Have you ever had these problems:
Fixed-point mode takeoff refused-insufficient number of GPS stars
One-click takeoff after drawing the route-positioning is not allowed to deviate from the route
Indoor drone-can only take off in self-stabilization mode
………
UAV positioning has always been a topic that cannot be avoided in the development of UAVs. Whether it is route planning or autonomous driving, more and more developers are pursuing more precise positioning solutions, such as GPS, RTK, etc., but for the completion of indoor development such as For drones for tasks such as patrols, pipeline inspections, and building group patrols, GPS signals are often blocked and weakened, or even completely lost. UAV positioning becomes a major problem. In response to the demand for drone positioning in a GPS-free environment, Amu Lab has launched a vision precision positioning solution, and customers are welcome to consult and cooperate.
1. Introduction to the visual positioning scheme
In previous articles, we have introduced some indoor positioning solutions, including lidar indoor obstacle avoidance.
However, the above schemes inevitably have some defects, such as the large size of lidar and the inability to restore the real scene; ultrasonic positioning information is susceptible to interference; UWB requires the deployment of base stations in advance. Visual positioning solves this problem very well. It is not only small in size and accurate in positioning, but also provides relatively rich visual information to help visual obstacle avoidance, three-dimensional reconstruction and other functions.
This visual positioning solution is developed based on the company’s mature P200 quad-rotor UAV platform, and is improved for the company’s next-generation scientific research UAV platform P300. At the same time, it supports customer customization. The entire system can be seamlessly connected to each UAV-like platform. The program has now passed the stages of development, simulation, and partial actual testing, and is undergoing comprehensive testing in various environments.
Two, system architecture
The visual positioning system consists of an onboard computer (P200 platform is equipped with NVIDIA TX2/Nano and supports X86 architecture computers), binocular camera (Intel T265) and RGBD camera (Intel D435i). The T265 camera provides high-precision positioning information through visual SLAM, and the D435i realizes three-dimensional reconstruction.
The software architecture is completely based on the Linux open source system, and communicates with the camera and flight controller through the ROS system to provide real-time positioning information. The positioning delay can be controlled to the nanosecond level, and the indoor positioning accuracy can reach the centimeter level.
3. Measured results
Indoor positioning 1
Indoor positioning 2
Three-dimensional reconstruction
The relevant content of this set of visual positioning solutions has been published on our open source autonomous drone software platform-Prometheus (Prometheus). Customers and readers are welcome to provide valuable suggestions and opinions. https://github.com/amov-lab/Prometheus
With the rapid development of technology, Amu Lab will keep up with the pace of technology and continue to recommend the latest technology and hardware in the drone industry to everyone. It is the greatest value of our training to see that the trainees who have been trained by us are advancing by leaps and bounds in technology. If you are in the drone industry, please pay attention to our official account. We will continue to release the most valuable information and technology in the drone industry.