Development and Demonstration of a Software-Defined Real-Time Control Interface for Micro Air Vehicles (Masterthesis)


Alexander Söhn

19.10.2015, 9:00, room 3945


The aim of this thesis was to develop a prototype of a flying platform, in which it will be possible to directly intercept and manipulate the copter’s main control loop. In order to avoid potential safety hazards, the possibility to partially and transiently enable and disable certain control actions shall be added to this platform. For example, a camera-based obstacle detection could be integrated. If the input from the remote control (RC) would lead to crashing into another drone, the RC input is disabled and a computer system takes over the control of the copter. In this thesis, the focus lies on the development of an interface that allows for manipulation of the control actions by software. In order to enable a computer that is running autonomous algorithms to control the copter, a specific interface is needed. Since autonomous algorithms include critical control tasks, this interface shall introduce as little latency as possible. There exist many suitable commercial of-the-shelf (COTS) quadrocopters, but these are lacking an interface that can be employed to connect to an embedded computer. In this thesis, a COTS quadrocopter shall be selected and fitted with an interface to an embedded computer. Firstly, requirements were derived from the desired use case. Secondly, COTS copters were evaluated and compared with respect to these requirements. After acquiring and examining the copter, various possibilities for interfacing the copter to a Zynq device were developed and discussed. Once the most suitable approach was implemented, the results were analysed. The outcome of this work is that a Zynq device is able to control the MAV. The timing requirements for the interface have been met. No additional latency, except for the delay that is inherent in the system, was introduced by the interface. Furthermore, the interface was analysed for race conditions, because these are a common problem in complex systems. It can be proven, that the interface does not introduce any race conditions. To conclude, a MAV that will serve as a testing platform for autonomous algorithms has been developed. The stability problems will be solved quickly, because the cause is already known and the timing requirements have been met. In the future, this thesis project will be ported to a bigger quadrocopter to test autonomous code on that platform.