Development and testing of an exploration strategy for a cooperative network of rover and UAV using the example of mapping gas concentrations

Abstract

The present thesis develops and demonstrates an exploration strategy for the reconnaissance of gas plumes with a cooperative network of airborne and ground-based unmanned vehicles by remote sensing. The cooperation of airborne and ground-based vehicles poses a clear distinction from other work in this field. For the measurement of the gas plume, the method of Tunable Diode Laser Absorption Spectroscopy (TDLAS) is applied. The method of TDLAS requires a multitude of measurements from different angles. Using the airborne vehicle as a reflector for the TDLAS beam provides more precise information on the measurement. To optimise the positioning of the vehicles and explore the gas plume, an engineering approach is pursued. A heuristic containing an exploration, an exploitation, and a moving costs term is developed. Furthermore, a system capable of implementing the strategy is arranged and presented. In a multi-step approach, the exploration concept is implemented and verified. The thesis presents the structure of the implementation of the exploration strategy in a Python simulation. With the simulation, the basic operating principle of the exploration strategy has been verified. Supplementary, the parameters used in the heuristic are examined to improve the performance of the exploration strategy. Besides, different variations of the strategy are analysed and discussed. Additionally, the implementation is transferred to a physics simulation with Robot Operating System (ROS) and Gazebo. In the simulation, the strategy is combined with the constraints of the vehicles. The simulation is performed with software in the loop simulations of the actual hardware to verify the compatibility. In a proof-of-concept demonstration, the implementation of the exploration strategy in the real hardware system (i.e. Unmanned Aerial Vehicle (UAV) and Unmanned Ground Vehicle (UGV)) with software-in-the-loop simulation of the TDLAS sensor is shown and analysed.