Periodic Reporting for period 2 - AERO-TRAIN (AErial RObotic TRAINing for the next generation of European infrastructure and asset maintenance technologies)
Periodo di rendicontazione: 2023-01-01 al 2024-12-31
The AERO-TRAIN project addressed the problem of the current lack of competencies, within the field of aerial robotics applied to Operation and Maintenance of infrastructures and brought technologies with high research content to use by the industry.
AERO-TRAIN filled gaps between the infrastructure O&M industry and Industry 4.0 with the ambition to keep our invaluable assets operational and safe.
The project had as an objective the education and training of 15 highly skilled ESRs, and promotion of innovative and entrepreneurial aspects. The overall goal of the network was to train ESRs to reduce the costs associated with O&M operations, while increasing the safety aspects related to the asset management, through the design, development and testing of aerial robotic technologies, autonomous systems and remote human-machine interfaces by de-skilling the operation of innovative aerial manipulation technologies.
Moreover, the project focused on developing new approaches for improving the robustness of aerial manipulators under conditions that reflected real application, such as the application of relevant forces for tool manipulation and robust perception in featureless and light challenging environments. AERO-TRAIN took a unique human-centred direction, and it addressed the scientific issues related to increasing the maneuverability for remote aerial manipulation by developing new immersive technologies, augmented and mixed reality for damage assessment by a remote expert, and collaborative human-machine intelligence for supported evaluation of criticalities.
The network was organized into three corresponding scientific Work Packages. To achieve its objectives, AERO-TRAIN researched and developed automation technologies based on aerial manipulators, and it identified that the solution lay in the synergy between three main research pillars: Intelligent Mechatronics, Artificial Intelligence, and Human-Machine Interaction.
AERO-TRAIN filled gaps between the infrastructure O&M industry and Industry 4.0 with the ambition to keep our invaluable assets operational and safe.
The project had as an objective the education and training of 15 highly skilled ESRs, and promotion of innovative and entrepreneurial aspects. The overall goal of the network was to train ESRs to reduce the costs associated with O&M operations, while increasing the safety aspects related to the asset management, through the design, development and testing of aerial robotic technologies, autonomous systems and remote human-machine interfaces by de-skilling the operation of innovative aerial manipulation technologies.
Moreover, the project focused on developing new approaches for improving the robustness of aerial manipulators under conditions that reflected real application, such as the application of relevant forces for tool manipulation and robust perception in featureless and light challenging environments. AERO-TRAIN took a unique human-centred direction, and it addressed the scientific issues related to increasing the maneuverability for remote aerial manipulation by developing new immersive technologies, augmented and mixed reality for damage assessment by a remote expert, and collaborative human-machine intelligence for supported evaluation of criticalities.
The network was organized into three corresponding scientific Work Packages. To achieve its objectives, AERO-TRAIN researched and developed automation technologies based on aerial manipulators, and it identified that the solution lay in the synergy between three main research pillars: Intelligent Mechatronics, Artificial Intelligence, and Human-Machine Interaction.
All 15 ESRs were employed at their respective institutions and enrolled in PhD programmes. All ESRs have been publishing and disseminating their scientific results. Some of the ESRs work has already received scientific awards.
Three AERO-TRAIN training events combined with teambuilding were realised and took place. The first training school took place in Seville, Spain with a science-based focus on design, control, planning, navigation, and state estimation for Unmanned Aerial Systems and aerial manipulators for Inspection and Maintenance. A use-case workshop was held at the end of the training module. The second training event, was held in Lyngby Denmark and had as a focus guidance, navigation and control as well as an entrepreneurship workshop. A second use-case and application of aerial technologies and AI workshop was held at the end of the module.
The third training school was held in Lulea, Sweden, with scientific courses in field experimentation and advanced image processing and transferable skills in grant writing.
The training program further involved the ESRs' participation in the integration weeks, secondments, and the Grand Challenge as well as the organization of the Final Event and Summer School, largely organized by the ESRs themselves.
Transferable skills webinars have been organized throughout 2022, including IPR, dissemination, literature review ethics and social responsibilities for scientists.
The Supervisory Board (SB) and the Consortium have given high priority early on, to providing the ESRs with a broad technological perspective, a common background and to promote entrepreneurship.
Three AERO-TRAIN training events combined with teambuilding were realised and took place. The first training school took place in Seville, Spain with a science-based focus on design, control, planning, navigation, and state estimation for Unmanned Aerial Systems and aerial manipulators for Inspection and Maintenance. A use-case workshop was held at the end of the training module. The second training event, was held in Lyngby Denmark and had as a focus guidance, navigation and control as well as an entrepreneurship workshop. A second use-case and application of aerial technologies and AI workshop was held at the end of the module.
The third training school was held in Lulea, Sweden, with scientific courses in field experimentation and advanced image processing and transferable skills in grant writing.
The training program further involved the ESRs' participation in the integration weeks, secondments, and the Grand Challenge as well as the organization of the Final Event and Summer School, largely organized by the ESRs themselves.
Transferable skills webinars have been organized throughout 2022, including IPR, dissemination, literature review ethics and social responsibilities for scientists.
The Supervisory Board (SB) and the Consortium have given high priority early on, to providing the ESRs with a broad technological perspective, a common background and to promote entrepreneurship.
The project progressed beyond the state of the art of aerial robotic technologies as a whole. Three WPs looked at novel technologies related to the development and use of aerial robots, their system autonomies and collaboration with humans, as well as challenges related to piloting and remote operation by non-skilled personnel.
WP1 directly addressed and investigated the main challenges of aerial manipulation. Scientific research was conducted to address challenges related to the control and stability of the platform during physical interaction with the environment, by developing novel mathematical models that represented the physics of the interaction process as accurately as possible and by addressing robustness issues as well as fluid-dynamic modeling approaches in aerial robot control. Thus, this WP addressed the limitations of existing platforms to achieve tool operation on the environment. Novel design of compliant aerial manipulators further advanced the knowledge of drone technology for safe robot-environment interaction.
WP2 aimed to develop assisted autonomy for aerial inspection tasks. Contrary to full autonomy, we investigated effective feedback between the operator, the robot, and the environment to fulfil real-world industrial monitoring. The research comprised new ways to display robotic capabilities to the user, assistance systems to help humans control the platform and understand the environment, and low-latency communication to offload computation and provide user feedback. The WP further included studies on augmented reality (AR), first-person view (FPV), and third-person view (TPV), and finally concluded that the level of assistance strongly depended on the use case. Multi-degree of freedom drone operation beyond visual line of sight in cluttered environments quickly overwhelmed the operator. Consequently, progress beyond the state-of-the-art focused on methods that would offload the sensory load by providing assisted autonomy through autonomous suggestion of optimal trajectories to the user, as well as automatic crack and corrosion detection from visual data, object-level mapping to movable objects such as valves or bricks enabling generalizable robotic object displacement. Moreover, edge computing was investigated to offload high-level computations to clusters.
WP3 focused on researching remote aerial manipulation, shared autonomy, teleoperation devices and offered services to the remote operator, robust guidance and control algorithms, communication techniques etc. Progress beyond the state of the art was achieved in edge-computing for remote control, UAV remote control under varying time delays, parametric uncertainties and payload variations, as well as disturbance rejection from wind factors and precise motion control.
WP1 directly addressed and investigated the main challenges of aerial manipulation. Scientific research was conducted to address challenges related to the control and stability of the platform during physical interaction with the environment, by developing novel mathematical models that represented the physics of the interaction process as accurately as possible and by addressing robustness issues as well as fluid-dynamic modeling approaches in aerial robot control. Thus, this WP addressed the limitations of existing platforms to achieve tool operation on the environment. Novel design of compliant aerial manipulators further advanced the knowledge of drone technology for safe robot-environment interaction.
WP2 aimed to develop assisted autonomy for aerial inspection tasks. Contrary to full autonomy, we investigated effective feedback between the operator, the robot, and the environment to fulfil real-world industrial monitoring. The research comprised new ways to display robotic capabilities to the user, assistance systems to help humans control the platform and understand the environment, and low-latency communication to offload computation and provide user feedback. The WP further included studies on augmented reality (AR), first-person view (FPV), and third-person view (TPV), and finally concluded that the level of assistance strongly depended on the use case. Multi-degree of freedom drone operation beyond visual line of sight in cluttered environments quickly overwhelmed the operator. Consequently, progress beyond the state-of-the-art focused on methods that would offload the sensory load by providing assisted autonomy through autonomous suggestion of optimal trajectories to the user, as well as automatic crack and corrosion detection from visual data, object-level mapping to movable objects such as valves or bricks enabling generalizable robotic object displacement. Moreover, edge computing was investigated to offload high-level computations to clusters.
WP3 focused on researching remote aerial manipulation, shared autonomy, teleoperation devices and offered services to the remote operator, robust guidance and control algorithms, communication techniques etc. Progress beyond the state of the art was achieved in edge-computing for remote control, UAV remote control under varying time delays, parametric uncertainties and payload variations, as well as disturbance rejection from wind factors and precise motion control.