The 36-month H2020
AErial RObotic System for In-Depth Bridge Inspection by Contact
(AEROBI) project started on 1st December 2015. This project, targeting the bridge inspection domain, is adapting and integrating recent research results in low-flying unmanned robots with arms, intelligent control in robotics, computer vision and sensing, in an innovative, integrated, low-flying, robotic system with a specialised multi-joint arm that will scan concrete beams and piers in bridges for potential cracks on the surface or for detecting and identifying significant visible defects such as concrete swelling or spalling.
The main objectives of the AEROBI project are:
- To develop an incrementally optimised aerial robotic system with a multi-joint arm for the in-depth inspection and assessment of reinforced concrete bridges
- To develop a computer vision system for the structural inspection of the bridge that will detect, identify and map structural defects (cracking, swelling, spalling)
- To determine appropriate methodologies for collecting data
- To extend already existing tools so that they can be able to precisely detect and quantify bridges' defects
- To develop a sensor system suitable for being installed on the robotic inspection system, as well as laser devices and quantitative, mechanical models for structural assessment
- To develop a quantitative structural assessment tool that, based on input from the inspection of the bridge, construction information and information on the operative environment, will automatically assess the structural condition and safety of the bridge at the time of inspection and at future times so that bridge managers can decide on an immediate intervention or on the time for the next inspection.
The AEROBI project has just completed its first year. Following a complete and comprehensive document on user requirements (all the aspects of the inspection of a bridge) produced by project partners Egnatia Odos AE and Netivei Israel, AEROBI started the development phase. The first step involved designing the overall system (air vehicle, data link and Ground Control Station). This helped define the specifications that will be the basis for system acceptance after the integration phase. Extensive work has been undertaken on the computer vision capabilities. Large sets of bridge defects/damage pictures have been used to understand precisely what needs to be detected, its extent to measured and its severity to be evaluated.
Due to the complexity of the system, the approach chosen is an iterative development, especially for the aerial vehicle capability. It was therefore necessary to perform preliminary trials early in the development phase to identify the possible challenges and risks. These preliminary trials took place on 19th and 20th December 2016 on a real bridge 150 km south-east of Seville. The bridge is operational and starting to show signs of ageing (the upper part was destroyed by a storm two years ago), so the trials were very realistic and enabled the unmanned aerial vehicle (UAV) navigation capabilities to be tested. The main focus of the trials was to detect and identify anomalies with the camera mounted on the UAV platform. The results have been used to refine the spcifications for the AEROBI platform and make the computer vision algorithms more robust. The campaign was concluded by a de-briefing session with the end-users who were given the opportuniy to amend or complete their needs and specifications.
At the same time, during the second part of the first year, the theoretical aspect of the structural assessment tool began to be laid out and formulated and the respective code developed. This entailed developing new relations between the deck, pier and bearings deformations and cracks observed and - through a reverse structural analysis - the developed internal forces at the respective bridge cross-sections that will allow the actual structural condition of the bridge and the respective safety factors to be estimated.
AEROBI has reached an important milestone as the project now has a first working version of the system which will support future developments. The end-users acknowledge this important step that demonstrated that the visual inspection of a span could be performed in less than 30 minutes.
The video of the first AEROBI trial can be found