Improved Reconstruction Of Jet Using Lidar-Vision Fusion

Just like most industrial or scientific installations, future fusion reactors will require frequent maintenance. The expected environmental conditions and the necessity of carrying out many tasks in parallel result in the requirement to replace person-in-the-loop maintenance with robotic maintenance. Many advanced technologies will be required to carry out the automated inspection and maintenance tasks in order to minimize the maintenance shutdown durations. LIDAR is one such promising technology which is only just starting to be applied in Fusion contexts. Though it is presently not radiation tolerant enough to be utilized in future reactor designs such as ITER or DEMO without further development, the low radiation levels in JET have presented an opportunity to evaluate the technology for use in fusion environments. In previous work, we presented initial results using data captured in JET in the form of a coloured 3D-point cloud created by LIDAR-Vision sensor fusion. In this paper, we present further work done to improve the quality of this data. This includes details on the improvement of model quality using ORB-SLAM as well as the use of recorded JET RH Boom kinematics data. We also carry out pointcloud-CAD data comparisons using numerical methods and show and improvement in reconstruction quality using ORB-SLAM and Boom Kinematics. Finally, the results are discussed and the relevance of this technology for future remote maintenance system inspection and navigation tasks is evaluated.