Autonomous Navigation of UAVs in the Indoor Environment for Search and Rescue Missions
This presentation talks about a system for autonomous navigation and target recognition for indoor search and rescue missions using small unmanned aerial vehicles (UAVs). The use of lightweight multicopters makes them ideal for maneuvering through tight spaces and locating victims in shorter time. Simultaneous Localization and Mapping (SLAM) techniques and Collision Avoidance Systems (CAS) are used to navigate the vehicle in the GPS-denied environments. SLAM, CAS, and target recognition software can help rescuers locate victims for disaster relief. Using a LIDAR and camera, it is possible to create a map of an indoor environment and determining and keeping track of the UAV's in the constructed map. The victims can be identified using onboard image processing. An RPLidar is used in conjunction with HectorSLAM algorithm localization and mapping. A Mobius Actioncam is used for the victim identification. The identification software runs on an NVidia Jetson TX1 microcomputer. The Jetson TX1 communicates with the onboard Pixhawk flight controller, while also transmitting data to a ground station using Xbee radio modules. Neural networks are used for the identification of victims as well as for collision avoidance with the wall so that the UAVs navigate the indoor environment safely. Simulation and test results will be presented. Work is underway to test the overall system in flight for realistic search and rescue missions in the indoor environment.