Volker Strobel edited DeclareMathOperator_argmin_arg_min_DeclareMathOperator__1.tex  over 7 years ago

Commit id: c6be511f2853408723803091344e942b9973f8e7

deletions | additions      

       

In a first setting, the commercially available Parrot AR.Drone2.0 was  equipped with an Odroid XU-4 single board computer, a Logitech 525 HD  webcam, a WiFi module, and a USB connection between module. Figure~\ref{fig:comparison} shows the setup. Camera images were processed on  the more powerful  Odroid board processor  and the AR.Drone2.0 resulting $x,y$-estimates were send over a USB data link to the MAV  flight controller controller.  to send and receive data between the two processors.Figure~\ref{fig:comparison} shows the setup.  The Odroid processor has a full operating system (Ubuntu 15.04) and  allows to run for running  arbitrary Linux software. While this setup allowed to  test algorithms on the rather powerful Odroid processor, However,  the additional weight through the modifications of the system,  resulted in unstable flight performance. Therefore, we decided to modify modified  the system to run it execute the localization algorithm  directly on-board on board  of the UAV MAV  and not onan external board such as  the additional  Odroid processor. This To this end, the software had to be ported from the the high-level language  Python to the low-level language C without being able to rely on many  existing software libraries.  Finally, this  removed the need for additional payload and made flight performance very stable. Also, it reduced the need to buy and attach an external  processor and a point of failure. Another advantages is that the  framework can be easily ported to any UAV supported by the paparazzi  software. The disadvantages are major disadvantage is  that the on-board processors of many MAVs have a lower performance than the Odroid processor.Additionally,  it required us to port the software from the high-level language  Python to the low-level language C without being able to rely on many  existing software libraries.  We decided to use the quadcopter \emph{Parrot Bebop Drone}  as prototype for all our tests. This UAV allow for navigating in