"Flying Soli" at MatLab Expo 2018: Radar-based Localization Showcase for Multicopter Solutions

Cooperative Project with MCI, University of Stuttgart and Infineon successfully presented at MatLab Expo in Munich.

Based on a cooperation between the Department of Mechatronics at the Management Center Innsbruck (MCI), the Flight Mechanics and Controls Lab at the University of Stuttgart, the Department of Embedded Systems at UAS Technikum Wien and the company Infineon Technologies, a multicopter provided by MCI was equipped with the 60 GHz radar chip GoogleSoli. The results of this radar-based localization system were tested at the MCI and successfully shown at the MATLAB Expo 2018 in Munich.

Based on this hardware setup a basic concept was developed which is shown in figure 1:


Flying Soli



Figure 1: Flying Soli Concept 



The multicopter software is based on the open-source flightcontroller Cleanflight and is executed on an XMC4500-1024 by Infineon. In order to execute CPU-intensive algorithms a Raspberry Pi 3 Model B+  is placed directly on the multicopter and communicates via serial interfaces with the flightcontroller. The radar chip is mounted on the bottom of the multicopter to detect targets in the area beneath the multicopter. The radar target is a trihedral cone, which reflects the radar signals directly back to the radar chip. The outputs of the radar downmixers are processed by various ADCs and the results are sent via USB/Serial to the Raspberry Pi. The Raspberry Pi calculates the 3D coordinates based on the received radar values, which is shown in figure 2:


Flying Soli



Figure 2: Sequence of Coordinate Calculation



The system architecture for the controller is shown in figure 3. The 3D coordinates from the Raspberry Pi are converted to a body-fixed distance vector and used as an input for an implemented distance control loop.


Project with MCI



Figure 3: Distance Vector Controller System Overview 



The body-fixed distance vector is transformed to a pseudo-geodetic coordinate system in order to compensate for roll and pitch movements. This vector is now compared to a desired reference vector (f.ex. 50cm above the target) and the error is fed into 3 decoupled controllers (x, y and z axis), which output accelerations for each axis. These accelerations are converted into roll and pitch angles and a total thrust to be used as commands for the low-level flight controllers. This ensures that once a radar target is detected the multicopter hovers and follows a specific radar target. The "hover" mode, shown in video 1 (Radar-based Distance Controller - Hover Mode), was presented at the MATLAB Expo 2018 and the "follow" mode is currently under development. In addition, using secondary reflectors as shown in figure 1 to determine and control the heading of the copter is also under development.

The radar-based distance controller and the radar signal processing have been generated based on MATLAB/Simulink. In order to use this showcase at public events a flightbox, shown in video 2 (Flightbox-Video) and figure 4 & 5 (Flightbox), has been developed to ensure that attendees of public events are not harmed during the presentation of this showcase. Hence, the radar target can be moved in x- and y-direction in order to test the "follow" mode in the future.

The results of this radar-based localization system was tested at the MCI and successfully shown at the MATLAB Expo 2018