can move along the surface of steel bridges was manufactured. The climbing robot measures 12 × 10 × 12 in. (30.5 × 25.4 × 30.5 cm) and weighs 40 lb with each battery good for approximately one hour. It is equipped with 2 GoPro cameras, a LiDAR, and an IMU. Its ground control station (GCS) has 16 channels with a control range of 2 km. It has a NUC i7 computer, a monitor, and a joystick. The crawler includes two independent channels of robot teleoperation and monitoring. A Dell Optiplex computer processes the data from the LiDAR and cameras and is then transferred to the GCS for live monitoring. For remote control, Rx receives commands from the operator via the GCS’s joystick. An Arduino Mega processes these signals and then exports outputs to control the robot’s motors and steering servos via power amplifiers. The structural crawler in Figure 3b is mostly applicable to large areas with little or no obstacles. Transverse climbing of the crawler around an I-shaped beam or girder is not a trivial task. A climbing robot may have an insufficient footprint or inappropriate orientation to make a safe turn from the inner to outer face of a flange of I-beams or girders. In addition, the unexpected disengagement of a climbing robot from its attached steel bridge may be a safety concern for passengers on underpass highways or over rivers. Therefore, innovative UAVs that can interact with a bridge deck for its enhanced inspection and maintenance were recently proposed. Two example vehicles are presented and demonstrated in Figures 3c and 3d. (Note that Figure 3c is still in laboratory test stage and Figure 3d was tested on an actual bridge.) Additionally, more recent generations of the structural crawlers described in Nguyen et al. (2021) and Otsuki et al. (2022) improved mobility around acute angles. Figure 3c shows an uncrewed multimodal vehicle, called BridgeBot, that was driven by four propellers in a flying mode and by four DC motors in a traversing mode. Both actuation systems were powered by batteries. The mechanics of train wheels and beam trolleys were followed to allow the inspection vehicle to be attached to the edge of an I-beam bottom flange. The four clamping wheels were 3D printed and coated with urethane to promote a higher coefficient of friction against the beam. Each wheel was independently driven by an electric DC motor. As shown in Figure 3c, the BridgeBot can mimic the operation of a traditional inspection platform as shown in Figure 3a. The uncrewed vehicle will be used to facilitate the I-girder bridge inspection and deploy structural crawlers, such as shown in Figure 3b, on the bottom flange of bridge girders. The uncrewed vehicle and the structural crawlers will allow the inspection and local maintenance of more than 90% of the bridges in the National Bridge Inventory. The uncrewed vehicle can fly in air and traverse along a girder with an effec- tive vehicle-bridge engagement mechanism for a smooth transition from flying to traversing mode, or vice versa. Design criteria of the hybrid vehicle include, but are not limited to, the following: Ñ In the flying mode, the vehicle is stable with necessary posi- tioning precision and navigation guidance in a GPS-denied environment. Ñ In the traversing mode, the vehicle with necessary posi- tioning precision moves at a constant speed to provide a stable station for various measurements. Figure 3. Bridge-attached platforms for close-in bridge inspection: (a) the moving platform in current practice (b) the proposed structural crawler with four magnet wheels, which carries a RGB camera for visual inspection (c) the proposed hybrid flying and traversing uncrewed vehicle (BridgeBot) attached to a wooden deck and (d) the proposed ceiling UAV in contact with the deck, which carries a dual-sensor (thermal and optical) camera. ME | AERIALNDTFORBRIDGES 70 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 70 12/20/22 8:15 AM

Ñ In the transition period between the flying and traversing modes, the vehicle engages with a girder safely and effi- ciently. Ñ Overall, the hybrid vehicle must have the required payload for measurement devices and structural crawlers to be deployed, and the required flight time for local maintenance or complete inspection of one bridge. The uncrewed multimodal vehicle was successfully tested to demonstrate its flying and traversing functions and its system performance. First, the vehicle as a drone flew to the underside of a simulated wooden bridge girder. Once directly underneath the girder, the vehicle grabbed the bottom flange of the girder with a specifically designed roller clamping system and then traversed the bridge at a constant speed. Finally, the vehicle simply detached from the bridge as it encountered any obstacles and flew to the next area of interest. The overall per- formance of the vehicle met the design requirements. However, the vehicle system could be improved by stiffening several components to avoid any potential vibration during landing that would cause damage (Reven et al. 2019). Figure 3d shows a ceiling UAV that is in direct contact with the ceiling of a room for close-distance ( 10 in. [ 25.4 cm]) inspection. The ceiling UAV is a hexacopter that is installed with a dual-sensor (thermal and optical) camera. The main difference of the ceiling UAV from a conventional hexacopter is the addition of a top frame to make the hexacopter in firm contact with the ceiling. The top frame is composed of two 10 in. (25.4 cm) wide legs 14 in. (35.5 cm) apart. Once in position against the building ceiling as shown in Figure 3d, the UAV enabled consistent imaging of the ceiling at a known standoff distance and thus provided high-quality thermal and RGB images as illustrated in Figure 3d. As the UAV approached the underside of a building ceiling, the required throttle to maintain a certain speed was reduced exponentially (Jiao et al. 2021). Therefore, moving the UAV along the building ceiling or a bridge deck is also energy-efficient in practice. Both the hybrid uncrewed vehicle in Figure 3c and the ceiling UAV in Figure 3d can be used to support various NDE methods, such as active thermography and GPR. The capa- bility of the active thermography and GPR for the detection of subsurface defects is demonstrated using embedded defects in reinforced concrete (RC) bridge decks in the following section. Example NDE Methods Appropriate for Deployment on Robots A 6 × 3.75 ft (1.8 × 1.1 m) RC bridge deck 8 in. (20 cm) thick with embedded defects mimicking delamination in application was designed and cast to test the effectiveness of various NDE methods for defect detection. Two types of delamination with different dimensions embedded into the RC deck were desig- nated as small and large delamination. All delamination was simulated using a combination of foam strips, each measuring 6 in. (15 cm) and 2 in. (5 cm) in length and width. The width of each strip is equal to the thickness of commercially available foam boards, which is 2 in. (5 cm). The length of each strip is equal to the length of a small plastic board, which is 6 in. (15 cm), to support the foam strips during concrete casting. Each small delamination consists of two strips side by side so that its dimension is 6 × 4 in. (15 × 10 cm). Each large delamination is composed of eight strips side by side in pattern to a total dimension of approximately 12 × 10 in. (30.5 × 25 cm) after the gap between the two strips has been taken into account. Small and large delamination objects were supported on 6 × 6 × 5/64 in. (15 × 15 × 0.8 cm) and 12 × 12 × 5/64 in. (30 × 30 × 0.8 cm) plastic boards, respectively. Each plastic board was embedded at two depths, 1.0 and 1.5 in. (2.5 and 3.8 cm) from the top and bottom concrete surface, respectively. This depth scenario represents the concrete cover used in dif- ferent RC members in bridges. The RC deck was heated at the delamination area by a 1 × 4 ft (0.3 × 1.2 m) trip heater for 10 min from both the top and bottom concrete surfaces, respectively. The heater was then removed so the data could be collected for 20 min during the cooling phase. Figures 4a–4d present the infrared images of the delamination area at 20 min of cooling. The concrete surface immediately above the embedded delamination is hotter than its surrounding area. The thicker the delamina- tion, the more the accumulation of heat and thus the higher the concrete surface temperature as indicated in Figure 4a through 4d when cooling inside the Highbay Laboratory at Missouri S&T. The passive thermography under sunlight from a drone at approximately 50 ft (15.2 m) above the concrete deck is presented in Figure 4e. Note that passive thermography was applied to field inspection (Sakagami 2015). In compari- son with Figure 4a through 4d, passive thermography remains effective in detecting the large delamination areas but less effective for the small delamination areas in four RC decks. 1.5 in. cover and 5 mm foam 1.0 in. cover and 5 mm foam 1.5 in. cover and 3 mm foam 1.0 in. cover and 3 mm foam 1.0 in. cover and varying foam thickness Figure 4. Thermal images of delamination with varying concrete covers and foam depths: (a) 1.5 in. (3.8 cm) cover and 5 mm foam (b) 1.5 in. (3.8 cm) cover and 3 mm foam (c) 1 in. (2.5 cm) cover and 5 mm foam (d) 1 in. (2.5 cm) cover and 3 mm foam (e) 1 in. (2.5 cm) cover and varying foam thickness. 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 71 2301 ME Jan New.indd 71 12/20/22 8:15 AM