Proceedings of the 25th International Conference on

Automation & Computing, Lancaster University,

Lancaster UK, 5-7 September 2019

Inspection Robots in Oil and Gas Industry: a

Review of Current Solutions and Future Trends
Leijian Yu1, Erfu Yang1*, Peng Ren2, Cai Luo3, Gordon Dobie4, Dongbing Gu5, Xiutian Yan1
1 3
Department of Design, Manufacture and Engineering College of Mechanical and Electronic Engineering,
Management, China University of Petroleum (East China),
University of Strathclyde, Qingdao 266580, China
4
Glasgow G1 1XJ, Scotland, UK Department of Electronic and Electrical Engineering,
*Corresponding author: [email protected] University of Strathclyde,
2
College of Information and Control Engineering, Glasgow G1 1RD, Scotland, UK
5
China University of Petroleum (East China), School of Computer Science and Electronic Engineering,
Qingdao 266580, China University of Essex,
Colchester C04 3SQ, UK

Abstract— With the increasing demands for energy, oil and inspection robots will steadily increase in the foreseeable
gas companies have a demand to improve their efficiency, future, and in 2015, the global capital expenditure of these
productivity and safety. Any potential corrosions and cracks robots will reach 2.85 billion dollars. Based on vehicle type,
on their production, storage or transportation facilities could the market of inspection robots in oil and gas industry can
cause disasters to both human society and the natural be divided into the following categories: Remotely
environment. Since many oil and gas assets are located in the operated vehicles (ROVs), Autonomous underwater
extreme environment, there is an ongoing demand for robots vehicles (AUVs), Unmanned ground vehicles (UGVs) and
to perform inspection tasks, which will be more cost-effective Unmanned aerial vehicles (UAVs) [5]. These robots have
and safer. This paper provides a state of art review of
different mechanisms and structures for different
inspection robots used in the oil and gas industry which
including remotely operated vehicles (ROVs), autonomous
inspection tasks. Some of them are focused on inspecting
underwater vehicles (AUVs), unmanned ground vehicles oil storage tanks, while some of them are designed for
(UGVs) and unmanned aerial vehicles (UAVs). Different pipeline inspection. Nevertheless, most of them need
kinds of inspection robots are designed for inspecting experienced engineers to manipulate them to conduct the
different asset structures. The outcome of the review suggests inspection process. Greater levels of robot’s flexibility and
that the reliable autonomous inspection UAVs and AUVs will autonomy can make the inspection more intelligent and
gain interest among these robots and reliable autonomous efficient. To have a comprehensive understanding
localisation, environment mapping, intelligent control inspection robot in oil and gas industry, this paper focuses
strategies, path planning and Non-Destructive Testing (NDT) on reviewing the key technologies in different kinds of
technology will be the primary areas of research. inspection robots, discussing the challenges and
highlighting the trends in future research, which will make
Keywords-inspection robots, oil and gas industry, the inspection process more efficient, intelligent and cost-
autonomous inspection, defects detection, asset integrity effective.
I. INTRODUCTION
The Oil and Gas sector is an underpins modern society.
In the 2018 Energy Outlook, British Petroleum (BP)
predicted that the absolute consumption of oil and gas
would have steady growth to 2040 [1]. The facilities used
in oil and gas industry for exploration, production, storage,
transportation and distribution are expensive and important
assets. If they are not well maintained, their failures may
cause catastrophic consequences for both the natural
environment and human society. In the Gulf of Mexico, the
Deepwater Horizon drilling explosion killed more than
600,000 animals [2]. The Amuay refinery blast injured
more than 80 people and damaged several buildings near Figure 1. Inspection robots in oil and gas industry [4].
the plant [3]. Therefore, regular and precise inspection is
necessary not only for production but also for safety. As the ROVs and AUVs are both underwater unmanned
Sensor network technologies are usable for this purpose. vehicles (UUVs), they will be covered in the related
However, in the hazardous environment, it is difficult or sections UUVs in this paper. The rest of the paper is
expensive to reach and deploy the sensor networks by organised as follows. In Section Ⅱ, the common sensors
human. used for inspection robots are presented. while the current
inspection robot solutions are introduced in Section Ⅲ.
For these reasons, the inspection robots are attractive to
Then, Section Ⅳ discusses the challenges and future trends
the industry. Fig. 1 from [4] shows that the demand of the
in high-level flexibility and autonomy of inspection robots. III. CURRENT INSPECTION ROBOTS REVIEW
Finally, this paper is concluded in Section Ⅴ.
A. Unmanned Ground Vehicles
II. INSPECTION METHODS AND SENSORS According to the shape and function of the onshore
For the inspection purposes in oil and gas industry, a facilities in oil and gas industries, the equipment can be
number of onboard NDT sensors are used to detect roughly divided into vertical structures and pipelines. The
corrosions and cracks. According to different types of vertical structures contain drilling, production and storage
sensors, the most commonly applied inspection assets, such as flare stacks and tanks. Pipelines are mostly
technologies can be roughly divided into four classes, i.e., focused on transportation purpose. To inspect these kinds
visual inspection, ultrasonic inspection, magnetic of facilities, UGVs are the most popular choice nowadays.
inspection and eddy current inspection.
1) Vertical Structure Inspection
Visual inspection is one of the oldest and popular For inspecting the vertical structures, wall climbing
inspection means. At the very beginning, experienced robots have gained great interests. The climbing
engineers used their naked eye to check the condition of technologies are the main difference between these robots.
assets. But now, cameras have allowed the inspection At the same time, the most important task in the design and
robots have the ability to pursue a view of the structure. development of a climbing robot is to develop an
When in visual degradation environment, LEDs can be appropriate mechanism to ensure that the robot adheres to
used to deal with the absence of proper illumination. Visual different types of walls and surfaces reliably without
inspection is simple and it is one of the easiest inspection sacrificing its mobility. According to the adhesion and
techniques to perform. However, it is only suitable for locomotion principles, the climbing methods can be
detecting damages on the surface and the inspection quality cauterised as magnetic adhesion mechanism, pneumatic
is sensitive to illumination [6]. adhesion mechanism and bio-inspired grasping grippers
[10], as shown in Fig. 2.
Besides the visual inspection mentioned previously, the
ultrasonic sensor is another primary sensor found on many
inspection robots. Ultrasonic sensors can emit and receive
the ultrasound, which is propagated into the material.
Cracks can be detected by measuring the time difference
between the generated and reflected ultrasound. The
comparison of these two signals tells the location and size
of the crack. There are many advantages in using ultrasonic
sensors, such as high accuracy, high sensitivity and
suitability for monitoring all kinds of materials. However,
it will not work when the defect lays along the line of wave
travelling [7].
Magnetic sensitive sensors work with ferrous material
assets. After applying a magnetic field on these facilities,
most of the magnetic flux lines will go through these metal
materials. If there is a defect, magnetic flux lines will be
bent. Some of the magnetic flux lines will leak out. The Figure 2. (a) a magnetic adhesion robot [11], (b) a pneumatic adhesion
magnet sensitive sensors can detect the magnetic leakage robot [14], (c) a cat insperied adhesive robot [15], (d) a gecko inspired
field. The detected signal can be analysed to reveal the robot [16].
changes in structure. This method can realise relatively
high-speed inspection and very good sensitivity to pitting. Vertical structures in the oil and gas company are
However, it can only work for ferromagnetic materials and usually made of carbon steel. Since this kind of material is
the sizes of defects detected are very limited [8]. ferromagnetic, magnetic adhesion can be highly desirable
in this kind environment. There is a lot of work utilising
Eddy current inspection is similar to magnet sensitive magnetic adhesion to inspect these facilities. One method
inspection in some ways. It uses eddy currents generated by is to use permanent magnets. Such as [11], they select
coils. When there is a crack in the structure, the eddy strength permanent magnets as the adhesion mechanism,
current will be altered. At the same time, the impedance of which can hold the robot firmly on the walls. Another
the coil will also be affected. Monitoring the change in advantage of these methods is robots do not need extra
impedance in the coils can tell the condition of the facility. power for the adhesion mechanism. In some
This method is sensitive to surface detects and can be used circumstances, variable adhesion is required and the speed
to inspect multilayer structures. Nonetheless, it is very of switching required is high, the electromagnetic adhesion
susceptible to magnetic permeability changes and cannot mechanism will become specifically useful [12].
detect the defects parallel to the surface [9].
Pneumatic adhesion mechanism is another widely used
Through the sensor technologies discussed above, we technology in vertical structure inspection robots. The
can see that each sensor has its own advantages and attraction force between the robot and the wall is
disadvantages. When choosing the proper sensor for proportional to the pressure difference between the
inspection robots, the application scenario and constraints pressure chamber or suction cups and the atmosphere.
need be taken into consideration. Unlike magnetic adhesion methods which can only work
on the ferromagnetic surfaces, pneumatic adhesion front or back module. Compared with other types of
mechanism is suitable for a wider range of materials. inspection robots, it has an advantage in curved pips. Snake
Using suction cups is a very popular method. In [13], three type [24] consists of several identical body segments with
suction cups, a supporting plate, a vacuum pump, and joints, which allows it to generate a wide range of different
some accessories were used to compose the suction motions. The screw type [25] moves forward through the
module. The ICM Rover equips vacuum adhesion rotary motion, achieving a good performance in vertical
mechanism as fall protector, while performing inspection pipelines.
tasks [14].
B. Unmanned Underwater Vehicles
There is also a variety of works adopting biomimetic Oil and gas companies have thousands of kilometres of
adhesion methods to realise the excellent climbing robots. pipelines and other assets in the sea to produce and
In [15], researchers developed a robot, which consists of transport their products. These undersea structures can
four legs with gripping devices made of 12 fishing hooks. easily develop corrosion and cracks since these structures
This robot can imitate the movement of rock-climbers and are often made from metal materials. In order to prevent
the way cats hold surfaces when they climb on vertical financial and environmental disasters caused by leaking out
direction. A gecko-inspired adhesives method was of the product, these facilities need to be inspected
proposed in [16]. Being inspired by the gecko toes, it used frequently. ROVs and AUVs are effective and affordable
a rigid tile supported by a compliant material and loaded platforms to perform underwater inspection tasks. The
by an inextensible tendon. This mechanism allowed the inspection ROVs are operated by the surface operator. They
climbing part to make a full contact with the surfaces. are alternative vehicles to human workers in conditions that
2) Onshore Pipeline Inspection are too deep and too dangerous for human beings. What is
For onshore pipeline inspection robots, the style of more, with the help of ROVs, inspection tasks can be
locomotion is a vital part, which can reflect the whole performed 24 hours and 7 days. Due to these advantages,
performance of the robots [17]. According to the the oil and gas technologies have developed ROV
difference of driving source and control ability of technologies from the 1980s [26]. After that, a large amount
movement mechanism, robots can be sorted into pig, of advanced ROVs with a group of sensors were used to
wheeled, tracks, legged, inchworm, snake and screw type carry out inspection tasks which vary in size and weight
[18]. Fig.3 gives some examples about these kinds of [27][28][29][30][31]. These ROVs consist of a vehicle
robots. body, control cabin, umbilical, video cameras, handling
system, launch system and power supplies. The main
features of these ROVs are summarized in TABLE I.
TABLE I. DIFFERENT ROVS COMPARATION
Max Forward Size
Weight
ROVs Depth Speed (mm)
(kg)
(meters) (knots)
VideoRay
305 38.5 4.2 375, 290, 220
PRO 4
Novaray
model 305 25.5 6 1020,997,229
2000
FALCON 300 60 3 1000,500,600
CONSTR 3220, 1700,
UCTOR 3048 4500 3.1
2165
220 HP
Mojave 300 85 3.5 1000,600,500

For the deep-sea inspection tasks, the major cost of

Figure 3. (a) The pig type [19], (b)The wheel type [20], (c) The track using an ROV is caused by the human operator. If these
type [21], (d) The legged type [22], (e) The inchworm type [23], (f) The tasks can be performed autonomously, this cost can be
snake type [24], (g) The screw type [25]. reduced. Based on the previous technologies and economic
constraints, using the AUVs may be a replacement for the
The pig type [19] itself is a simply device which collects
ROVs for inspection purpose and currently it has gained
the data of the pipeline. The pig carries out the inspection
great research interest. Compared with ROVs, AUVs
tasks along the flow of oil or gas. What is more, the pig has
demand more accurate and efficient sensors, guidance
no driving mechanism and is driven through the pipeline by
systems and algorithms for the execution of this kind of
oil or gas flow. The wheel type [20] uses wheels to touch
mission. The LIRMM [32] presented a Deformable Virtual
the pipe wall. It can easily adapt to various pipelines with
Zones (DVZ) method that is a sensor-based control
springs. The track type [21] is often treated as the
approach. It builds the model of a virtual zone around the
alternative to the wheeled robot. Wheels are bounded with
robot. Once the obstacles are detected by proximity
the belt, which can enlarge the surface contact area and
sensors, there will be deformation and the control signal
reduce chances of losing pipe wall contact. The legged type
will be calculated by minimising the deformation in the
[22] uses legs to contact with the pipe wall. This type of
DVZ. The FlatFish project [33] utilized the asset layout-
robots can produce highly sophisticated motions and is
based navigation methods. Since almost all of the
suitable for pipes with obstacles. The inchworm type [23]
underwater assets are connected together, they can be
uses the traction generated by the large force applied to the
inspected by following pipelines and tie-backs to reach the
different part of facilities. The sonar and camera are used to the UAV is aligned with the pipeline. Finally, automated
target and inspect the facilities. In [34], image-processing navigation along the pipeline is realised.
technologies are adopted to AUV’s navigation. At first, it
detects the pipeline corners in the captured image. After IV. CHALLENGES AND RESEARCH DIRECTIONS
that, the obstacles are identified by Hough Transform (HT) From the survey above, it can be concluded that robots
and velocity and angle values can be calculated according play an important role for inspection purpose in oil and gas
to this information. Finally, the AUV can move along the industries. In these robots, UAV systems can be efficient
pipeline by itself and inspect these structures alternatives for achieving the same goals to UGVs. AUVs
autonomously. outperform ROVs in terms of efficiency and cost in
C. Unmanned Aerial Vehicles underwater inspection tasks. The autonomous inspection
UAVs and AUVs do not need human interaction, which
UAVs equipped with relative sensors can work as an can further reduce the operational cost and time-
excellent alternative to traditional inspection techniques. It consuming. Meanwhile, the autonomous inspection
can not only save time but also lower the cost. The North systems have become the main development direction of
Sea E&P company conducted a survey and showed that energy industries [39]. For autonomous inspection robots,
using UAVs to inspect assets can be twenty times faster the capabilities of autonomous localisation, navigation and
and half the cost of traditional inspection methods [35]. inspection play crucial roles. UAVs and AUVs are both 6
This kind of method has gained great interest in inspecting Degree of freedoms (DoFs) robots, and they both adopt
assets in oil and gas industry. Until now, all of the visual inspection technologies. The same development
commercial inspection UAVs are manually controlled. frame can be implied to both of them. What is more, some
During the inspection procedure, the UAV will be hardware and software can work both for UAVs and AUVs,
controlled by one experienced pilot, while the live video such as px4 open source autopilot [40]. Approaches
for inspection purpose is monitored by another discussed here are suitable for both UAVs and AUVs.
experienced inspection engineer. Famous oil and gas According to the survey above, the autonomous inspection
companies in the world such as BP, Shell, Apache, BG framework can be concluded as Fig.4. It consists of Visual
Group and Statoil have cooperated with Cyberhawk, Simultaneous Localisation and Mapping (VSLAM), the
which is the world leader in drone inspection to inspect control part, path planning and defect detection.
their facilities. Intel Falcon 8 plus used by Cyberhawk has
shown a reliable and efficient performance [36]. The Intel
Falcon 8 plus has a patented V-shaped design with eight
rotors, which makes the UAV more stable and ensures
unobstructed data capture procedure. By carrying three
redundant inertial measurement units (IMUs) with
efficient data fusion technology, the flying system can
perform reliable responsiveness and stability during
flights. The inspection module consists of an RGB camera
and a thermal camera, which helps UAV navigate while
capturing detailed data for orthography and 3D
reconstruction that can be used for inspecting the assets
and further analysation [37]. Another notable inspection
UAV is ELIOS, the first collision-tolerant drone Figure 4. Framework of autonomous inspection
developed by Flyability. This UAV is surrounded by a
carbon fibre shell, which can protect the body from the A. Autonomous Localisation and Mapping
collision. When collision to the obstacles, the UAV will
bounce off and roll along the surface to find the path. What Autonomous localisation and environment mapping
is more, it can finish the inspection task in any light are the basic requirements for navigation. Robots need to
conditions with the help of the HD camera, thermal camera know where it is before planning the path. However, in the
and an LED lighting system. assets inspection scenario, the electromagnetically guided
localisation systems are constrained. As UAVs and AUVs
For UAV inspection, one UAV needs two experienced both utilise vision information to inspect the structure.
human engineers. If the inspection tasks can be performed Vision-based localisation and navigation systems have
autonomously, the inspection procedure can not only be become popular these days. They are based on the images
speeded up but also improve cost efficiency by reducing captured by the onboard camera, and VSLAM algorithms
engineering labor fees and minimising the accidents due to can locate robots’ position, estimate its state and
human operate errors. The Petroleum Institute of Abu simultaneously build the map of the surrounding
Dhabi has developed a UAV autonomous tracking and environment. These features make it suitable for visual
navigation controller for inspecting straight oil and gas guide systems. Some related works have proved the reality
pipelines [38]. Their autonomous procedure contains four of these methods. Graph-based SLAM has been deployed
parts: Firstly, the gradient of Gaussian is used to extract in UAVs and AUVs to realise the robot localisation [41].
the edge of the object in image sequences. Secondly, the There is no doubt that the previously introduced methods
HT is implied to identify the pipeline. After that, a PID have made a great improvement in intelligent and
controller is designed to achieve angular and lateral autonomous 6-DoF robots. However, these approaches are
correction, which ensures the position and orientation of still in the start-up stage. Unlike UGVs, UAVs and AUVs
cannot stop operating while it in the uncertain state. It
means that VSLAM should be able to locate the position V. CONCLUSION
and respond to the dynamic environment quickly. For the In this paper, four kinds of inspection robots in oil and
inspection tasks, the robots should be close to the facilities, gas industries have been surveyed. Among these robots,
which will cause the scenery changes very fast. In other autonomous UAVs are efficient alternatives to UGVs. At
words, perceiving the dynamic environment is a the same time, AUVs can be more efficient than ROVs.
significant problem that needs to be solved. The tolerance The autonomous UAVs and AUVs will be the research hot
of image blur, object occlusions and illumination changes pot in oil and gas industries. What is more, with the
should also be taken into consideration. development of VSLAM, intelligent control strategies and
B. Intelligent Control Strategies and Path Planning path planning methods, the autonomous localisation and
Another challenge is to integrate the VSLAM with navigation performance of UAVs and AUVs still can be
intelligent control strategies and path planning to realise further improved. The autonomous inspection will also
high-level reliable and flexible navigation. Intelligent come true with the help of advanced computer vision
navigation methods can make the inspection progress algorithms, especially with the deep learning method. The
more efficient and safer. This survey also indicates that authors believe that the maturity level of autonomous
visual localisation systems are not fully integrated with the inspection UAVs and AUVs will be developed,
navigation systems. The previously presented works can considering the robustness and reliability.
only work under specific assumptions, such as the ACKNOWLEDGMENT
pipelines are straight or some parameters are known
[34][38]. In other words, their works can just be a kind of This work is supported in part by the UK Oil and Gas
simulation for the autonomous efficient inspection. It is Technology Centre (OGTC) under the LOCUST research
clear that UAVs and AUVs are more efficient than other project (2019-2021, Grant No.: AI-P-028). Mr Leijian Yu
robots which heavily relied on their agile maneuvering is funded by the China Scholar Council and the
capabilities. Therefore, it is essential to pay more attention International Fees Only Studentship from the University of
to it and make intelligent control strategies and path Strathclyde (2018-2021). The authors would also like to
planning a further step. thank the OGTC robotics team at the University of
Strathclyde for their kindly support.
C. Inspection Technologies
It can be clearly seen from the review that one REFERENCES
experienced engineer is needed to monitor the image [1] "Overview | Energy economics | Home", BP global, 2019.
sequences captured by the UAVs and AUVs. What is more, [Online].Available:https://www.bp.com/en/global/corporat
e/energy-economics/energy- outlook/ introduction/overvie
to realise a fully autonomous inspection system,
w.html.
autonomous detecting corrosions and cracks is essential.
[2] "Gulf oil leak: Biggest ever, but how bad?", BBC News,
With the development of image recognition, especially 2019. [Online]. Available: https: //www.bbc.co.uk/ news/sc
with the deep learning algorithms, it is possible to inspect ience-environment-10851837.
the assets in oil and gas industries autonomously. Deep
[3] "Blast halts Venezuela refinery", BBC News, 2019.
learning methods, especially convolutional neural [Online]. Available: https://www.bbc.co.uk/news/world-
networks, which has made many breakthroughs in latin-america-19378657.
computer vision tasks [42][43][44]. It can extract features [4] "Inspection Robotics in Oil & Gas Industry Market |
autonomously for all kinds of objects, which is more Growth, Trends, and Forecast (2018 - 2023)", Mordorintell-
accurate and robust than traditional man designed features, igence.com, 2019. [Online]. Available: https://www.mordor
such as SIFT [45] and HOG [46]. It has led to great intelligence.com/industry-reports/inspection-robotics-in-oi
improvements in many vision inspection tasks. Beijing l-and-gas-industry-market.
University of Science and Technology has adopted the [5] R. ltd, "Inspection Robotics in Oil & Gas Industry: Global
deep learning methods to detect steel surface, which is Market 2017-2025 by Robot Type, Sub-system and
vital to the steel product. With the help of deep learning Region", Researchandmarkets.com, 2019. [Online]. Availa-
method, the detection precision of hot rolled plates has ble: https://www.researchandmarkets.com/reports/4331666
been improved around 16% than traditional methods [47]. /inspection-robotics-in-oil-and-gas-industry.
In [48], a deep learning algorithm was used to detect [6] S. Hajjaj and I. Khalid, "Design and Development of an
breaches, dents, burrs and abrasions on the sealing surface Inspection Robot for Oil and Gas Applications",
of a container. The accuracy and efficiency of this method International Journal of Engineering & Technology, vol. 7,
no. 435, p. 5, 2018.
has been proved by the real-world test in the filling line.
The autonomous detection of concrete structure cracks can [7] P. Shull, Nondestructive evaluation： theory, techniques,
and applications. CRC press, 2016.
also be realised by deep learning. Dung C. V. developed a
[8] Y. Shi, C. Zhang, R. Li, M. Cai and G. Jia, "Theory and
deep learning based autonomous concrete crack detection Application of Magnetic Flux Leakage Pipeline Detection",
algorithm, which can reach approximately 90% of the Sensors, vol. 15, no. 12, pp. 31036-31055, 2015.
maximum Average precision (AP) and F-Score (F1) on [9] M S Safizadeh, M. Hasanian, "Gas pipeline corrosion
validation and test datasets [49]. With so many successful mapping using pulsed eddy current technique", International
works in inspection tasks with deep learning, it is possible Journal of Advanced Design and Manufacturing
to develop autonomous corrosion and cracks detection Technology, vol. 5, no. 1, pp. 11-18, 2011.
approaches in the future. As a result, the inspection quality, [10] M. F. Silva, J. A. T. Machado, and J. K. Tar, “A Survey of
Technologies for Climbing Robots Adhesion to Surfaces,”
speed, as well as the overall inspection process will be in 2008 IEEE International Conference on Computational
improved. Cybernetics, 2008, pp. 127–132.
[11] "Scorpion2 and Swift Remote-Access Tank Shell Inspection [31] "Forum Energy Technologies", F-e-t.com, 2019. [Online].
Solution", Eddyfi, 2019. [Online]. Available: https://www.s Available: https://www.f-e-t.com.
ilverwingndt.com/scorpion2/. [32] P. K. Paim, B. Jouvencel, and L. Lapierre, “A Reactive
[12] M. F. Silva, R. S. Barbosa, and A. L. C. Oliveira, “Climbing Control Approach for Pipeline Inspection with an AUV,” in
Robot for Ferromagnetic Surfaces with Dynamic Proceedings of OCEANS 2005 MTS/IEEE, pp. 1–6.
Adjustment of the Adhesion System,” J. Robot., vol. 2012, [33] J. Albiez et al., “Repeated close-distance visual inspections
pp. 1–16, Mar. 2012. with an AUV,” in OCEANS 2016 MTS/IEEE Monterey,
[13] Y. Guan et al., “A Modular Biped Wall-Climbing Robot 2016, pp. 1–8.
With High Mobility and Manipulating Function,” [34] C. Wang and L. Cui, “The Implementation of Automatic
IEEE/ASME Trans. Mechatronics, vol. 18, no. 6, pp. 1787– Inspection Algorithm for Underwater Vehicles Based on
1798, Dec. 2013. Hough Transform,” in Proceedings of the 2018 7th
[14] "ICM Rover – Advanced Vacuum Adhesion Inspection International Conference on Sustainable Energy and
Crawler", Robotics.innotecuk.com, 2019. [Online]. Availa- Environment Engineering (ICSEEE 2018), 2019.
ble: http://robotics.innotecuk.com/icm-rover-advanced-vac [35] "Drones halve the cost of inspection, Cyberhawk says",
Offshore Energy Today, 2019. [Online]. Available:
uum-adhesion-inspection-crawler/. https://www.offshoreenergytoday.com/drone-inspections-
[15] A. Sintov, T. Avramovich, and A. Shapiro, “Design and halve-the-cost-of-inspection-cyberhawk/.
motion planning of an autonomous climbing robot with [36] "Intel and Cyberhawk Inspect Gas Terminal through Lens
claws,” Rob. Auton. Syst., vol. 59, no. 11, pp. 1008–1019, of Commercial Drone Technology - Cyberhawk",
Nov. 2011. Cyberhawk, 2019. [Online]. Available: https:// thecyberhaw
[16] E. W. Hawkes, E. V. Eason, A. T. Asbeck, and M. R. k.com/intel-cyberhawk-inspect-gas-terminal-lens-commerc
Cutkosky, “The Gecko’s Toe: Scaling Directional
Adhesives for Climbing Applications,” IEEE/ASME Trans. ial-drone-technology/.
Mechatronics, vol. 18, no. 2, pp. 518–526, Apr. 2013. [37] I. System, "Intel® Falcon™ 8+ System", Intel, 2019.
[17] Wang, D.J., Li, R.P. and HUANG, G.M., “Progresses in [Online]. Available: https://www.intel.co.uk/content/www/
Study of Pipeline Robot,” in 2012 Proceedings of uk/en/products/drones/falcon-8.html.
International Conference on Modelling, Identification and [38] A. Shukla, H. Xiaoqian, and H. Karki, “Autonomous
Control, Sep, 2012. tracking and navigation controller for an unmanned aerial
[18] J. M. Mirats Tur and W. Garthwaite, “Robotic devices for vehicle based on visual data for inspection of oil and gas
water main in-pipe inspection: A survey,” J. F. Robot., vol. pipelines,” in 2016 16th International Conference on
27, no. 4, pp. 491–508, Jun. 2010. Control, Automation and Systems (ICCAS), 2016, pp. 194–
[19] A. A. Mazreah, F. B. I. Alnaimi, and K. S. M. Sahari, “Novel 200.
design for PIG to eliminate the effect of hydraulic transients [39] V. Sudevan, A. Shukla, and H. Karki, “Inspection of
in oil and gas pipelines,” J. Pet. Sci. Eng., vol. 156, pp. 250– Vertical Structures in Oil and Gas Industry: A Review of
257, Jul. 2017. Current Scenario and Future Trends,” pp. 65–68, Jul. 2018.
[20] A. Kakogawa and S. Ma, “Design of a multilink-articulated [40] "Open Source for Drones - PX4 Open Source Autopilot",
wheeled pipeline inspection robot using only passive elastic PX4 Open Source Autopilot, 2019. [Online]. Available:
joints,” Adv. Robot., vol. 32, no. 1, pp. 37–50, Jan. 2018. https://px4.io/.
[21] Y.-S. Kwon and B.-J. Yi, “Design and Motion Planning of [41] A. Annaiyan, M. A. Olivares-Mendez, and H. Voos, “Real-
a Two-Module Collaborative Indoor Pipeline Inspection time graph-based SLAM in unknown environments using a
Robot,” IEEE Trans. Robot., vol. 28, no. 3, pp. 681–696, small UAV,” in 2017 International Conference on
Jun. 2012. Unmanned Aircraft Systems (ICUAS), 2017, pp. 1118–
[22] A. A. Gargade and S. S. Ohol, “Development of Actively 1123.
Steerable In-pipe Inspection Robot for Various Sizes,” in [42] F. Gao, F. Ma, J. Wang, J. Sun, E. Yang, and H. Zhou,
Proceedings of the Advances in Robotics on - AIR ’17, “Visual Saliency Modeling for River Detection in High-
2017, pp. 1–5. Resolution SAR Imagery,” IEEE Access, vol. 6, pp. 1000–
[23] C. Nițu, B. Grămescu, A. S. Hashim, and M. Avram, 1014, 2018.
“Inchworm Locomotion of an External Pipe Inspection and [43] F. Gao, T. Huang, J. Sun, J. Wang, A. Hussain, and E. Yang,
Monitoring Robot,” Springer, Cham, 2019, pp. 464–470. “A New Algorithm for SAR Image Target Recognition
[24] F. Trebuňa, I. Virgala, M. Pástor, T. Lipták, and Ľ. Miková, Based on an Improved Deep Convolutional Neural
“An inspection of pipe by snake robot,” Int. J. Adv. Robot. Network,” Cognit. Comput., pp. 1–16, Jun. 2018.
Syst., vol. 13, no. 5, Sep. 2016. [44] Z. Yue et al., “A Novel Semi-Supervised Convolutional
[25] H. Tourajizadeh, M. Rezaei, and A. H. Sedigh, “Optimal Neural Network Method for Synthetic Aperture Radar
Control of Screw In-pipe Inspection Robot with Image Recognition,” Cognit. Comput., pp. 1–12, Mar. 2019.
Controllable Pitch Rate,” J. Intell. Robot. Syst., vol. 90, no. [45] D. G. Lowe, “Distinctive Image Features from Scale-
3–4, pp. 269–286, Jun. 2018. Invariant Keypoints,” Int. J. Comput. Vis., vol. 60, no. 2, pp.
[26] Shukla A, Karki H. Application of robotics in offshore oil 91–110, Nov. 2004.
and gas industry—A review Part II[J]. Robotics and [46] N. Dalal and B. Triggs, “Histograms of Oriented Gradients
Autonomous Systems, 2016, 75: 508-524. for Human Detection,” in 2005 IEEE Computer Society
[27] "VideoRay ROVs - Underwater Remotely Operated Conference on Computer Vision and Pattern Recognition
Vehicles", Videoray.com, 2019. [Online]. Available: (CVPR’05), vol. 1, pp. 886–893.
http://www. Videoray.com/. [47] H. Di, X. Ke, Z. Peng, and Z. Dongdong, “Surface defect
[28] "Nova Ray® Subsea Systems ROV (Remotely Operated classification of steels with a new semi-supervised learning
Vehicle) The Current Experts™", Novaray.com, 2019. method,” Opt. Lasers Eng., vol. 117, pp. 40–48, Jun. 2019.
[Online]. Available: http://www.novaray.com/. [48] Y. Li et al., “Research on a Surface Defect Detection
[29] "Electric Underwater Robotics (ROVs) | Saab Seaeye", Algorithm Based on MobileNet-SSD,” Appl. Sci., vol. 8,
Saabseaeye.com, 2019. [Online]. Available: https://www.sa no. 9, p. 1678, Sep. 2018.
abseaeye.com. [49] C. V. Dung and L. D. Anh, “Autonomous concrete crack
detection using deep fully convolutional neural network,”
[30] "Underwater Drones (ROV, USV) for sale in California | Autom. Constr., vol. 99, pp. 52–58, Mar. 2019.
Deep Ocean Engineering", Deepocean.com, 2019. [Online].
Available: https://www.deepocean.com/.