1. Background
Whether it is civilian or military. Small clustering of drones/robots is a trend. At present, many logistics robots are used, and many robots cooperate to complete tasks.
With the development of technology and entering the era of robots, communication and collaboration between machines is an inevitable trend. Data sharing of cluster robots, sensor status data sharing, and mutual control interface. To achieve that any robot in the cluster can obtain the status data of other robots, control the movement of other robots, and realize group collaborative work. The choice of communication mechanism is more important and involves the networking of cluster robots.
This article is to discuss and test the use of ROS communication mechanism to realize multiple drone clusters. Each drone in the network can realize mutual data reading and mutual control, and the formation of drone network formation and perception coordination .
Unlike ordinary formation performances, in formation performances, the ground station has one control and multiple controls, and the drone formation can be set in advance. Multi-aircraft collaboration means that each UAV must interact with all UAVs or the surrounding UAVs and control each other. The entire UAV formation is not preset, but the UAVs sense and control each other. As a result, the algorithm converges to an optimal UAV trajectory to achieve collaborative work.
2. System functional block diagram
There are two aspects to consider in group work:
The first communication hardware system must support the networking and interconnection of each node (drone/unmanned vehicle). The requirement is duplex communication to form a network. There are many data transmission modules for duplex communication, but large-scale networking is possible. A little bit less. Generally, a TCP/IP LAN is a star-shaped network with a router as the central node. All data is forwarded through the router's IP address, and each node can communicate with each other. Then the data flow bandwidth of the central node of the router, It is the upper limit of the data transmission bandwidth of the entire network. We usually talk about 100M bandwidth and gigabit bandwidth. Gigabit data capacity is definitely larger than 100M bandwidth. There are also some communication hardware systems without a central node. Generally, the price is relatively high and the bandwidth is limited. It is not described here. You can also find some solutions by consulting the communication manufacturers.
The structure of the second software system, what is the structure of the group of nodes that each drone/unmanned vehicle runs on, if the ROS solution is adopted, ROS is a node that supports distributed operation, that is, no one The drone and UAV 2 can communicate with each other by means of ROS nodes. The current ROS1 uses a central point of communication, that is to say, in Figure 1, the ROS Master is the center of the entire ROS node, forming a star network structure. This ROS Master is a communication bridge with other nodes. The bad factor is that the central node of ROS Mater is abnormal and the entire network fails.
In terms of hardware, the WIFI router (TCP/IP) solution is also abnormal in the central node of the router, and the network fails.
At the software level, the ROS1 method is the same, and the ROS Master abnormal network also fails.
However, the advantages of using the above are: There are many mature options for WIFI LAN equipment.
ROS1 also provides distributed multi-nodes, which solves the problem of software architecture. There is no need to consider the software problem of mutual communication. The ROS1 network message list contains all the data and control instructions of all current network nodes. UAV 1 obtains the UAV According to the sensor data of 2, UAV 1 and UAV 2 can mutually control each other's flight speed and flight position. This forms the basis of swarm intelligence. Provide an interface for mutual awareness and mutual control.
However, the number of groups that can be accommodated in this way needs to be considered. We believe that this way of accommodating 10 unmanned equipment is relatively easy in both 3D simulation and real UAV flight. If you want to accommodate more unmanned systems, the message flow should be streamlined and hardware devices with higher communication bandwidth should be selected. In theory, a network segment of the LAN can accommodate 255 devices.
Another point is the launch of ROS2, which supports DDS, a communication method without a central node. The software communication level has no central point. The stability is greatly improved. At the software message communication level, it is a group intelligence that is very suitable for networking. Because ROS2 was launched soon, the follow-up is still worth studying. Especially the real distributed mode of running DDS.
ROS1 system software operation diagram
Part of the message node graph
Using the ROS distributed message mode, for example, UAV No. 1 sends instructions to UAV No. 2, just add a prefix /uav2/**** or /uav1/****, you can read or send Control instructions to UAV #2. The ROS system maintains a message instruction queue (Topic) of all drones. According to the index message instruction queue, the sensor messages and control instructions of all drones in the cluster can be shared.
Simulation video
Using the ROS environment simulation code is highly consistent with the real machine code, and there is no need to consider the problem that the simulation code cannot be used on the real machine.
5 UAVs in a simulation environment for multi-aircraft formation switching
Actual test video
UAV formation test: 4 UAVs outdoor flight formation test
The above is a test of 4 UAVs. Because the general GPS is used and the accuracy is limited (0.5 to 2 meters), the formation of the formation is not ideal. But you can see the basic effect of multi-machine collaboration, which is the software and hardware basis for future intelligent algorithm collaboration.
5 UAV formation collaborative test video
UAV formation test based on ROS node multi-machine coordination
The architecture is PX4 firmware 410 wheelbase UAV, ROS, MAVROS .
The communication network hardware equipment is as follows:
The system involves networking hardware introduction: https://homer-wiki.readthedocs.io/en/latest/
The actual working hardware bandwidth status, communication method and communication bandwidth are very important considerations. We recorded the communication bandwidth data of 5 ROS1.
Only the system bandwidth occupied by the ROS master node is as follows:
The peak value is 347KB, which is the bandwidth occupied by a main drone to start ROS data flow interaction.
The bandwidth occupied by 5 UAVs simultaneously starting the system is as follows:
The peak value is 1.16MB/S, the design bandwidth of this hardware image and data transmission is 100M, and the theoretical transmission bandwidth is 12MB/S. Therefore, the network coordination of the five UAVs is far from being online. It is estimated that the network of about 10 ROS drones can still be realized.
If all image and video data are to be shared, the pressure on bandwidth will be greater. However, it is relatively easy to share data such as general sensor position data, speed data, status data, and control command data, and the bandwidth can still meet the requirements. Secondly, the communication data can be streamlined. In the cooperative state, only data such as position, speed, attitude, and status can be retained. It can also accommodate more drones when the bandwidth is limited.
Pull- saw bandwidth test : (communication bandwidth will be affected by distance. See-saw test is necessary, but it is limited to the venue. Later we will conduct ground-saw bandwidth test)
- ROS course recommendation (used to supplement ROS foundation): https://bbs.amovlab.com/plugin.php?id=zhanmishu_video:video&mod=video&cid=32
- Recommended courses for drone ROS simulation (the course about simulating cluster drones will be added later): https://bbs.amovlab.com/plugin.php?id=zhanmishu_video:video&mod=video&cid=18
For this product and project, we want to eventually evolve into a fusion of about 10 drones for obstacle avoidance and intelligent collaboration of images. For example, the pan/tilt of one of the drones finds the target of interest and links all the drones to achieve the initial stage. Group intelligence. We are also seeking cooperation with powerful universities and research institutes. We provide hardware platforms and common interfaces to jointly realize the UAV intelligent collaborative control system.
The WeChat public account pays attention to "Amu Lab" to obtain the dry goods of cutting-edge technology and evaluation of drones for the first time.