Improved consistency control algorithm

Abstract

Due to the fact that a traditional unmanned aerial vehicle formation control system is highly dependent on a host, and a traditional consistency control algorithm is not good in convergence effect on a time-varying reference state, the invention provides an improved second-order consistency control theory. According to the theory, in the aspect of communication, all the unmanned aerial vehicles only need to establish communication links with part of the unmanned aerial vehicles, it is guaranteed that the whole unmanned aerial vehicles are a minimum directed spanning tree, and then deployment of a consistency algorithm can be achieved. Therefore, the consistency control algorithm added with the time delay item provided by the invention can carry out better convergence in a time-varying reference state. And the unmanned aerial vehicles can keep a more stable flight formation in actual flight.

Description

Technical Field

[0001] The present invention relates to the field of consistency control, and in particular to consistency convergence regulation of time-varying reference states. Background Art

[0002] With the rapid development of drone applications, the limitations of a single drone are becoming increasingly greater, and people's demand for multiple drones is also showing a rapid development momentum. Drone systems have achieved a lot of research results in cluster control, path planning, formation transformation, etc. In recent years, consistency theory based on graph theory and matrix theory has developed rapidly. At present, drone formation methods can be divided into virtual pilot-follower method, virtual leader method, behavior-based method, consistency, etc.

[0003] Compared with other control algorithms, the biggest advantage of consistency control is that it can be deployed in a distributed structure. The algorithm can be deployed on each drone, and the bandwidth and computing power requirements are relatively low. The idea of ​​consistency control is that all drones are independent individuals, and all calculations are scattered on each drone. In the end, the speed of all drones tends to be consistent, thereby achieving a formation effect. The consistency theory has been optimized from the first order to the current second order. Because for the position of the drone, the first order corresponds to the speed. Due to the actual reasons of the drone, the control of speed cannot achieve sudden changes in speed. Therefore, the second-order consistency is introduced to control the acceleration of the drone, that is, to control the speed of the propeller. Compared with the first-order control, the second-order drone control has achieved better results in actual convergence. In terms of communication, all drones only need to establish communication links with some drones to ensure that the whole is a minimum directed spanning tree to achieve the deployment of the consistency algorithm.

[0004] However, the traditional second-order consistency control algorithm does not have a very good convergence effect on the time-varying reference state. The farther the distance between the node and the reference state in the communication topology diagram is, the greater the delay. Therefore, the consistency control algorithm with the delay term proposed in the present invention can converge better under the time-varying reference state, so that the UAV can maintain a more stable flight formation in actual flight. Summary of the invention

[0005] In view of the fact that the existing second-order consistency algorithm does not have an ideal convergence effect on the time-varying reference state, an improved second-order consistency algorithm control theory is proposed. The present invention adds a delay term to the original algorithm so that it can have a better convergence effect under the time-varying reference state.

[0006] The technical solution adopted by the present invention is as follows:

[0007] Step 1: Using the knowledge of graph theory, the drone cluster system is regarded as a directed generated graph Γ = (K, E u , Au ) generated by the system. The graph contains the node set K and the edge set E u and the weighted adjacency matrix A u . K represents the node and the number of the drone; E u is an edge set, indicating whether there is a communication connection between drones; A u is a weighted adjacency matrix, which represents the credibility of messages between drones.

[0008] Step 2:

[0009] The biggest advantage of distributed drones is that there is no leader. All drones only need to have a minimum spanning tree between the communications to achieve the interconnection of the entire drone. Therefore, the communication topology used in the present invention is a directional communication topology, which can greatly save the occupation and delay of communication resources. Therefore, the present invention uses five drones to form a simulation cluster, and its communication topology diagram is as follows Figure 1 shown.

[0010] Step 3: Since the traditional formation method often requires all drones to refer to the leader or reference state, all drones need to be connected to the host, which is a waste of communication resources. And when the leader or reference state is unexpected, the overall formation system cannot maintain the formation. Therefore, a consistent distributed formation control method is proposed, and its formula is shown in the figure below:

[0011]

[0012] In the formula, a ij >0,a ii = 0, indicating the information weight between agents. When agent j can receive the message of agent i, p ij =1, otherwise p ij = 0. i is the location information of drone i.

[0013] Step 4: The typical first-order consistency control algorithm controls the speed of the drone, but since the drone is an inertial system, we cannot directly control the speed, so the second-order consistency control theory is proposed. The second-order consistency control theory is as follows:

[0014]

[0015] Where ε id , ε jd They represent the expected deviation of the agent ij in the formation, γ is an adjustable parameter, which is related to the stability and convergence of the system. is i The derivative with respect to time, and acceleration. When |(ε i -ε id)-(ε j -ε jd )|→0 and , the agents in the system can reach a consistent state by obtaining the state information of their neighbors, so this paper will Name it as a synergy term.

[0016] Step 5: However, the traditional second-order consistency control algorithm can achieve a better fit to the reference state than the first-order algorithm. However, if the reference state is a real moving drone, and the motion trajectory of the reference state is not a fixed flight path but time-varying, the second-order consistency control algorithm cannot achieve a good fit. The drone far away from the reference state has a large delay in fitting the reference state. Therefore, an improved second-order consistency control theory is proposed. On the basis of the original second-order consistency control algorithm, the delay parameter is introduced. And the delay term:

[0017]

[0018] Under this control law, only some agents can receive the reference state, and the system can still reach a consistent state when the first-order reference state is time-varying.

[0019] Step 6: To better adjust the stability and convergence of the system, modify the adjustable parameters. The final improved consistency algorithm is:

[0020]

[0021] Where γ1 and γ2 are adjustable parameters. It should be noted that a stable formation is generally maintained in the formation, so the expected deviation ε in the formation id , ε jd are constants, so their derivative with respect to time t is 0, that is, u jd =0.

[0022] The beneficial effects of the present invention are as follows:

[0023] 1. Compared with the traditional formation method, the present invention has higher stability and robustness, and can still control the drone cluster to reduce losses when unexpected situations occur.

[0024] 2. The present invention improves the fitting effect of the traditional consistency control algorithm under the time-varying reference state, and can deploy the consistency algorithm in a more complex and changeable motion environment.

[0025] 3. The present invention is simulated and verified under simulation software, which proves the feasibility of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0027] Figure 1 It is the communication topology diagram between drones.

[0028] Figure 2 It is the convergence diagram of the first-order consistency control algorithm.

[0029] Figure 3 This is the effect diagram of the second-order consistency algorithm.

[0030] Figure 4 This is a graph showing the convergence effect of the improved consistency algorithm on the time-varying reference state. Specific implementation methods

[0032] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0033] like Figure 1 As shown, the communication topology diagram of the UAV fitting effect in this experiment is as follows Figure 1 As shown in the figure. All drones only need to receive information from their neighbors to ensure that the entire drone cluster is a minimum spanning tree, and the formation control of the drone cluster formation can be realized. The step length δt = 0.1s, the initial position and yaw angle of drones 1 to 5 are defined as shown in Table 1, and their initial values ​​can be set arbitrarily. Figure 1 It can be seen that only UAV No. 1 can receive the reference state. The overall simulation process is that the UAV takes off to a height of 5m, and then all UAVs hover until t = 15s and begin to form a formation according to the preset reference formation. The preset formation is a three-dimensional rectangular formation with a spacing of 2m. When t = 30s, all UAVs have formed a preset formation and start to move as a whole according to the preset reference state speed. The preset reference speed is time-varying, and its x-axis speed is set to 0.5m / s and the y-axis speed is set to

[0034] Table 1. Initial position and speed settings of simulated drone.

[0035]

[0036] like Figure 2 As shown, it is the convergence effect of conventional first-order consistency control theory on the reference state. Figure 3 This is the convergence effect diagram of the conventional second-order consistency control theory. Figure 2 and Figure 3 From the comparison, we can see that the second-order consistency has more advantages than the first-order in the convergence effect of the reference state and in actual deployment. Since the drone is an inertial system, we cannot directly control the speed of the drone, and its speed has a differentiable change process, that is, we can only control the speed and acceleration of each motor of the drone. Therefore, the biggest advantage of the second-order consistency over the first-order consistency is that it can be deployed on an actual drone. However, when the reference state in the second half is not a constant value, but begins to change irregularly, the fitting effect of the traditional second-order consistency algorithm is not very good, and there is a large delay effect.

[0037] Figure 4 This is the convergence effect diagram of the improved second-order consistency algorithm. It can be seen that it has a better fitting effect on the time-varying reference state. To explain it another way, the ordinary second-order consistency algorithm has always had a delay problem, but because the reference state has always remained stable, its moving target at this moment is the reference state of the previous moment, but because the reference state has always remained constant, it cannot be seen.

[0038] The method of the present invention has been verified through experiments to have a better fitting effect in actual UAV flight than the second-order consistency control theory. The algorithm of the present invention has been verified by the simulation system, so the algorithm of the present invention can be deployed in actual UAVs. Compared with the traditional UAV formation control system, it can have a better fitting effect and a more stable UAV formation motion control effect.

Claims

1. Using the knowledge of graph theory, the drone cluster system is regarded as a directed generated graph Γ = (K, E u , A u ) generated by the system. The graph contains the node set K and the edge set E u and the weighted adjacency matrix A u . K represents the node and the number of the drone; E u is an edge set, indicating whether there is a communication connection between drones; A u is a weighted adjacency matrix, which represents the credibility of messages between drones.

2. The biggest advantage of distributed drones is that there is no leader. All drones only need to have a minimum spanning tree between communications to achieve the interconnection of the entire drone. Therefore, the communication topology structure adopted by the present invention is a directional communication topology, which can greatly save the occupation and delay of communication resources. Therefore, the present invention uses five drones to form a simulation cluster, and its communication topology diagram is shown in Figure 1.

3. Since the traditional formation method often requires all drones to refer to the leader or reference state, all drones need to be connected to the host, which is a waste of communication resources. And when the leader or reference state is unexpected, the overall formation system cannot maintain the formation. Therefore, a consistent distributed formation control method is proposed, and its formula is shown in the figure below: In the formula, a ij >0,a ii = 0, indicating the information weight between agents. When agent j can receive the message of agent i, p ij =1, otherwise p ij =0.

4. The typical first-order consistency control algorithm controls the speed of the drone, but since the drone is an inertial system, we cannot directly control the speed, so the second-order consistency control theory is proposed. The second-order consistency control theory is as follows: Where ε id , ε jd They represent the expected deviation of the agent ij in the formation, γ is an adjustable parameter, which is related to the stability and convergence of the system. is i The derivative with respect to time, and acceleration. When |(ε i -ε id )-(ε j -ε jd )|→0 and When , each agent in the system can reach a state consensus by obtaining the state information of its neighbors, so this paper will Name it as a synergy term.

5. However, although the traditional second-order consistency control algorithm can achieve a better fit to the reference state than the first-order algorithm, if the reference state is a real moving drone and the motion trajectory of the reference state is not a fixed flight path but time-varying, the second-order consistency control algorithm cannot achieve a good fit. The drone far away from the reference state has a large delay in fitting the reference state. Therefore, an improved second-order consistency control theory is proposed. On the basis of the original second-order consistency control algorithm, the delay parameter is introduced. And the delay term: Under this control law, only some agents can receive the reference state, and the system can still reach a consistent state when the first-order reference state is time-varying.

6. In order to better adjust the stability and convergence of the system, the adjustable parameters are modified. The final improved consistency algorithm is: Where γ1 and γ2 are adjustable parameters. It should be noted that a stable formation is generally maintained in the formation, so the expected deviation in the formation ε id , ε jd are constants, so their derivative with respect to time t is 0, that is, u jd =0.

Images