A hardware-in-the-loop testing facility for unmanned aerial vehicle sensor suites and control algorithms

In the past decade Unmanned Aerial Vehicles (UAVs) have rapidly grown into a major field of robotics in both industry and academia. Many well established platforms have been developed, and the demand continues to grow. However, the UAVs utilized in industry are predominately remotely piloted aircraft offering very limited levels of autonomy. In contrast, fully autonomous flight has been achieved in research, and the degree of autonomy continues to grow, with research now focusing on advanced tasks such as navigating cluttered terrain and formation flying.The gap between academia and industry is the robustness of control algorithms. Academic research often focuses on proof of concept demonstrations with little or no consideration to real world concerns such as adverse weather or sensor integration.One of the goals of this thesis is to integrate real world issues into the design process. A testing environment was designed and built that allows sensors and control algorithms to be tested against real obstacles and environmental conditions in a controlled, repeatable fashion. The use of this facility is demonstrated in the implementation of a safe landing zone algorithm for a robotic helicopter equipped with a laser scanner. Results from tests conducted in the testing facility are used to analyze results from flights in the field.Controlling the testing environment also provides a baseline to evaluate different control solutions. In the current research paradigm, it is difficult to determine which research questions have been solved because the testing conditions vary from researcher to researcher. A common testing environment eliminates ambiguities and allows solutions to be characterized based on their performance in different terrains and environmental conditions.This thesis explores how flight tests can be conducted in the lab using the actual hardware and control algorithms. The sensor package is attached to a 6 DOF gantry whose motion is governed by the dynamic model of the aircraft. To provide an expansive terrain over which the flight can be conducted, a scaled model of the environment was created.The feasibility of using a scaled environment is demonstrated with a common sensor package and control task: using computer vision to guide an autonomous helicopter. The effects of scaling are investigated, and the approach validated by comparing results in the scaled model to actual flights. Finally, it is demonstrated how the facility can be used to investigate the effect of adverse conditions on control algorithm performance. The overarching philosophy of this work is that incorporating real world concerns into the design process leads to more fully developed and robust solutions.