A Method for Semi-Autonomous Map Generation and Path Finding In UAV-UGV Cooperation for Search and Rescue Mission in Dynamic and Unmapped Environment

Abstract

Catastrophes, accidents and wars can create circumstances where rescue missions need to be conducted without delay. These rescue missions can be as dangerous for rescuers as they are for the ones waiting to be rescued. Landscape may be deformed due to the events, or environment can have dynamic events occurring and therefore prior mapping of the area is not valid. Therefore, there is need for solutions that allow operations to be performed without delay and without endangering rescuers. To tackle this issue this thesis introduces a concept for search and rescue missions where unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV) agents are used in collaboration to map the area and to find a path for UGV to reach the target destination. Concept is possible to implement with fully or partially autonomous agents. Concept allows scaling up and down operations depending on resources. In addition to this high-level concept, this thesis introduces methods for map and path generation for UGV using depth images taken from aerial perspective of the mission area by the UAV. The algorithm for the map generation filters away false obstacles that are present in raw images from above, but do not actually form obstacles for the UGV on the ground. These false obstacles can be for example foliage of trees or powerlines. As in some situations it is important to be able to remain undetected, this mapping and path generation method is developed to support passive sensors. Path generating part of the method is modified version of A* search algorithm. This method allows adjusting of the route optimization according to traversability of the UGV used in the mission.