollero, addresses a very specific industrial inspection problem: landing a UAV with precision on pipelines in refineries without GNSS, where occlusions, interference and reflective surfaces make perception particularly challenging.
The article addresses this challenge by integrating two components: (1) LiDAR-based localisation and (2) a pipeline detection, mapping and tracking module that combines a depth camera and two 2D LiDAR sensors, utilising cylindrical registration via non-linear optimisation, geometric association and Kalman filters to maintain persistent hypotheses and filter out false positives.
This system has been validated in controlled indoor experiments and in fully autonomous outdoor missions at a real refinery, achieving sub-3 cm accuracy in the final pipe estimation during the landing phase. The paper also includes a supplementary video showing the tests carried out.
This work has been carried out as part of the European SIMAR Project (under the Horizon Europe programme), with the aim of advancing the automation of inspections in refineries, thereby improving safety and reducing operating costs.
You can access the article published on IEEE Xplore (Early Access) via this link: https://ieeexplore.ieee.org/document/11373726
Access the demo video for this article below:
Abstract:
Industrial inspection in complex environments such as refineries poses significant challenges for autonomous systems, particularly when GNSS signals cannot be used. In this context, the use of drones capable of operating with precision and safety opens up new opportunities to optimise maintenance tasks, reduce risks to personnel and increase operational efficiency.
An advanced architecture has recently been developed that enables a drone to perform high-precision autonomous landings on industrial pipelines, even in GNSS-denied environments. The system combines different sensors — LiDAR, depth cameras and laser scanners — to localise the drone in space and reliably detect the geometry of the pipelines, overcoming the limitations of single-sensor-based solutions.
One of the key elements is the intelligent fusion of data from various sensors, which enables individual pipes to be identified and tracked even in scenarios involving multiple nearby lines, reflective surfaces or adverse lighting conditions. This information is integrated in real time to generate a precise landing reference, aligned both in position and orientation with the target pipe.
Thanks to this approach, the drone is not only able to approach safely, but also to land with centimetre-level precision. This capability is particularly relevant for operations requiring physical contact with the infrastructure, such as the deployment of non-destructive inspection systems or crawling robots directly onto the pipes.
The system has been validated in both controlled environments and at a real refinery, demonstrating robust and reliable performance during fully autonomous missions. The results confirm that the combination of multi-sensor perception and temporal tracking provides stable and accurate estimates even during the most critical phases of the manoeuvre.
This breakthrough represents a significant step towards more autonomous, safe and versatile aerial inspection solutions for the Oil & Gas industry and other industrial sectors. The ability to operate without GNSS and to interact directly with complex infrastructure significantly broadens the range of possible applications for industrial drones in real-world scenarios.
