Stratégie de transition entre localisation autonome et localisation intérieure de haute précision pour un robot mobile manipulateur

Abstract

This thesis studies the localization transition of a mobile manipulator between autonomous localization and high-precision localization provided, in an instrumented area, by an indoor localization system designed for metrology. The objective is to preserve autonomous navigation outside this area and to use the external measurement when the robot is inside it, without disrupting the reference frames required for navigation and manipulation. The work is conducted on the RanGen, a four-wheel-steering mobile base carrying a Kinova arm equipped to use this external measurement. It addresses the digital twin, the integration of this measurement with the arm, frame calibration, and the transition between global localization sources. The working hypothesis is that a transition architecture, based on consistent ROS 2 interfaces and geometric calibrations, can connect autonomous navigation, high-precision localization, and local manipulation. The methods combine digital modeling, ROS 2 integration, rigid-body calibrations, and comparative tests on the real robot. The results show functional continuity between simulation and the real robot. In manipulation, a feedback loop correcting the arm pose based on the external measurement reduces the 3D RMSE from about 7.7 mm to 1.7 mm. In navigation, the high-precision localization reduces the median error at the Nav2 target poses from 78.0 mm to 3.5 mm, and the yaw error from 2.17◦ to 0.24◦. The final demonstration validates the sequence linking autonomous navigation, the transition to this high-precision localization, and the execution of an assembly task by the robotic arm. The thesis concludes that this transition is functional in the experimental context studied. Recommended future work concerns the robustness of the external measurement setup mounted on the arm and the definition of a decision criterion for determining when the robot should return to autonomous localization if this measurement is no longer available or no longer sufficiently reliable.

Date	12 Aug 2026
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	David St-Onge (Supervisor) & Lionel Birglen (Co-supervisor)