A Multi-Modal Mapping Unit For Autonomous Exploration And Mapping Of Underground Tunnels

The ability of autonomous exploration and mapping using aerial robots in GPS-denied dark visually-degraded environments is critical in multiple applications including those of inspection, exploration, search and rescue in tunnel environments, monitoring of mines, and mapping underground voids. This paper presents the design of a multi-modal mapping unit tailored to such missions. When combined with our previous work of the receding horizon volumetric exploration path planner, the unit enables robotic autonomy in such environments without any prior knowledge of the environment. The multi- modal mapping unit detailed in this work tightly synchronizes visible light cameras with inertial sensors as well as LEDs that flash only when the cameras' shutters are open. When used in conjunction with a visual-inertial odometry pipeline, the multi-modal mapping unit enables reliable robot navigation in darkness. With the further support of multiple miniature Time-of-Flight 3D depth sensors, dense and accurate maps are derived. The proposed system was evaluated using field experiments involving exploration and mapping of a railroad tunnel in conditions of darkness. Results show that a consistent and dense mapping of such challenging degraded visual environments was achieved.