For the current autonomous drone development platform, in addition to the autonomous drone development platform produced by our Amu Lab, domestic users have another choice, which is the PX4-Vision autonomous drone development platform.
The hardware is pixhawk 4, an UP Board onboard computer, and a GPS. Since it is an autonomous drone, the focus is on the configuration of the onboard computer.
The configuration is as follows: the avoidance ROS function package 
of https://github.com/PX4/
avoidancepx4 is pre-installed .
By default, the file avoidance.launch is automatically started at boot, and the obstacle avoidance algorithm VHF+ (local_planner) for local path planning is started.
The CPU used to run this algorithm is as follows:
This algorithm is based on the 3D point cloud data to achieve the VFH algorithm, the calculation is relatively small, of course, if you do not open the desktop CPU usage will be lower. The overall CPU is around 40%.
Sensor configuration:
There are many types of optical flow and laser fixed-height radars to choose from. If there is optical flow, it should be able to fly indoor fixed-point mode. But it is not configured by default, and the configuration method of related optical flow module and fixed height radar is not found on the official website of PX4. This should be studied.
The Structure Core depth camera is a depth camera from Occipital . Another Microsoft HoloLens headset is also from this company. The detailed parameters are as follows: https://structure.io/structure-core/specs The
perception distance is 0.3 To about 5 meters, look at the ambient light, global shutter, similar to Intel's D435I.
The main function:
Run avoidance's routine code, the default is to run the VFH algorithm. The octomap algorithm with global path in this ROS function package of avoidance can also be run. We will evaluate this VFH algorithm in the video evaluation. We will test the octomap algorithm later. .
- This UAV can only be used and flown outdoors at present, relying on GPS to achieve obstacle avoidance and local path planning. Mainly study the path planning related to obstacle avoidance.
- Mainly study the path planning related to obstacle avoidance.
- The available functions are obstacle avoidance in fixed-point mode and obstacle avoidance in mission mode.
- The available functions are obstacle avoidance in fixed-point mode and obstacle avoidance in mission mode. There are VIO related content in the development kit, and it also supports Intel realsense sensor. I didn't see the code of visual odometer, but you can use T265 to implement related functions.
- It mainly provides a binocular vision path planning algorithm framework and a set of hardware development platform, which uses the better depth cameras on the market.
The flight control firmware code is still in the open source process.
| product | PX4-Vision | P200 (Amu Laboratory) |
|---|---|---|
| Wheelbase | 290mm | 410mm |
| Use range | Outdoor available | Available indoor and outdoor |
| Scalability | —— | it is good |
| Lidar | not support | stand by |
| Vision sensor | stand by | stand by |
| Function support | ROS-based development | ROS-based development |
Summary: Used for outdoor path planning, using binocular structured light camera, mainly used for path planning, ROS platform, some tracking functions of P200 can be put on it to realize some recognition and tracking algorithms. If it is VHF obstacle avoidance, the laser radar method is better in some situations. The binocular structured light sensor has limited sensing distance and accuracy, so we will find that when there are no obstacles in the mission mode, this PX4-Vision is still executing the obstacle avoidance algorithm action, which causes the route to go wrong. Very ideal. If it is the obstacle avoidance of global path planning, the amount of calculation is relatively large. In a small-scale scene, both laser and vision can get some effects.
Introduction to PX4-Vision:
PX4-Vision unboxing video
Introduction of autonomous drone development platform px4-vision
Outdoor obstacle avoidance test **: **
Outdoor obstacle avoidance test video
There are two modes of outdoor obstacle avoidance: fixed-point obstacle avoidance and task mode obstacle avoidance . The obstacle avoidance remote control in fixed-point mode can only control forward flight and yaw, and cannot reverse and roll. If the light interference is large, it will always be in an obstacle avoidance state, which is related to the immature technology of the current binocular camera. Among them, obstacle avoidance in mission mode uses the VFH+ algorithm, and the binocular structured light camera used is integrated with the global path planning algorithm, but it is not turned on by default.
The state of flying a route mission in the VHF+ obstacle avoidance state, and the flying altitude is 3.5 meters to 4 meters.
Our agent sales address: https://item.taobao.com/item.htm?spm=a1z10.5-cs.w4002-22617251051.20.2fa84720wgl3BH&id=613762545009