Autonomous aerial inspection of GNSS-denied and confined critical infrastructures

The >50k large vessels across the world must be regularly monitored for corrosion and defects by human surveyors, including dangerous and dirty confined GNSS-denied areas such as ballast water tanks and cargo holds. One person is killed every week from accidents in these enclosed spaces, which despite having large surface areas, consist of many smaller, confined compartments with narrow passages (40cmx60cm). However, a radical new approach is possible using unmanned aerial systems (UAS or drones), by combining the latest developments in (1) collision-tolerant UAS, (2) multi-modal SLAM, (3) path planning, (4) autonomous drone racing, (5) aerial manipulation, (6) miniaturized NDT sensors, and (7) ML-based defect identification. Only through a complete integration of these technologies is it possible to address the challenges of deploying aerial robots in these challenging conditions. Equipped with automated AI-based scanning, mapping, navigation and contact-based NDT, this has the potential to completely remove the need for human inspection. Using a digital twin approach brings “superhuman” results: comprehensive semantic-aware detailed 3D mapping (1 cm resolution) of large structures (>300 m), high resolution visual and NDT analysis (100um) and improved traceability with automatically generated trend analysis. The ML for system mapping and NDT is trained with sociotechnical inputs from experienced human inspectors.Currently, a typical inspection costs >1M€ and requires 15 days (8 days inspection and 7 days travel to low cost Far Eastern docks). A UAS-based inspection will take 1 day, with 1-2 days travel to an EU port at a cost of 200k€, saving the industry >9B€ p.a. with 2.4MT of CO2 reduction. This consortium includes many of the world leaders in the field of UAS-based inspection teamed with vessel owners and inspectors, enabling an end-to-end survey solution which would save 50 lives/yr, and provide safer, more reliable, and accurate inspection data.