Research Of Positioning And Navigation Of The Tokamak Transfer Cask

Positioning and navigation technology is one of the key researchesin robotics. Reliable positioning and navigation technology is critical formobile robots to perform well in the fields of industrial applications,intelligent transportation, indoor cleaning, search and rescue, and evenplanet exploration. Under the polluted area, such as nuclear facilities,mobile robots with self-positioning and navigation abilities are extremelyhelpful.The Transfer Cask System(TCS) is a critical element of theInternational Thermonuclear Experimental Reactor, it is devoted totransportation of components between the Tokamak building and Hot Cell.The driverless transfer cask is supposed to be automatically guidedamong confined spaces with highly reliability. In this thesis, anexperimental platform of the transfer cask is designed, then how to realizereliable and efficient positioning and navigation under hazardousenvironments is especially studied. Concrete researches are performed inthe following aspects:1. As the practical demands of the component transfer cask, we analysis and compare the existing navigation techniques, and anexperimental platform of the transfer cask based on laser sensor isdesigned.2. We research on navigation methods based on artificial landmarkswith high remission value，which can be detected by a laser scanner.Firstly we analysis the usage of artificial landmarks with a thoroughunderstanding of the working principles of laser scanner. Then, wepropose an optimized technology of landmark distribution, validate theprocess of creating sparse landmark maps and high-precision positioningand navigation on a the designed experimental platform.3. From the aspect of using artificial landmarks, given the currentcomplexity and time-consuming mapping procedure, we introduce arelatively efficient method of combining Simultaneous Localization AndMapping(SLAM) to automatically generate the landmarks maps using thereflectance.4. From the aspect of SLAM, we study on the important dataassociation problem in SLAM, analyze the existing closed loop detectionmethods. Artificial landmarks are proposed for reliable loop closuredetection. With SLAM techniques the number of artificial landmarkscould be substantially reduced, since most of the time the pose the vehicleis known. In the meanwhile, the proposed method reduces the difficultyof data association, and provides better loop closure result. Finally, the methods are applied to unmanned experimental tokamaktransfer cask platform. Friendly human-machine interfaces are developed.Large indoor environments SLAM tests are performed and experimentalresults have demonstrated that the proposed approach is capable ofproviding a high positioning accuracy and efficiency of generatingconsistent maps.