This project is part of the BioSensing and Instrumentation Lab (BSAIL) at the University of Georgia. Our goal was to develop a low cost drone and imaging system for high throughput phenotyping for agricultural crops. Most high-end drones and sensor systems used in the research field can cost tens of thousands of dollars so we set out to develop a custom built option at a fraction of the price.

Our drone uses a Pixhawk 4 flight controller which is an open hardware project aimed at providing high-quality and low cost autopilot hardware. The flight controller runs the ArduPilot flight stack. Our first sensor system is a low cost multispectral camera using Raspberry Pi’s with RGB and NIR cameras.

Our goal with this documentation is to provide the parts and procedures of building this drone so that it can be applied by others on similar projects. As we continue this research we will expand this documentation as necessary.

Parts

A solid walkthrough of the basic configuration of the flight controller can be found here. Follow the steps in this guide and then return here for subsequent setup.

For the onboard computer (Raspberry Pi 3B+, in our case), we need to set up the telemetry 2 port to handle MAVLink communication. We do this by adjusting several parameters on the board. We can do this using QGroundControl that was installed in the aforementioned guide. Connect to the flight controller, either by USB or telemetry radio and open the Parameter’s tab in QGC.

Find the parameter SERIAL2_PROTOCOL and set it’s value to MAVLink1. Also set SERIAL2_BAUD to 921600, this is the baud rate we will use for the UART communication.

For the Raspberry Pi 3 B+, we will be using a Xenial (16.04) Ubuntu image from Ubiquity Robotics. This image comes with ROS Kinetic pre-installed and so is perfect for our on-board computer for the drone.

The image can be downloaded here We used the “2019-06-19-ubiquity-xenial-lxde” file.

Use an image writing software like Etcher to write this image to your micro SD card for the Pi. A 16GB card minimum is recommended but even larger is preferred if planning to capture large amounts of high resolution images and/or video.

Upon first boot, the Pi will resize it’s file system to fill the SD card, this may take a few moments.

The username is ubuntu with password ubuntu.

The image comes with a Wi-Fi access point which will come in use later but for now we will need to connect to the internet to grab updates and software packages.

First, disconnect form the access point then connect to internet enabled Wi-Fi/ethernet and perform software updates.

sudo apt-get update
sudo apt-get upgrade

This will take some time.

Since we are not running one of Ubiquity’s robots, run this command to disable their startup scripts.

sudo systemctl disable magni-base

We then need to configure some settings on the Pi 3. Run the configuration dialog:

sudo raspi-config 

Select Interfacing Options then P1 Camera and Yes to enable. Do the same for P2 SSH. We also need to configure several serial options on the config page. Select P5 Serial, select No for "Would you like a login shell to be accessible over serial?" and Yes for "Would you like the serial port hardware to be enabled?".

For Raspberry Pi 3B+, the Bluetooth module occupied uart serial port. To disable the Bluetooth, add dtoverlay=pi3-disable-bt and enable_uart=1 the end of /boot/config.txt. Also, edit the content of /boot/cmdline.txt to dwc_otg.lpm_enable=0 console=tty1 root=/dev/mmcblk0p2 rootfstype=ext4 elevator=deadline fsck.repair=yes rootwait quiet splash plymouth.ignore-serial-consoles.

MavROS is a communication node for ROS with a proxy for a Ground Control Station. It allows us to send MAVLnk messages from the onboard computer to the flight controller. More information can be found here.

To install on the Raspberry Pi 3, run the following commands:

sudo apt-get install ros-kinetic-mavros ros-kinetic-mavros-extras

wget https://raw.githubusercontent.com/mavlink/mavros/master/mavros/scripts/install_geographiclib_datasets.sh

To run geographiclib script, make sure to add execute permissions then run the script.

chmod a+x install_geographiclib_datasets.sh
sudo ./install_geographiclib_datasets.sh

Now is a good time to run a test. Hook up the Raspberry Pi 3 to the Pixhawk as shown below.

Power them both up and run the below code to in a terminal. (The first command sets the publishing rate to 10 Hrtz for all MavROS topics, which we found was needed for the publishing of topics to go through)

rosrun mavros mavsys rate --all 10
roslaunch mavros apm.launch fcu_url:=/dev/ttyAMA0:921600

Open up a new terminal window and issue the below command.

rostopic echo /mavros/imu/mag

Should see IMU data published to the terminal as illustrated below.

The second Raspberry Pi we will use is a Raspberry Pi Zero. Ours runs the lightweight version of Raspbian Buster. That image can be downloaded here.

After downloading, use image writing software like Etcher to write this image to your micro SD card and when finished DO NOT remove from your computer just yet. Continue below.

To interface the Raspberry Pi 3 and Pi Zero, we chose to go with setting up the Zero as an 'ethernet gadget'. This allows us to network the two microcontrollers together so that we can SSH into the Pi Zero from the Pi 3 without using a wireless connection or ethernet port. We can also SCP images collected from the Pi Zero camera over to the Pi 3 for easier retrieval of data post flight.

To begin the connections we must do some configuration on the SD card with the Raspbian OS on the Pi Zero first. On the root directory (i.e. /boot) open the file config.txt with your preferred text editor using sudo. Scroll down to the bottom of this file and add dtoverlay=dwc2 to a new line. Save and exit.

Next open cmdline.txt and scroll across to find the command rootwait. After rootwait add modules-load=dwc2,g_ether being careful not to add and extra spaces or newline characters.

Lastly, we need to enable SSH so that we can communicate with the Pi Zero once it's connected to the Pi3. We can do this by placing a file name ssh, without an extension, onto the boot partition.

Now we can connect the Pi's and test the connection. You will need a Type B Micro to type A USB cable (Male to Male) that has data capabilities (i.e. not just voltage and ground). Note that the Micro USB plug goes into the port labeled USB on the Pi zero (not the PWR IN). This will provide both the data connection and power to the Pi Zero.

Power up the Pi 3 and the Pi Zero should power itself up. After the Pi 3 has booted up, open a terminal window and test the connection.

ping raspberrypi.local

If successful, next SSH into the Pi Zero.

ssh [email protected]

The default password is raspberry.

Lastly, we need to enable the camera interface similar to how we did with the Pi 3.

sudo raspi-config 

Select Interfacing Options then P1 Camera and Yes to enable.

We are using a GPIO interrupt to trigger the camera on the Pi Zero. When the image capture command is received by the Pi 3, our main onboard computer running ROS, it sets a GPIO pin high that is connected to a GPIO pin on the Pi Zero that has code running to catch this signal and trigger its camera.

Our code uses the GPIO21 on both the Pi 3 and Pi Zero for the interrupt but it can be tailored to any pin if need be, just update the code. Also the two Pi's need to share a common ground. A wiring diagram illustrates our setup below.

Now it is time to configure the Pixhawk's camera trigger. Power up the flight controller and connect it to QGroundControl through either USB or telemetry. Open the Parameters tab and in the side pane select CAM. Set the following parameters:

• CAM_RELAY_ON : High
• CAM_TRIGG_TYPE : Relay
• CAM_DURATION : 5 ds

Then in the side pane select RELAY and the following parameters:

• RELAY_DEFAULT : Off
• RELAY_PIN : Pixhawk AUXOUT5

The Pixhawk camera trigger is now on AUXOUT5 which we then connect to another GPIO Pin on the Raspberry Pi 3 that is configured with an interrupt. The wiring is illustrated below.

We now have everything set up to capture images using the Pixhawk and on board computer. Connect the camera's to each Pi: the RGB to the Pi 3, and NIR to the Pi Zero. Boot up both the flight controller and Pi and open a terminal to run these commands:

rosrun mavros mavsys rate --all 10
roslaunch mavros apm.launch fcu_url:=/dev/ttyAMA0:921600

Open a new terminal tab and run the image capture server that starts the ROS service for the camera.

rosrun multispectral_camera capture_image_server.py

To capture an image call the service in a new terminal.

rosservice call /capture_image 1

When the ROS capture_image_server is terminated (i.e. Ctrl-c) the images from the Pi Zero will be automatically SCP over to the Pi 3 and are stored in ~/images.

Each camera triggering will capture an RGB and NIR photo. We are currently using these two images to compute NDVI, Normalized Difference Vegetation Index, which is a scale that estimates plant chlorophyll content. An example we captured with this system is below.