Page 40 - Read Online

Page 22 of 26 Blewitt et al. Soft Sci 2024;4:13 https://dx.doi.org/10.20517/ss.2023.49

Improvements and advancements in worm robot mapping are required to enhance their applicability to real
[24]
pipe inspection although many robots have been demonstrated with cameras at their heads . Image
[80]
processing could be used to create a global SLAM algorithm using landmark recognition , though as pipe
environments are quite uniform in appearance, local information from IMU sensors and ultrasonic sensors
may be needed to improve positional estimations . Although, in some respects, pipe environments make
[81]
visual SLAM difficult, the uniform geometry of the pipes can aid odometry. Zhang et al. showed that as
feature detection is performed in a cylindrical environment of known diameter, it is relatively simple to
[82]
calculate the distance moved between frames . Worm-like robots tend to explore smaller pipelines
(< 70 mm), although not much research has been conducted into small-diameter pipe SLAM due to the size
and rigidity of sensors, making them difficult to integrate. Lim et al. developed a robotic system where a
[83]
sponge was propelled through 15 mm pipes using hydraulics as the driving force . The Sponge pulled a
CMOS camera and a sensor unit behind it which were used to build a map of the pipes. An accelerometer
and single-axis gyroscope were used to measure the roll of the camera with image processing employed to
detect the direction of forward movement. This data was all used in combination to create a map of the
pipelines. However, Lim et al. found that the camera algorithm was unreliable as pipe conditions
changed . SLAM in worm-like pipe inspection robotics remains an open challenge which is required to
[83]
broaden the applicability of worm-like robot inspection.

CONCLUSION AND OUTLOOK
Recently, the application of worm-like robots for pipe inspection has seen significant development.
Currently, there exist many worm-like robots capable of traversing small-diameter pipes both vertically and
horizontally and around tight bends. Many have also presented their ability to carry a small load such as a
camera, raising the possibility of using them for inspection. However, the robots presented in this work
have not been demonstrated in realistic pipe set-ups and have not been shown to exhibit specific capabilities
that may be required to carry out inspection tasks. To be specific, the pipe set-ups that the robots have been
tested upon are, in most cases, plastic pipes, whereas the pipes used in the industry tend to be made from a
metal such as steel. Steel may pose a problem for worm robots as they present a more frictional surface, and
worm robots rely on pushing their body along the pipe environment. Moreover, autonomous navigation
and mapping of complex pipework has also not been demonstrated, likely due to two areas of difficulty. The
first of these is the turning in a soft robot. The controlled curvature of soft actuators and robots is a complex
control problem, and the pipe environment provides both guidance (in the case of swept bends) and
obstruction (in the case of junctions such as T-junctions). Hence, a control system that can effectively
predict the robot state and sense the robot’s environment is required. Worm robots have demonstrated
their ability to turn, although a control method that could be considered robust enough for real pipe
environments remains an open challenge. The other area of difficulty comes with mapping the pipe
environment. Though this has been demonstrated in several non-worm-like pipe robots, there are few
examples in worm robots. This is likely as estimations of robot shape and movement are required to create
effective mapping algorithms and as worm robots are generally made from soft actuators this presents a
significant difficulty. On top of these difficulties, there are also the problems associated with attaching
sensors to small compliant robots. For the pipe inspection capabilities of worm-like robots to be realised,
demonstrations of worm-like robots with more sophisticated sensing and inspection abilities are required.

DECLARATIONS
Authors’ contributions
Made substantial contributions to conception and design of study, found and analysed the relevant
literature for interpretation: Blewitt G