Periodic Reporting for period 1 - AUTOASSESS (Autonomous aerial inspection of GNSS-denied and confined critical infrastructures)
Reporting period: 2023-10-01 to 2024-09-30
Currently, across the world vessels must be regularly monitored for corrosion and defects to ensure structural safety. This has been done up until now by human surveyors. This entails humans going into dangerous and confined GNSS-denied areas, like water ballast tanks and cargo holds. These spaces consist of tight enclosed spaces, difficult to access, low lighting, slippery surfaces, air pockets with low/no oxygen and toxic gases. On average, one person is killed every week from accidents in enclosed spaces in marine structures, underlying the need to look for an alternative surveying solution.
Next to this dangerous environment a vessel inspection is very costly, and can reach 1MEUR per vessel. Given the fact that a fifth of the global fleet (60000) is inspected per year, this means 11BEUR cost per year.
AUTOASSESS is aiming to address both problems with the main objective of removing human surveyors out of harms way while at the same time reducing inspection costs. This can be achieved by combining the latest developments of different drone technologies across different areas (from collision-tolerant UAS to aerial manipulation and ML-based defect identification). By combining different technologies, the AUTOASSESS project aims to address the challenges of deploying aerial robots in challenging conditions. Equipped with automated AI-based scanning, mapping, navigation and contact-based NDT, the proposed solution is aimed to remove the need for human inspection.
The AUTOASSESS system comprises four main components that utilize state-of-the-art technologies. Together, they provide a complete, autonomous solution for surveying marine vessel structures.
Here is how it works:
User Interface and Decision Support System (UI-DSS): Overseeing the whole operation is the UI-DSS that provides access to all vessel data, including past inspections. Surveyors and other end users can use the system to get insight into structural integrity, identify hotspots and see thickness measurements or corrosion images annotated directly on the 3D map of the vessel. Additionally, missions can be planned from the UI-DSS which are sent directly to the robots. There are three fully autonomous robots: two for the ballast tanks (EUAS and IUAS) and one for the cargo hold (TIUAS).
Exploration Unmanned Aerial System (EUAS): This lightweight, fast, long endurance drone is designed for initial exploration, in the tight spaces of the ballast tanks. These areas are too cramped and complex for mobile robots or tethered drones. The EUAS quickly flies through the entire area, with no detailed path planning and takes pictures for a overall assessment of the tanks and further offline mapping which can then be used to plan detailed inspections using the IUAS.
Inspection Unmanned Aerial System (IUAS): This drone performs closeup inspection and thickness measurements. The IUAS uses higher accuracy sensors (eg: LiDAR) and visits specific locations (as identified in the UI-DSS via post processing of the data form the EUAS) to get high quality pictures of defects and perform NDT thickness measurements. This helps document problems and plan repairs.
Tethered Unmanned Aerial System (TIUAS): In the larger, open spaces like cargo holds, a tethered drone is used. The tether provides unlimited power, allowing it to perform all tasks —exploration, close-up inspection, and thickness measurements— without needing to stop and recharge. It uses the same advanced navigation and mapping technology as the IUAS, but allows long and very fine detail inspection.
For more information on AUTOASSESS project, please visit: https://www.autoassess.eu/(opens in new window)
Next to this dangerous environment a vessel inspection is very costly, and can reach 1MEUR per vessel. Given the fact that a fifth of the global fleet (60000) is inspected per year, this means 11BEUR cost per year.
AUTOASSESS is aiming to address both problems with the main objective of removing human surveyors out of harms way while at the same time reducing inspection costs. This can be achieved by combining the latest developments of different drone technologies across different areas (from collision-tolerant UAS to aerial manipulation and ML-based defect identification). By combining different technologies, the AUTOASSESS project aims to address the challenges of deploying aerial robots in challenging conditions. Equipped with automated AI-based scanning, mapping, navigation and contact-based NDT, the proposed solution is aimed to remove the need for human inspection.
The AUTOASSESS system comprises four main components that utilize state-of-the-art technologies. Together, they provide a complete, autonomous solution for surveying marine vessel structures.
Here is how it works:
User Interface and Decision Support System (UI-DSS): Overseeing the whole operation is the UI-DSS that provides access to all vessel data, including past inspections. Surveyors and other end users can use the system to get insight into structural integrity, identify hotspots and see thickness measurements or corrosion images annotated directly on the 3D map of the vessel. Additionally, missions can be planned from the UI-DSS which are sent directly to the robots. There are three fully autonomous robots: two for the ballast tanks (EUAS and IUAS) and one for the cargo hold (TIUAS).
Exploration Unmanned Aerial System (EUAS): This lightweight, fast, long endurance drone is designed for initial exploration, in the tight spaces of the ballast tanks. These areas are too cramped and complex for mobile robots or tethered drones. The EUAS quickly flies through the entire area, with no detailed path planning and takes pictures for a overall assessment of the tanks and further offline mapping which can then be used to plan detailed inspections using the IUAS.
Inspection Unmanned Aerial System (IUAS): This drone performs closeup inspection and thickness measurements. The IUAS uses higher accuracy sensors (eg: LiDAR) and visits specific locations (as identified in the UI-DSS via post processing of the data form the EUAS) to get high quality pictures of defects and perform NDT thickness measurements. This helps document problems and plan repairs.
Tethered Unmanned Aerial System (TIUAS): In the larger, open spaces like cargo holds, a tethered drone is used. The tether provides unlimited power, allowing it to perform all tasks —exploration, close-up inspection, and thickness measurements— without needing to stop and recharge. It uses the same advanced navigation and mapping technology as the IUAS, but allows long and very fine detail inspection.
For more information on AUTOASSESS project, please visit: https://www.autoassess.eu/(opens in new window)
AUTOASSESS has progressed significantly towards the goal of developing autonomous inspection of marine vessels using aerial robotics. The main achievements of the project in the first period are as follows.
Novel Exploration robot:
Designed and prototyped the EUAS (“Wabble“) which is a lightweight flexible solution able to achieve 20+ minutes of flight time. The robot is highly resilient, and can withstand collisions at high speeds without sustaining damage.
Innovative Non-Destructive Sensing technology:
Created and tested miniaturized ultrasonic (UT) and Pulse Eddy Current (PEC) sensors. The tests proved that PEC could be a valid alternative to UT.
Machine Learning Advances:
Performed dataset collection and iterative annotation pipeline that combines non-expert and expert inputs to enhance labeling efficiency. Built initial models for defect detection, probabilistic segmentation, unsupervised domain adaptation and overall surveying.
Integration and Interface:
Defined data formats and pipelines for seamless communication between ground UAVs, mapping server and user interface. User interface now supports NDT measurements as well as Gaussian splatting for photo-realistic 3D rendering.
Autonomy and Path Planning:
Introduced Semantics-aware Predictive Planning that optimizes inspection efficiency by reducing inspection time, especially in repetitive environments. Developed learning-based control algorithms for fast, agile, precise maneuvering through narrow openings and confined spaces.
Testing and Validation:
Conducted early field tests in ballast tanks and cargo holds.
Novel Exploration robot:
Designed and prototyped the EUAS (“Wabble“) which is a lightweight flexible solution able to achieve 20+ minutes of flight time. The robot is highly resilient, and can withstand collisions at high speeds without sustaining damage.
Innovative Non-Destructive Sensing technology:
Created and tested miniaturized ultrasonic (UT) and Pulse Eddy Current (PEC) sensors. The tests proved that PEC could be a valid alternative to UT.
Machine Learning Advances:
Performed dataset collection and iterative annotation pipeline that combines non-expert and expert inputs to enhance labeling efficiency. Built initial models for defect detection, probabilistic segmentation, unsupervised domain adaptation and overall surveying.
Integration and Interface:
Defined data formats and pipelines for seamless communication between ground UAVs, mapping server and user interface. User interface now supports NDT measurements as well as Gaussian splatting for photo-realistic 3D rendering.
Autonomy and Path Planning:
Introduced Semantics-aware Predictive Planning that optimizes inspection efficiency by reducing inspection time, especially in repetitive environments. Developed learning-based control algorithms for fast, agile, precise maneuvering through narrow openings and confined spaces.
Testing and Validation:
Conducted early field tests in ballast tanks and cargo holds.
The beyond-state-of-the-art key innovation results achieved so far in AUTOASSESS are as follows:
- Creation of a novel, resilient, collision-tolerant, long-flight-time drone platform for confined area exploration
- Development of novel miniaturized ultrasonic and eddy current modules for contact-based inspection
- Invention of an innovative Semantics-aware Predictive Planning that improves the efficiency of exploration.
- Development of domain adapted probabilistic segmentation for inspection
- Novel design for enhanced contact-based to applied on underactuated multicopters
- Creation of a novel, resilient, collision-tolerant, long-flight-time drone platform for confined area exploration
- Development of novel miniaturized ultrasonic and eddy current modules for contact-based inspection
- Invention of an innovative Semantics-aware Predictive Planning that improves the efficiency of exploration.
- Development of domain adapted probabilistic segmentation for inspection
- Novel design for enhanced contact-based to applied on underactuated multicopters