While the data collection speed of RABIT is significantly higher than of the manual data collection (approximately three times higher than a team of five NDE technicians), it could be significantly increased through the use of air-coupled and/or rolling probes that would eliminate RABIT’s test point stops. As an example, the use of air-coupled acoustic and vertical electrical impedance (alternative to ER) testing, along with GPR and high-definition imaging, was implemented on an NDE platform that enabled data collection at a walking speed (Pashtouni et al. 2020). Climbing Robots for Bridge Superstructure and Substructure The development of climbing robot systems for bridge inspec- tion has received great attention recently (Tirthankar et al. 2018 Nguyen and La 2021). Inspired by the way that animals and insects move, robots have demonstrated the feasibility of climbing over different connectors and surfaces on bridges (Minor et al. 2000 Nguyen and La 2019 Nguyen et al. 2020). However, each bridge has many locations to be checked, and they are usually not close together, so it will take a long time for those climbing robots to complete the inspection of a bridge, not to mention that the calculation to move also takes a lot of time and requires intelligent algorithms. Studies on using drones for inspection have found that drones allow a quick, efficient overview without being limited to the bridge element material. However, a comprehensive inspection of bridges requires multiple positions for in-depth testing, while the current drone capabilities can provide only a visual inspec- tion. Unlike the approaches mentioned above, a new hybrid robotic design is presented, which considers the advantages of a drone’s flying flexibility and a mobile robot’s steady climbing capability to perform in-depth inspections of bridges. With the new design, the in-depth inspection of the bridge will also be conducted faster because of the drone’s maneuverability. The mobile robot part is equipped with permanent magnets that can change the distance from the steel surface. Changing the distance between the magnet and the steel surface allows the robot to switch its operating modes between landing, taking off, and moving. The design concept of this robot is illustrated in Figure 5. The robot is integrated with multiple sensors: Intel camera T265 and GPS for robot location tracking, infrared sensors for a safe landing, and a GMR sensor for crack detection. The onboard computer for processing is a Raspberry Pi 4. The PX4 flight controller is for controlling the robot. The robot is sur- rounded by a sphere cage to protect it from a collision with the bridge. Overall, the robot is designed to work in two modes: ME | NDEOFBRIDGES 0 10 20 10 20 30 30 40 50 60 70 80 Longitudinal distance (ft.) Longitudinal distance (ft.) 90 100 0 10 20 30 40 50 60 70 80 90 100 Sound Fair Poor Serious (kOhm-cm) (ksi) 110 120 130 140 150 160 ER 0 10 20 10 20 30 30 40 50 60 70 80 90 100 110 120 130 140 150 160 IE 0 10 20 10 20 30 30 40 50 60 70 80 Longitudinal distance (ft.) Longitudinal distance (ft.) 90 100 110 120 130 140 150 160 GPR 0 10 0 3000 4000 5000 6000 (dB) –30–25–20–15–10 –8 –7 –6 –5 –4 –3–2.5–2–1.5–1–0.5 0 0.5 1 1.5 2 2.5 20 10 20 30 30 40 50 60 70 80 90 100 110 120 130 140 150 160 USW Figure 4. Condition maps from four NDE technology surveys: (a) ER (b) IE (c) GPR (d) USW and (e) section of the deck surface image. 60 M AT E R I A L S E V A L U AT I O N • J A N U A R Y 2 0 2 3 2301 ME Jan New.indd 60 12/20/22 8:15 AM Lateral distance (f Lateral distance (f Lateral distance (f Lateral distance (f

mobile and drone. The mobile part allows the robot to cling to steel surfaces and move like other conventional mobile robots. A new robot version will be able to cling to any material type surface, including concrete. On the robot, the body is attached to magnets to create an attractive force when the robot clings to a steel surface. The distance between the magnet array and the steel face is controlled by two pull motors. This distance is adjusted depending on the working condition of the robot. If the robot needs to cling strongly to the structure, then the distance is 0. Several images of the robot climbing on vertical structures of the bridge are shown in Figure 5. The robot was in mobile mode, using its wheels to climb on the elements. As shown, the robot can climb on tall elements and access dif- ficult-to-reach areas to perform inspections. If the robot needs to climb on the bridge, then the distance is larger than 0 (e.g., 0.3 cm) to allow easy crawling of the robot. The drone mode allows the robot to fly between inspection areas. In this mode, the robot uses a high-resolution camera to capture images or record video of the surface of the bridge elements and joints. At the same time, the robot sends images to the inspector for live viewing. To navigate the robot in the drone mode, the inspector wears a VR headset (5.8 GHz) to receive video data from the robot’s camera for observation of the environment at a distance of up to around 2 km. The robot uses its camera to detect available surfaces to land on the element (switching from the flying to the mobile mode). The robot relies on the values of two infrared sensors along with an intelligent landing algorithm to determine if the point in front is a safe position to land. The robot was deployed to collect data on several bridges. Both mobile and drone modes were tested to validate the design. The left side of Figure 6 shows a snapshot of a video recording of a highway bridge taken by the robot in the flying mode. There are two high-resolution images on the right of the figure showing cracked pier surface areas from the robot’s flight close to the pier. Advanced Data Visualization Like the automation discussed above, 3D visualization of NDE data is another area where improvements can be made for a more comprehensive and effective assessment of concrete bridges. Visualization allows NDE data to be presented in an intuitive manner, which will, in turn, facilitate the under- standing and further analysis of NDE test results. Unlike the conventional data processing methods, which are used to detect the presence of anomalies through measurement of GPS module RASPBERRY Pi 4 Flight controller Battery Frame GMR sensor Mobile robot IR sensor Magnets Tracking camera Figure 5. Flying robot with sensor integration: (a) overall design (b) as built and (c) images of the robot climbing on vertical surfaces of steel bridge elements. J A N U A R Y 2 0 2 3 • M AT E R I A L S E V A L U AT I O N 61 2301 ME Jan New.indd 61 12/20/22 8:15 AM