A Framework of Using Customized LIDAR to Localize Robot for Nuclear Reactor Inspections

While remote inspection of industrial structures, such as nuclear reactors, using robotic crawlers currently presents significant advantages in terms of safety, accuracy and cost, other challenges emerge due to poor context-awareness and positional accuracy. This results in a lack of visibility for path planning and difficulty in precise localization of NDE (Non-Destructive Evaluation) inspection data. LIDAR (Light Detection and Ranging) are one form of sensors that estimate distances at various angles to map the surrounding environment using optical techniques. Existing commercial LIDARs offer a long range of measurement, allowing mapping of the surroundings. However, such sensors often have centimeter accuracy and a minimum scan range, resulting in a blind area and are generally unsuitable for compact spaces and areas with high density of neighboring objects. This paper presents a framework for using a customized 2D laser scanner, an IMU (Inertial Measurement Unit) and a data fusion approach for localization inside high-density volumes such as nuclear reactors. The laser scanner offers precise measurements with submillimeter accuracy for items located in the short range. The IMU calculates the robot attitude angles, which are critical for inclination angle corrections. The facilities are often made of metallic materials with highly reflective surfaces, which remains problematic for the laser scanner. A mock-up nuclear dome, of realistic material construction, was utilized to benchmark the performance of this framework. The distance and orientation error observed were below 2 mm and 1°, respectively. The framework will be further processed to produce a close-range environment mapping.