Collaborative navigation hierarchical fusion topological structure optimization system and method

Abstract

The invention relates to a collaborative navigation hierarchical fusion topological structure optimization system and method, and the system comprises a hierarchical fusion topological structure construction module, a genetic algorithm search module and a correlation detection module, a hierarchical fusion topology optimization model is established, a first-layer fusion topology is generated, a genetic algorithm search module performs search based on the first-layer fusion topology by using a genetic algorithm to obtain a second-layer fusion topology, and a correlation detection module performs correlation detection on the second-layer fusion topology and removes correlation information. And an updated second-layer fusion topology is obtained and output to the genetic algorithm search module for iterative updating until an optimal hierarchical fusion topology is obtained. The hierarchical fusion topological structure is modeled by means of the algebraic graph theory, then the optimization model is established, the genetic algorithm is designed for solving, resource consumption is reduced by optimizing the topological structure, and meanwhile it is guaranteed that the navigation performance meets the requirement.

Description

Technical Field

[0001] The present invention relates to a technology in the field of unmanned aerial vehicle control, specifically a collaborative navigation hierarchical fusion topology structure optimization system and method. Background Art

[0002] Collaborative navigation is a technology that uses communication and relative observation between intelligent agents to achieve navigation resource sharing. However, existing technologies cannot select appropriate topological structures based on positioning requirements and communication and computing resource constraints to achieve a compromise between positioning accuracy and resource expenditure. They consume a lot of resources and are difficult to adapt to dynamic scenarios. Summary of the invention

[0003] In view of the problem that the prior art does not consider topology optimization and consumes too much communication computing resources, the present invention proposes a collaborative navigation hierarchical fusion topology structure optimization system and method, models the hierarchical fusion topology structure with the help of algebraic graph theory, and then establishes an optimization model and designs a genetic algorithm for solving it. By optimizing the topology structure, resource consumption is reduced while ensuring that the navigation performance meets the requirements.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a collaborative navigation hierarchical fusion topology structure optimization system, comprising: a hierarchical fusion topology structure construction module, a genetic algorithm search module and a correlation detection module, wherein: the hierarchical fusion topology structure construction module establishes a hierarchical fusion topology optimization model and generates a first-layer fusion topology according to the sensor configuration and resource constraints of an intelligent body; the genetic algorithm search module searches based on the first-layer fusion topology using a genetic algorithm to obtain a second-layer fusion topology; the correlation detection module performs correlation detection on the second-layer fusion topology, removes information with correlation, obtains an updated second-layer fusion topology and outputs it to the genetic algorithm search module for iterative update until an optimal hierarchical fusion topology is obtained.

[0006] The hierarchical fusion topological structure means that in a multi-agent system, the information fused by the agents is layered in order, and the fusion topology of each layer is modeled as a graph variable with the help of the concept of adjacency matrix, and the mathematical expression of the multi-layer topological structure is finally obtained.

[0007] The hierarchical fusion topology structure building module includes: an input unit and an initialization unit, wherein: the input unit generates the sensor configuration and resource constraint information of the intelligent agent, and the hierarchical fusion topology optimization model is established through the initialization unit and the initial fusion topology is generated.

[0008] The genetic algorithm search module includes: a coding unit, a fitness unit and a genetic unit, wherein: the coding unit encodes the fusion topology, the fitness unit evaluates the fitness of each generated offspring and generates a corresponding fitness value, and the genetic unit generates new offspring through mutation operations on the coding of the surviving offspring, and finally searches out the optimal fusion topology.

[0009] The correlation detection module includes: a correlation mapping unit and a multipath detection unit, wherein: the correlation mapping unit lists all existing layers in parallel for mapping according to the information fusion process, and the multipath detection unit detects whether there is correlation based on the results obtained by the correlation mapping unit and removes the information with correlation. Technical Effects

[0010] The present invention uses algebraic graph theory and a hierarchical structure to describe the propagation and fusion mechanism of agent cluster information at different levels with several layers, and obtains a hierarchical fusion topological structure; the topological structure is optimized by a genetic algorithm and the information with correlation is removed by multipath detection. Compared with the prior art, the present invention reduces resource consumption by optimizing the hierarchical fusion topology, while ensuring that the navigation performance meets the requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the system of the present invention;

[0012] Figure 2 It is a schematic diagram of the layered structure of the present invention;

[0013] Figure 3 This is an example of a single-layer fusion topology of the present invention;

[0014] Figure 4 This is a schematic diagram of the principle of the correlation checking module of the present invention;

[0015] Figure 5 This is a schematic diagram of a scenario for implementing this embodiment;

[0016] Figure 6 This is a simulation experiment effect diagram of this embodiment. DETAILED DESCRIPTION

[0017] like Figure 1 As shown, a collaborative navigation hierarchical fusion topology structure optimization method based on the above system involved in this embodiment includes:

[0018] Step 1: Establish a mathematical model of the collaborative navigation topology structure based on the adjacency matrix concept of algebraic graph theory, including: the fusion topology G of sensor information p =(v p ,ε p ) and the fusion topology G of state estimation r=(v r ,ε r ). When the multi-agent system has N agents and M sensors, then G p The nodes can Where i represents the agent i in the multi-agent system, s jk Corresponds to the kth sensor on agent j. G p The edge set Represents whether any sensor on any agent transmits information to the information fusion center on another specified agent. ij Transmit information to the information fusion center of the specified agent k, then the edge G p The adjacency matrix of This means that if the sensor s ij , the information is transmitted to the information fusion center on agent k, then otherwise G r Node set Where i represents the agent i in the multi-agent system, The state estimate of agent j corresponding to agent i. r The edge set represents whether any state estimate on any agent is transmitted to the information fusion center on another agent. If agent i transmits an existing state estimate s to agent k, ij , then the edge G r The adjacency matrix of Indicates that Represents the state estimate s existing on agent i ij Transmitted to the fusion center on agent k, otherwise

[0019] Step 2: The hierarchical fusion topology construction module determines the combination optimization model and constructs the search problem based on the given multi-agent system information to obtain the initial fusion topology, which includes:

[0020] 2.1) Based on the input, an optimization model is established, where: the optimization variable is the hierarchical fusion topology structure; the objective function is the communication and computing resource consumption of the intelligent agent; the constraints include the positioning accuracy requirements and communication and computing resource constraints of the intelligent agent.

[0021] 2.2) Use a hierarchical structure to construct the search problem, split the hierarchical fusion topology into multiple layers, and describe and model them separately. Figure 2 As shown, the positioning accuracy after each layer is taken as state S k , taking each layer as action Ak , the corresponding navigation resource expenditure is taken as k Execute A k The cost of reaching the next state can be described by a series of alternating sequences of S and A, thus constructing a search problem. As an action, a single-layer fusion topology example is Figure 3 shown.

[0022] 2.3) Initialize the hierarchical fusion topology optimization problem and construct the initial state: set the adjacency matrix of the first layer of fusion topology to zero, indicating that at the beginning of a navigation epoch, there is no information exchange and fusion between agents; the corresponding positioning error is set to infinity, indicating that the error of the agent's navigation solution is infinite, that is, no effective navigation solution is generated; generate an initialized single-layer fusion topology with the help of the concept of the adjacency matrix.

[0023] Step 3: The genetic algorithm search module encodes the current state and the initialized single-layer fusion topology, measures the quality of each offspring with fitness, generates offspring through genetic operations, constructs a genetic algorithm and searches, and outputs the search to obtain the optimal new layer of fusion topology, including:

[0024] 3.1) Design coding and generate initial population: Use binary coding and randomly generate an initial population with binary coding.

[0025] 3.2) Design the fitness function and the corresponding genetic operation. The fitness function can be expressed as f = (500P + M) - C S. P represents the number of agents that meet the positioning accuracy requirements, M represents the amount of information used in the new layer of fusion topology, C = 10E9, and S represents whether any of the navigation resources exceeds the maximum limit. If it exceeds, it is 1, and the rest are 0. The genetic operation in this step is random mutation and exchange.

[0026] 3.3) Determine the new layer of information fusion through genetic algorithm search: The genetic algorithm module outputs the optimal adjacency matrix of the new layer of information fusion topology as the information that the corresponding intelligent agent in the new layer needs to be newly fused. The correlation detection module determines the correlation between the candidate information and the original state estimation one by one. When there is information that does not meet the requirements, the corresponding elements of the adjacency matrix are set to zero. When the adjacency matrix is ​​not a zero matrix, it is recorded as a new action A. k+1 , the corresponding state is also advanced to S k+1 Otherwise, it returns to the previous state S k-1 Re-search, using a different action A than before k ', when S k+1 If all constraints are met, the search result is successful and output; if it reaches the initial state and still needs to be rolled back, the search result is a failure.

[0027] Step 4: The correlation detection module checks whether the information to be fused is correlated with the original state estimation. The correlation mapping unit builds a map based on the hierarchical fusion topology, such as Figure 4 As shown in the figure, the information fusion process is displayed in parallel as multiple layers. The arrows between the layers represent the transmission of information between agents. If any agent sends out multiple arrows and these arrows converge on the same agent, it means that the information corresponding to the agent is correlated. By removing this information, the updated hierarchical fusion topology can be obtained.

[0028] Step 5: Repeat steps 3 and 4 to obtain the optimized hierarchical fusion topology structure.

[0029] After specific experiments, in a Figure 5 In the setting of a multi-agent cluster consisting of six agents, experiments were conducted with the parameters shown in Table 1. The performance of the fully measured distributed topology structure under the same circumstances was used as a comparison, and the experimental data obtained are shown in Table 2.

[0030] like Figure 5 As shown in the figure, it is a multi-agent cluster composed of six agents. They have relative distance measurement with each other as shown in the figure. It is required that all navigation solutions of agents 1 to 6 exist on agents 2 and 6, and the accuracy of all navigation solutions on agent 2 can be better than 4.2m, and the accuracy of all navigation solutions on agent 6 can be better than 6m. Through experiments, under the simulation parameters shown in Table 1, the topological structure obtained is as follows Figure 6 shown.

[0031] Table 1 Simulation parameters Settings Number of carriers 6 Pseudorange observation error variance <![CDATA[1m 2 ]]> Relative ranging error variance <![CDATA[0.04m 2 ]]> Maximum computing power resources of carrier (fusion center) Both are 100000 Non-fusion center vector number 2、6 Maximum communication resources of carrier (fusion center) Both are 1200 Navigation solution accuracy requirements (agent 2) <![CDATA[≤2.05m(trace(P)≤4.2m 2 )]]> Navigation solution accuracy requirements (agent 6) <![CDATA[≤2.83m(trace(P≤8m 2 )]]>

[0032] The performance of the fully measured distributed topology commonly used in collaborative navigation in the simulation experiment is further compared. The fully measured distributed topology can meet the navigation solution accuracy of the simulation experiment, but cannot meet the corresponding navigation resource constraints at the same time, as shown in Table 2.

[0033] Table 2

[0034] In the hierarchical fusion topology output by the present invention, the navigation resource expenditure of each agent has obviously decreased. Compared with the fully measured distributed topology, the maximum communication resource expenditure of each agent in the hierarchical fusion topology has decreased by 33.8%, and the total communication resource expenditure has decreased by 38.4%; the maximum computing resource expenditure of each agent has decreased by 60.7%, and the total computing resource expenditure has decreased by 78.9%; however, the positioning error of agent 2 has only increased by 24.4%, and the positioning error of agent 6 has only increased by 2.0%, and the hierarchical fusion topology is closer to the simulation requirements, and there is no pursuit of excessively high positioning accuracy on agent 2, resulting in unnecessary navigation resource expenditure.

[0035] Compared with the prior art, the present invention takes state estimation into consideration as a special observation by modeling the mathematical model of the collaborative navigation topology structure, making full use of the information in the state estimation, reducing repeated information propagation and fusion, ensuring the accuracy of the navigation solution, and greatly reducing the collaborative navigation resource expenditure. Through the genetic algorithm search module, the communication resources and computing power resource constraints of each agent are considered at the same time during optimization, so that compared with the fully measured distributed topology structure, the present invention also has a significant reduction in the maximum value of the navigation resource expenditure of a single agent, and significantly reduces the requirements for the hardware facilities of the agent. In addition, through the genetic algorithm search module, the present invention can consider the accuracy of the navigation solution obtained by each agent in the system and whether there is a navigation solution for the specified agent on each agent in the system for the collaborative navigation positioning accuracy, which is more comprehensive than the prior art.

[0036] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A collaborative navigation hierarchical fusion topology structure optimization system, characterized in that: include: A hierarchical fusion topology structure building module, a genetic algorithm search module and a correlation detection module, wherein: the hierarchical fusion topology structure building module establishes a hierarchical fusion topology optimization model and generates a first-layer fusion topology according to the sensor configuration and resource constraints of the intelligent body; the genetic algorithm search module searches based on the first-layer fusion topology using a genetic algorithm to obtain a second-layer fusion topology; the correlation detection module performs correlation detection on the second-layer fusion topology, removes the information with correlation, obtains the updated second-layer fusion topology and outputs it to the genetic algorithm search module for iterative update until the optimal hierarchical fusion topology is obtained.

2. The collaborative navigation hierarchical fusion topology structure optimization system according to claim 1 is characterized in that: The hierarchical fusion topological structure means that in a multi-agent system, the information fused by the agents is layered in order, and the fusion topology of each layer is modeled as a graph variable with the help of the concept of adjacency matrix, and the mathematical expression of the multi-layer topological structure is finally obtained.

3. The collaborative navigation hierarchical fusion topology structure optimization system according to claim 1 is characterized in that: The hierarchical fusion topology structure building module includes: an input unit and an initialization unit, wherein: the input unit generates the sensor configuration and resource constraint information of the intelligent agent, and the hierarchical fusion topology optimization model is established through the initialization unit and the initial fusion topology is generated.

4. The collaborative navigation hierarchical fusion topology structure optimization system according to claim 1 is characterized in that: The genetic algorithm search module includes: a coding unit, a fitness unit and a genetic unit, wherein: the coding unit encodes the fusion topology, the fitness unit evaluates the fitness of each generated offspring and generates a corresponding fitness value, and the genetic unit generates new offspring through mutation operations on the coding of the surviving offspring, and finally searches out the optimal fusion topology.

5. The collaborative navigation hierarchical fusion topology structure optimization system according to claim 1 is characterized in that: The correlation detection module includes: a correlation mapping unit and a multipath detection unit, wherein: the correlation mapping unit lists all existing layers in parallel for mapping according to the information fusion process, and the multipath detection unit detects whether there is correlation based on the results obtained by the correlation mapping unit and removes the information with correlation.

6. A collaborative navigation hierarchical fusion topology structure optimization method based on the system described in any one of claims 1 to 5, characterized in that: include: Step 1: Establish a mathematical model of the collaborative navigation topology structure based on the adjacency matrix concept of algebraic graph theory, including: the fusion topology G of sensor information p =(v p ,ε p ) and the fusion topology G of state estimation r =(v r ,ε r ), when the multi-agent system has N agents and M sensors, then G p Node Where i represents the agent i in the multi-agent system, s jk Corresponding to the kth sensor on agent j, G p The edge set It represents whether any sensor on any agent transmits information to the information fusion center on another specified agent. ij Transmit information to the information fusion center of the specified agent k, then the edge G p The adjacency matrix of This means that if the sensor s ij , the information is transmitted to the information fusion center on agent k, then otherwise G r Node set Where i represents the agent i in the multi-agent system, The state estimate of agent j on agent i, G r The edge set represents whether any state estimate on any agent is transmitted to the information fusion center on another agent. When agent i transmits the existing state estimate s to agent k, ij , then the edge G r The adjacency matrix of Indicates that Represents the state estimate s existing on agent i ij Transmitted to the fusion center on agent k, otherwise Step 2: The hierarchical fusion topology construction module determines the combination optimization model and constructs the search problem based on the given multi-agent system information to obtain the initial fusion topology, which includes: 2.1) Based on the input, an optimization model is established, where: the optimization variable is the hierarchical fusion topology structure; the objective function is the communication and computing resource consumption of the intelligent agent; the constraints include the positioning accuracy requirements and communication and computing resource constraints of the intelligent agent; 2.2) Adopting a hierarchical structure to construct the search problem, the hierarchical fusion topology is split into multiple layers, which are described and modeled separately. The positioning accuracy after each layer is taken as the state S k , taking each layer as action A k , the corresponding navigation resource expenditure is taken as k Execute A k The cost of reaching the next state is described by a series of alternating sequences of S and A, thus constructing a search problem; 2.3) Initialize the hierarchical fusion topology optimization problem and construct the initial state: set the adjacency matrix of the first-layer fusion topology to zero, indicating that at the beginning of a navigation epoch, there is no information exchange and fusion between agents; the corresponding positioning error is set to infinity, indicating that the error of the agent navigation solution is infinite, that is, no effective navigation solution is generated; generate the initialized single-layer fusion topology with the help of the concept of the adjacency matrix; Step 3: The genetic algorithm search module encodes the current state and the initialized single-layer fusion topology, measures the quality of each offspring with fitness, generates offspring through genetic operations, constructs a genetic algorithm and searches, and outputs the search to obtain the optimal new layer of fusion topology, including: 3.1) Design coding and generate initial population: Use binary coding to randomly generate an initial population with binary coding as ; 3.2) Design the fitness function and the corresponding genetic operation. The fitness function f = (500P + M) - C S, where P represents the number of agents that meet the positioning accuracy requirements, M represents the amount of information used in the new layer of fusion topology, C = 10E9, and S represents whether any of the navigation resources exceeds the maximum limit. If it exceeds, it is 1, and the rest are 0. The genetic operation in this step is random mutation and exchange; 3.3) Determine the new layer of information fusion through genetic algorithm search: The genetic algorithm module outputs the optimal adjacency matrix of the new layer of information fusion topology as the information that the corresponding intelligent agent in the new layer needs to be newly fused. The correlation detection module determines the correlation between the candidate information and the original state estimation one by one. When there is information that does not meet the requirements, the corresponding elements of the adjacency matrix are set to zero. When the adjacency matrix is ​​not a zero matrix, it is recorded as a new action A. k+1 , the corresponding state is also advanced to S k+1 Otherwise, it returns to the previous state S k-1 Re-search, taking a different action A than before k ', when S k+1 If all constraints are met, the search result is successful and output; if it reaches the initial state and still needs to be rolled back, the search result is failed; Step 4: The correlation detection module checks whether the information to be fused is correlated with the original state estimation. The correlation mapping unit maps the hierarchical fusion topology. The information fusion process is displayed in parallel as multiple layers. The arrows between the layers represent the transmission of information between agents. If any agent sends multiple arrows and these arrows converge on the same agent, it means that the information corresponding to the agent is correlated. By removing this information, the updated hierarchical fusion topology can be obtained. Step 5: Repeat steps 3 and 4 to obtain the optimized hierarchical fusion topology structure.

Images