The NDE industry is under constant pressure to increase inspection speeds, while simultaneously reducing costs to keep up with the ever-expanding demands of providing robust inspection for new infrastructure as well as ongoing inspections for currently operating facilities, and the increasing rise in the need for extensions in the planned life of existing plants. Currently, setting up an automated phased array ultrasonic inspection requires significant manpower, especially on components with complex geometry, this often exposes operators to hazardous environments. This is a particular problem with conventional ultrasonic NDT where operators must regularly exchange probes (an ‘intervention’). Furthermore, inspections are often carried out during planned outages, and the necessary installation time of rigging can represent a significant part of the inspection cost. To alleviate these challenges, several specialised robotic systems have been developed in industry for performing NDE in areas with well-defined geometries. However, these systems are often limited by a high degree of manual intervention, a lack of general-purpose design, and unsophisticated brute-force data acquisition with little to no data interpretation. The development of next generation, automated NDE solutions present considerable improvements to the current state of design such as reduced inspection time, greater separation of data capture and analysis, data localization – data are intrinsically encoded with the position they were captured. These benefits lead to a reduction in plant downtime & operator dosage. The platform presented will achieve these improvements through a set of universal automated deployment tools, implemented through hardware and software advances. By creating a platform consisting of a motorised magnetic base paired with a miniature robotic arm, a very capable and adaptable system is formed. This allows for different sensing modalities with an initial focus on phased array ultrasonics to be delivered accurately and repeatably to the target inspection site. Furthermore, by introducing additional perceptual sensors such as cameras, laser scanners, & a force-torque sensor the system can understand the environment in which it is operating. Through these sensors the user may guide the robot through the plant remotely in a safe and controlled manner. In addition to this these sensors may be used to generate scan paths of critical areas with unknown geometry on the fly as well as adapt the path in a conformable manner.

In this paper, a novel method for crawler positioning is presented utilizing an onboard depth-sensing camera which can operate in semi-structured, self-similar environments, using only measurements of the sample under inspection to navigate. Nondestructive evaluation at manufacture is a vital aspect of assuring the fitness for purpose of high-value marine assets, moreover, it is almost always a regulatory requirement to ensure build quality standards have been met. Traditionally these inspections were deployed manually by a trained operator in a laborious and timeconsuming manner. More recently, robotic crawler-based solutions have become available on the marketplace, however, these solutions are limited in their capabilities and still require significant manual intervention and set-up for each application. Additionally, GPS or prior knowledge of their surroundings which are critical to their operation are often unavailable in an active work environment. An autonomous, self-localizing system would provide significant benefits in these situations, but certain challenges arise from limited situational awareness and poor positional accuracy. The accuracy and robustness of the novel method were assessed and experimentally validated through ground truth readings from a Vicon motion capture system. The localization algorithm's ability to function on different materials and under various lighting conditions was also explored. Using the example of the receipt inspection of steel plate under 240 lux lighting, the system proved capable of positioning the crawler at the desired position within 5.7mm.