Looking Back
In September 2019, we participated in our second aerial competition in Madrid, the “International Micro Air Vehicle Competition and Conference” (IMAV). This competition is sponsored by Airbus, Parrot, the Spanish Postal Service, Correos, and many other companies. It was held at the Polytechnic University of Madrid.
With our quadcopters for autonomous indoor navigation, we received 3rd place and a special award for the best package-handling. Twelve international teams from four different continents participated in the competition. Our interdisciplinary team combined expertise in the fields of computer science, mechanical engineering,and electrical engineering.
The team would like to express special thanks for the support of the Voss Foundation and numerous other sponsors from the drone industry.
What was the goal of the competition?
The aim was to develop autonomous drones for use in an automated warehouse. Among other things, they were to inventory shelves, transport packages, and navigate and land autonomously.
How did hardware development go?
The quadcopter was developed from scratch by our team members from various engineering courses (mechanical engineering and electrical engineering). As part of our competition preparations, two almost identical versions of the drone were built so that the tasks at the competition could be completed simultaneously.The two versions also offered a backup in the event of a software or hardware defect. The focus was on integrating all the components required for the complete onboard calculation of the computer vision tasks. Another design factor was the ability to transport and drop off packages at destinations. The main material for the components is carbon fiber panels, while the chassis is made of lightweight wood. All parts were custom-made and CNC milled. The mounts for the cameras, the landing gear and the packaging were designed and 3D printed by our team.
How did Software Development go?
The software stack of the Raupe Nimmersatt (The Very Hungry Caterpillar) drones is based on the Robot Operating System (ROS), the most common platform for developing robotics applications. Furthermore, we used packages like MAVROS for the application and integration of the flight hardware into the navigation and control system.
During the prototype phase, we realized that the computing power of a single computer was insufficient to fulfill all the required tasks. Therefore, we decided to distribute the tasks between an NVIDIA Jetson Nano and a Raspberry Pi 4. The ROS core was handled by the Jetson Nano, our main device, while the Raspberry Pi communicated with the Jetson Nano via Ethernet. For controlling the flight hardware, we connected the Pixhawk flight controller to the Jetson via USB. To quickly test our software stack and ROS nodes, we modeled the entire environment inside a Gazebo simulation. Both the T265 and the PiCam were connected to the Jetson. To check the positional data of the T265 and compensate for drift, we used Aruco markers as solid reference points in the environment. For marker recognition, we used a camera mounted on the belly of the drone. This setup allowed us to quickly determine and verify a global position within the environment. Our team developed two nodes, Mapper and Fusion, to merge this data.
Technical Specifications
Empty Weight: 1.200gMTOW: 2.500g
Frame Size: 330mm
Time of construction and development: July 2019 – August 2019
Motors: 4x F80 Pro 2200kv
Regulator: 4x HGLRC Forward FD50A
Battery: 2p 14,8V 2600mAh LiPo (provided by MyLipoShop)
Propeller: 7×3.8in
max. thrust: 8.500g
Special details
- Intel Realsense T265
- Nvidia Jetson Nano
- Raspberry Pi 4
- Holybro Pixhawk PX4
- 2x Pi Cam