End to end navigation stack for nuclear power plant inspection with mobile robot

摘要

This paper describes a novel approach for nuclear facility inspection with novel automated 3D mapping system as an open source end to end navigation stack available at https://github.com/JanuszBedkowski/msas_enrich_2023 . Incidents such as Fukushima, Majak or Chernobyl as well as the decommissioning and dismantling of old nuclear facilities ( e.g. Sellafield, Asse or Murmansk) are showing great importance of the robotic technology. Rapid inspection requires reliable, accurate, precise and repeatable simultaneous localization and mapping. Proposed SLAM approach uses only non repetitive scanning pattern Lidar (Livox Mid360) and integrated inertial measurement unit. The novelty is based on feature less single core SLAM implementation. It fuses Normal Distributions Transform and motion model for simultaneous map building and current pose estimation. Motion model bounds an optimization result, thus it is stable and reliable. It requires less than 10 ms for pose update, trajectory tracking and emergency behavior. This method is a candidate for real time application since a calculation time is bounded and it uses only one core of Intel Celeron CPU G1840 2.8 GHz. It was tested both (i) during EnRicH 2023 https://enrich.european-robotics.eu/ — the European robotics hackathon, (ii) in laboratory conditions. This open source project provides also software of base station, thus it is first end to end solution available in literature.