Unmanned aerial vehicle cluster multi-center data link relative navigation consistent initialization method
By constructing a multi-center relative navigation joint error state model and a unified auxiliary common coordinate system, the problems of member state unification and initial state determination in UAV swarms under dynamic networking were solved, realizing the establishment of state correlation and navigation continuity, and ensuring the accuracy of swarm navigation.
Patent Information
- Application Number
- CN202611534133.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-30
- Publication Date
- 2026-11-10
AI Technical Summary
In dynamic networking scenarios, it is difficult to unify the states of members in different reference coordinate systems within a drone swarm. The initial state of newly added members is difficult to determine accurately, and the state correlation between newly added members and existing members is difficult to establish effectively, leading to navigation errors and inconsistencies.
By constructing a multi-center relative navigation joint error state model, establishing a unified two-dimensional auxiliary common coordinate system, sharing navigation state information and covariance, and using dynamic network pseudo-measurement relationships for state updates, the state consistency initialization between new members and existing members is achieved.
It achieves uniformity of the status of each member during dynamic networking, ensures the accuracy and relevance of the initial status of newly added members, avoids navigation errors and inconsistencies, and maintains the continuity and accuracy of the cluster's relative navigation.
Smart Images

Figure CN123065475A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of cluster collaborative navigation and information fusion technology, specifically, it relates to a method for initializing relative navigation consistency of a multi-center data link in a UAV cluster. Background Technology
[0002] Relative navigation is a crucial technology for determining the positional relationships among members and enabling collaborative mission execution in unmanned aerial vehicle (UAV) swarms. During collaborative reconnaissance, area surveillance, and disaster relief missions, when a global navigation satellite system (GNSS) is available, each member can calculate their relative position by sharing their absolute navigation results and combining this with the positional relationships between members. However, in environments where satellite navigation is denied, such as those with electromagnetic interference or signal blockade, members struggle to continuously obtain reliable absolute position information. Therefore, relative navigation based on sharing absolute navigation results is insufficient to meet the autonomous operation requirements of the swarm.
[0003] To improve autonomous navigation capabilities in satellite navigation denied environments, existing technologies typically employ data link relative navigation methods that fuse inertial navigation information with inter-unit relative ranging information. These methods recursively deduce the motion states of members through the inertial navigation system and acquire ranging information via data link communication between swarm members. They then combine filtering estimation and multi-source information fusion algorithms to calculate the relative navigation states among members, thus maintaining the positional relationships within the UAV swarm. Furthermore, to enhance relative navigation estimation capabilities, some methods share member state information and relative measurement information, enabling each member to collaboratively estimate using the observations of other members, and maintain the error correlation between members through joint state modeling.
[0004] As the size of UAV swarms expands and the mission area and functional division of labor dynamically change, the swarm typically consists of multiple local operational units. Each operational unit performs relative navigation in its own reference coordinate system, and multiple nodes maintain the local joint navigation state. When members join, leave, or operational units regroup during a mission, the swarm member relationships, communication topology, and state dimensions in the filters all change. It is necessary to initialize the new members and re-establish the navigation state association with the original members.
[0005] However, existing multi-center relative navigation methods typically rely on a fixed set of members and a fixed reference datum for state estimation, which still has shortcomings in dynamic networking scenarios. Firstly, different operational units usually establish relative coordinate systems using different reference members, leading to differences between local coordinate datums. When a new member is added, state information from different reference datums cannot be directly fused, easily causing coordinate inconsistencies during state stitching. Secondly, in satellite navigation denied environments, the inertial navigation position of a new member experiences cumulative drift. Directly using this position as the initial value for the new state will linearize distance measurements at state points deviating from the true position, resulting in significant errors and affecting the filtering initialization process. Furthermore, the state estimation of a new member depends on the existing member states and relative measurement information between members, exhibiting an error correlation with the original member states. Ignoring this correlation during initialization makes it difficult to maintain consistency in the joint filtering state.
[0006] In summary, existing technologies suffer from several drawbacks, including difficulty in unifying member states under different reference coordinates during dynamic networking, difficulty in accurately determining the initial state of newly added members, and difficulty in effectively establishing state correlations between newly added members and existing members. Summary of the Invention
[0007] To address the shortcomings of existing technologies, such as difficulty in unifying member states under different reference coordinates during dynamic networking, difficulty in accurately determining the initial state of newly added members, and difficulty in effectively establishing state correlations between newly added and existing members, the technical solution provided by this invention is as follows: A method for initializing relative navigation consistency in a multi-center data link of an unmanned aerial vehicle (UAV) swarm includes: The steps to establish a multi-center relative navigation joint error state model for each node in a UAV swarm are as follows: construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains; and establish the corresponding state covariance based on the correlation between the states of each node. The steps include detecting changes in the members of the drone swarm, determining new and off-network members based on the member set before and after the change, deleting the status information corresponding to off-network members, augmenting the status information corresponding to new members, and isolating the relative measurement information involving new members to form the temporary joint navigation state corresponding to the members to be initialized. The steps involve establishing a unified two-dimensional auxiliary common coordinate system based on preset common reference members, converting the horizontal position information of each node to the unified auxiliary common coordinate system, and sharing the navigation status information, auxiliary common coordinate estimation information and corresponding covariance information of each node through a data link to form shared data for initializing new members. The steps include: acquiring shared data corresponding to the new member and distance measurement information between members; after achieving height closure at each node using altimeters; establishing a two-dimensional position constraint relationship for the new member based on the shared data, the closed-loop height difference, and the distance measurement information; and determining the initial two-dimensional position correction amount and corresponding covariance of the new member relative to the common reference member based on the position constraint relationship. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic networking pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic networking pseudo-measurement relationship, and the cross-correlation between the status of the new member and the status of the original member is established. The steps include: releasing the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster relative navigation status update, and performing status prediction, measurement update, and error feedback correction based on the updated joint navigation status, and outputting the relative navigation results among the UAV cluster members.
[0008] Furthermore, in a preferred embodiment, the establishment of a multi-center relative navigation joint error state model for each node in the UAV cluster includes constructing a joint error state vector based on the navigation error state of each node and the navigation error states of other nodes maintained by the node, and constructing a state covariance matrix based on the error correlation between different node states. The state covariance matrix is used to characterize the error correlation between the navigation states of each node.
[0009] Furthermore, in a preferred embodiment, the detection of changes in UAV swarm members includes determining a set of newly added members and a set of members that have left the network based on the stable set of members before and after the change; deleting the corresponding state variables for the set of members that have left the network, while retaining the covariance relationship between the states of the remaining members; adding the corresponding state variables to the original joint navigation state for the set of newly added members, and setting a temporary state and a temporary covariance for the initialization process.
[0010] Furthermore, in a preferred embodiment, the establishment of a unified two-dimensional auxiliary common coordinate system includes using a preset common reference member as a coordinate reference benchmark, generating two-dimensional auxiliary common coordinate estimation information based on the horizontal positional relationship between the navigation status information of each node and the common reference member, and sending the auxiliary common coordinate estimation information and the corresponding covariance information to the member to be initialized via a data link.
[0011] Furthermore, in a preferred embodiment, determining the initial position correction of the new member relative to the common reference member includes establishing a distance constraint relationship based on the received shared data of multiple members and the corresponding distance measurement information, taking the two-dimensional position correction of the new member as a parameter to be determined, and under the positioning approximation with height closure and broadcast coordinates as given input, performing a Gauss-Newton iterative solution with a numerically stable term on the two-dimensional position correction based on the distance constraint relationship and the weight information corresponding to the TOA ranging error, and determining the initial position correction and the corresponding covariance based on the solution result.
[0012] Furthermore, in a preferred embodiment, the construction of dynamic network pseudo-measurement relationships includes constructing a self-machine pseudo-measurement relationship based on the initial position correction of the new member and the common reference member status information to constrain the positional relationship between the new member status and the common reference member status; and constructing a other-machine pseudo-measurement relationship based on the navigation status information shared by the original members to jointly update the original member status information and the new member status information.
[0013] A method for initializing relative navigation consistency in a multi-center data link of an unmanned aerial vehicle (UAV) swarm includes: Establish a multi-center relative navigation joint error state model for each node in a UAV cluster. Construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains. Establish a corresponding state covariance module based on the correlation between the states of each node. The module detects changes in the members of the drone swarm, determines the new and off-network members based on the member set before and after the change, deletes the status information corresponding to the off-network members, augments the status information corresponding to the new members, and isolates the relative measurement information involving the new members to form a module for the temporary joint navigation status of the members to be initialized. A unified two-dimensional auxiliary common coordinate system is established based on preset common reference members. The horizontal position information of each node is converted to the unified auxiliary common coordinate system. The navigation status information, auxiliary common coordinate estimation information and corresponding covariance information of each node are shared through the data link to form a module for shared data initialization of new members. The module acquires shared data corresponding to the new member and distance measurement information between members. After the height closure is achieved at each node using an altimeter, a two-dimensional position constraint relationship for the new member is established based on the shared data, the closed-loop height difference, and the distance measurement information. Based on the position constraint relationship, the module determines the initial two-dimensional position correction amount and corresponding covariance of the new member relative to the common reference member. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic network pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic network pseudo-measurement relationship, and a module for establishing the cross-correlation between the status of the new member and the status of the original member is established. This module releases the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster's relative navigation status update. Based on the updated joint navigation status, it performs status prediction, measurement updates, and error feedback correction, and outputs the relative navigation results among the UAV cluster members.
[0014] A computer storage medium for storing a computer program, which, when read by the computer, is executed by the computer using the method described.
[0015] A computer, including a processor and a storage medium, executes the method when the processor reads a computer program stored in the storage medium.
[0016] A computer program product, which, as a computer program, implements the method when the computer program is executed.
[0017] Compared with the prior art, the advantages of the technical solution provided by the present invention are as follows: This invention achieves consistent initialization of the relative navigation state of multiple centers during dynamic networking by constructing a unified auxiliary public coordinate system, determining the initial relative position of new members based on data link shared information, and establishing the state correlation between new members and existing members by combining dynamic networking pseudo-measurement updates. This solves the problems of difficulty in unifying state information under different reference coordinates, inaccurate initial state of new members, and difficulty in establishing state correlation after the addition of new members.
[0018] This invention introduces auxiliary common coordinates that include common reference member errors, transforms the positions of members in different local clusters to a unified coordinate definition, and uses common reference member information to construct positional relationships between members, so that the relative positional states under different reference bases can be expressed in the same coordinate system, avoiding the problem of inability to directly fuse states due to differences in local coordinate bases.
[0019] This invention constructs positional constraints based on shared navigation state via data link, auxiliary common coordinate estimation and corresponding covariance, and combined with TOA measurements of multiple members. It uses an iterative weighted least squares method to solve for the position correction of newly added members, enabling them to obtain an initial position relative to a unified reference coordinate based on the measurement information of existing members. This avoids the measurement linearization bias caused by directly using drifting inertial navigation positions as initial state values.
[0020] This invention sets up dynamic network pseudo-measurements during the initialization process of adding new members, and uses self-machine pseudo-measurements and other-machine pseudo-measurements to jointly update the position status of the new members with the status of the original members. This enables the state estimation of the new members to establish cross-correlation with the navigation status of the original members, avoiding the problem of inconsistent joint state estimation caused by ignoring the error correlation between members.
[0021] This invention adjusts the filter state dimension according to changes in cluster members, performs state pruning on off-network members, performs augmented modeling on on-network members, and isolates ordinary measurements involving uninitialized members during initialization. This enables the filter state structure to adapt to changes in members during dynamic networking, while avoiding the influence of uninitialized member states on the relative navigation estimation of existing members.
[0022] This invention enables the member states after dynamic networking to continue participating in the relative navigation recursion process by removing measurement isolation after initialization and continuing to perform time updates, measurement updates, and error feedback corrections based on the adjusted member set, thus maintaining the continuous update of the cluster's relative navigation state. Attached Figure Description
[0023] Figure 1 This is a schematic diagram of the dynamic network trajectory for relative navigation of the cluster.
[0024] Figure 2 The overall flowchart for initializing relative navigation consistency in a multi-center data link is shown.
[0025] Figure 3 This is a graph showing the average, maximum, and minimum values of the relative position error of cluster members when using the initialization method of this invention.
[0026] Figure 4 The graph shows the average, maximum, and minimum relative position errors of cluster members under the comparative condition of using the pure inertial navigation divergence position as the initial value of the new state.
[0027] Figure 5 This is a graph showing the average, maximum, and minimum values of the relative distance error between cluster members when using the initialization method of this invention.
[0028] Figure 6 The graph shows the average, maximum, and minimum values of the relative distance error between cluster members under the comparative condition of using the pure inertial navigation divergence position as the initial value of the new state. Detailed Implementation
[0029] To make the advantages and benefits of the technical solution provided by the present invention clearer, the technical solution provided by the present invention will now be described in further detail with reference to the accompanying drawings, specifically: Implementation Method 1: This implementation method provides a method for initializing relative navigation consistency in a multi-center data link of an unmanned aerial vehicle (UAV) swarm, including: The steps to establish a multi-center relative navigation joint error state model for each node in a UAV swarm are as follows: construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains; and establish the corresponding state covariance based on the correlation between the states of each node. The steps include detecting changes in the members of the drone swarm, determining new and off-network members based on the member set before and after the change, deleting the status information corresponding to off-network members, augmenting the status information corresponding to new members, and isolating the relative measurement information involving new members to form the temporary joint navigation state corresponding to the members to be initialized. The steps involve establishing a unified two-dimensional auxiliary common coordinate system based on preset common reference members, converting the horizontal position information of each node to the unified auxiliary common coordinate system, and sharing the navigation status information, auxiliary common coordinate estimation information and corresponding covariance information of each node through a data link to form shared data for initializing new members. The steps include: acquiring shared data corresponding to the new member and distance measurement information between members; after achieving height closure at each node using altimeters; establishing a two-dimensional position constraint relationship for the new member based on the shared data, the closed-loop height difference, and the distance measurement information; and determining the initial two-dimensional position correction amount and corresponding covariance of the new member relative to the common reference member based on the position constraint relationship. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic networking pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic networking pseudo-measurement relationship, and the cross-correlation between the status of the new member and the status of the original member is established. The steps include: releasing the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster relative navigation status update, and performing status prediction, measurement update, and error feedback correction based on the updated joint navigation status, and outputting the relative navigation results among the UAV cluster members.
[0030] The establishment of a multi-center relative navigation joint error state model for each node in the UAV cluster includes constructing a joint error state vector based on the navigation error state of each node and the navigation error states of other nodes it maintains, and constructing a state covariance matrix based on the error correlation between different node states. The state covariance matrix is used to characterize the error correlation between the navigation states of each node.
[0031] The detection of changes in the drone swarm members includes determining the set of newly added members and the set of members that have left the network based on the stable set of members before and after the change; deleting the corresponding state variables for the set of members that have left the network, while retaining the covariance relationship between the states of the remaining members; adding the corresponding state variables to the original joint navigation state for the set of newly added members, and setting temporary states and temporary covariances for the initialization process.
[0032] The establishment of a unified two-dimensional auxiliary common coordinate system includes using a preset common reference member as a coordinate reference benchmark, generating two-dimensional auxiliary common coordinate estimation information based on the horizontal positional relationship between the navigation status information of each node and the common reference member, and sending the auxiliary common coordinate estimation information and the corresponding covariance information to the member to be initialized via a data link.
[0033] The determination of the initial position correction of the new member relative to the common reference member includes establishing a distance constraint relationship based on the received shared data of multiple members and the corresponding distance measurement information, taking the two-dimensional position correction of the new member as a parameter to be determined, and under the positioning approximation with height closure and broadcast coordinates as given input, performing Gauss-Newton iterative solution with numerical stability term on the two-dimensional position correction based on the distance constraint relationship and the weight information corresponding to the TOA ranging error, and determining the initial position correction and the corresponding covariance based on the solution result.
[0034] The construction of dynamic network pseudo-measurement relationships includes constructing a pseudo-measurement relationship for the self based on the initial position correction of the new member and the state information of the common reference member, so as to constrain the positional relationship between the state of the new member and the state of the common reference member; and constructing a pseudo-measurement relationship for the other machine based on the navigation state information shared by the original members, so as to jointly update the state information of the original members and the state information of the new members.
[0035] A method for initializing relative navigation consistency in a multi-center data link of an unmanned aerial vehicle (UAV) swarm includes: Establish a multi-center relative navigation joint error state model for each node in a UAV cluster. Construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains. Establish a corresponding state covariance module based on the correlation between the states of each node. The module detects changes in the members of the drone swarm, determines the new and off-network members based on the member set before and after the change, deletes the status information corresponding to the off-network members, augments the status information corresponding to the new members, and isolates the relative measurement information involving the new members to form a module for the temporary joint navigation status of the members to be initialized. A unified two-dimensional auxiliary common coordinate system is established based on preset common reference members. The horizontal position information of each node is converted to the unified auxiliary common coordinate system. The navigation status information, auxiliary common coordinate estimation information and corresponding covariance information of each node are shared through the data link to form a module for shared data initialization of new members. The module acquires shared data corresponding to the new member and distance measurement information between members. After the height closure is achieved at each node using an altimeter, a two-dimensional position constraint relationship for the new member is established based on the shared data, the closed-loop height difference, and the distance measurement information. Based on the position constraint relationship, the module determines the initial two-dimensional position correction amount and corresponding covariance of the new member relative to the common reference member. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic network pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic network pseudo-measurement relationship, and a module for establishing the cross-correlation between the status of the new member and the status of the original member is established. This module releases the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster's relative navigation status update. Based on the updated joint navigation status, it performs status prediction, measurement updates, and error feedback correction, and outputs the relative navigation results among the UAV cluster members.
[0036] A computer storage medium for storing a computer program, which, when read by the computer, is executed by the computer using the method described.
[0037] A computer, including a processor and a storage medium, executes the method when the processor reads a computer program stored in the storage medium.
[0038] A computer program product, which, as a computer program, implements the method when the computer program is executed.
[0039] Implementation Method Two: This implementation method is a further detailed description of the technical solution provided in Implementation Method One, specifically: This embodiment provides a method for consistent initialization of relative navigation in a multi-center data link for unmanned aerial vehicle (UAV) swarms. In satellite navigation denied environments, when members join, leave, or work units reorganize within the UAV swarm, the method initializes the relative navigation state of newly added members, enabling them to obtain initial relative positions under a unified reference coordinate system and establish a consistent joint navigation state with existing members. The method first establishes a full-error state model among swarm members and adjusts the filter state according to changes in swarm members. Then, it shares navigation information among members using a unified auxiliary common coordinate system and determines the initial relative position of the newly added member using data link ranging information. Based on this, it constructs a dynamic network pseudo-measurement to jointly update the states of both newly added and existing members, establishing state correlations between members. After initialization, the newly added member is incorporated into the normal relative navigation recursion process, continuously outputting relative navigation results among swarm members.
[0040] Specifically, this embodiment includes the following processing steps: First, a multi-center relative navigation full-error state model is established for each node in the cluster. Each node constructs a joint error state vector containing state information of multiple members based on its own navigation error state and the navigation error states of other members it maintains, and establishes a corresponding covariance matrix based on the state correlations between members. The joint error state vector describes the position error, velocity error, attitude error, and time synchronization related errors of each member, while the covariance matrix records the correlations between the state estimation errors of different members. Based on the established joint error state model, state propagation is performed using inertial navigation information to obtain the predicted state of each member at the current moment, providing a state basis for subsequent dynamic network initialization.
[0041] When a change in cluster membership is detected, newly added and offline members are determined based on the member sets before and after the change. For offline members, the corresponding state variables are deleted from the current joint state, and the covariance matrix is adjusted synchronously, while preserving the state relationships between the remaining members. For newly added members, the corresponding state variables are added to the original joint state, and a temporary state and covariance used during initialization are set for the new member. Subsequently, ordinary relative measurements involving newly added members are isolated, ensuring that the state of newly added members that has not yet been initialized does not directly participate in the original relative navigation filtering process.
[0042] During the initialization process of adding new members, a unified two-dimensional auxiliary common coordinate system is established between different local clusters. Each member selects a common reference member, and based on the positional error of the common reference member and the member's own positional relationships, under the condition that each member is equipped with an altimeter and completes altitude closure, the member's horizontal position is transformed to be expressed in the unified reference coordinate system. Each member shares its own navigation status information, auxiliary common coordinate estimation information, and corresponding covariance through data link communication according to a preset communication sequence. At the same time, it shares member identifiers, common reference member identifiers, and time information, so that the member to be initialized can obtain the unified coordinate information and corresponding uncertainty description of the original cluster members.
[0043] After receiving data packets from existing cluster members, the member to be initialized matches the received data based on its member identifier, common reference member identifier, and time information to form a data set corresponding to each broadcast member. The member to be initialized further acquires arrival time measurement information corresponding to each broadcast member and, based on the measurement information, the position status of the broadcast members, and the known height difference after loop closure, writes the three-dimensional slant range information into a distance constraint relationship for the two-dimensional position. Within the initialization time window, the two-dimensional position correction amount of the member to be initialized relative to the common reference member is used as a parameter to be determined, and a position correction calculation model is established based on multiple sets of distance constraint relationships.
[0044] Based on the aforementioned position correction calculation model, the converged broadcast coordinates are used as given input. Under the positioning approximation where the residual influence of the height closure loop is negligible, the measurement weights are determined solely based on the TOA ranging error, and the two-dimensional position correction for the members to be initialized is iteratively calculated. During the iteration process, small positive quantities are added to the normal equation matrix to improve numerical stability. The two-dimensional correction is updated based on the residual between the predicted distance and the actual ranging until the preset convergence condition is met. If convergence is not achieved after reaching the upper limit of the number of iterations, the positioning result is not accepted. After effective convergence, the initial two-dimensional position correction and its two-dimensional covariance are output for use in dynamic network pseudo-measurement updates; the astronomical correction is not solved for or spliced separately.
[0045] After obtaining the initial relative positions of the members to be initialized, a dynamic network pseudo-measurement relationship is constructed based on the initial relative position estimation results and auxiliary common coordinate information. First, a self-machine pseudo-measurement is constructed based on the position correction results of the member to be initialized, and the position state of the new member is associated with the state of the common reference member, so that the state of the new member can be mapped to the original navigation state system. Then, a third-machine pseudo-measurement is constructed based on the state information broadcast by the original members and the new members, and the state information of the original members is propagated to the filtered state of the new members, while the state information of the new members is fed back to the filtered state of the original members.
[0046] Based on the constructed self-machine pseudo-measurement and other-machine pseudo-measurement, the joint error state is updated, and the state estimate and covariance matrix are updated simultaneously. During the update process, the mutual covariance between members is calculated according to the information transmission relationship between the new member state and the original member state, so that the new member state and the original member state establish a statistical correlation, thereby forming a joint state consistent with the original multi-center relative navigation filter structure.
[0047] Once a new member has completed valid initialization and related members have completed their state updates, the ordinary measurement isolation between the new member and existing members is lifted based on the initialization completion status, allowing the new member to participate in subsequent relative navigation measurement updates. For members that have not completed initialization, the initialization process continues.
[0048] After dynamic network initialization, joint error state prediction and measurement updates are continuously performed based on the updated member set. Each member uses inertial navigation information, data link ranging information, and other auxiliary information to recursively update the joint state, and feeds back the estimated navigation error to the inertial navigation state to correct the member's position, velocity, and attitude information. Based on the corrected navigation state, the relative positional relationships between cluster members are calculated, and the relative navigation results of the UAV cluster are output.
[0049] In a preferred embodiment, the auxiliary common coordinates include position error information of common reference members, enabling the horizontal position difference between different members to be calculated under a unified reference relationship. During initialization, by sharing the estimated value of the auxiliary common coordinates and its covariance, the members to be initialized complete the two-dimensional position solution, and provide positional relationship and uncertainty information for subsequent pseudo-measurement updates.
[0050] In a preferred embodiment, the arrival time measurement information can be obtained from the distance measurement results between members generated during the data link communication process. The member to be initialized determines its initial relative position based on the measurement information of multiple members, so as to avoid insufficient position constraints caused by a single measurement relationship.
[0051] In a preferred embodiment, the dynamic networking pseudo-measurement update process adopts a complete covariance update method to synchronously maintain the state correlation between the new member and the original member, so that the joint navigation state after the new member joins can continue to participate in subsequent relative navigation calculations.
[0052] Implementation Method 3: This implementation method is described in detail with reference to the accompanying drawings. Specific embodiments are provided to further illustrate the technical solutions offered above. Specifically: To overcome the limitation that existing multi-center data link relative navigation is only applicable to fixed formation scenarios, this implementation proposes a consistent initialization method for UAV swarm multi-center data link relative navigation. The initial relative position of newly added members is determined under a unified reference coordinate system and updated using pseudo-measurements (i.e., equivalent measurements constructed from positioning results or shared states). The initialization results are mapped to the filters of each member, and the covariance of the newly added state and its cross-covariance with the original state are established synchronously, providing initial conditions for the recursive relative navigation after dynamic networking.
[0053] The method includes the following steps: Step 1: Establish the full error state equation for multi-center relative navigation for each node in the cluster; Each node jointly models the position, velocity, attitude, and clock errors of itself and other members according to member number, maintains state estimation and complete covariance, and establishes discrete propagation equations:
[0054] in, For the node that performs filtering, For the filtering epoch, and These are the state transition matrix and the noise driving matrix, respectively. For covariance equals Process noise.
[0055] Step 2: Identify changes in cluster networking in real time and design a variable-dimensional relative navigation filter; Each node compares the set of stable member IDs before and after the network topology change to identify newly joined and disconnected members. For disconnected members, a matrix is constructed to select the retained states. Execution status and covariance reduction:
[0056]
[0057] The left side of the equation represents the estimated state and covariance of the retained state, respectively. When deleting an off-network state, the cross-covariance among the remaining members is retained. For a new member, the augmented temporary state and covariance are:
[0058] in, and To preserve the state and its covariance, This is the temporary positive definite diagonal covariance for newly added states. Isolation involves normal TOA updates for uninitialized members; other valid recursions continue. No re-initialization occurs only when a member leaves the network; if both members join and leave simultaneously, deletion is performed first, followed by augmentation.
[0059] Step 3: Unify auxiliary common coordinates, and share navigation nominal values, auxiliary common coordinate estimates and their covariance in preset time slots based on the time division multiple access round-robin broadcast mechanism; Each node is equipped with an altimeter and stabilizes its altitude through an altitude-assisted closed-loop system, providing a common reference for members. Establish a two-dimensional auxiliary common coordinate system. Let... Earth-fixed coordinate system For a common navigation coordinate system, the subscript is... If only the East and North components are represented, then the members The auxiliary common coordinates and the nominal horizontal position of the inertial navigation system are:
[0060] , These are the actual and inertial navigation fixed Earth positions, respectively. This is the coordinate transformation matrix. The two-dimensional auxiliary common coordinates do not include the altitude to be estimated. Geographic location error and latitude / longitude error are denoted as follows:
[0061] First pass the latitude and longitude error Converted to local metric error, then incorporated into the common navigation system through coordinate transformation; the corresponding coefficients are:
[0062]
[0063] , The radii of curvature of the east-west and north-south meridians at the reference member location. Two The submatrix corresponds to the latitude and longitude error column; the two-dimensional position equation uses its east and north rows:
[0064]
[0065] in, It is a two-dimensional position correction amount; the approximation omits the influence of the remaining height during the positioning stage. , Record separately The first two rows and the third column of the first two rows; , Similarly, each node broadcasts the nominal navigation value, closed-loop altitude, and estimated two-dimensional auxiliary common coordinates according to preset time slots. and covariance They also share member ID, reference ID, validity flag, and timestamp.
[0066] Step 4: The members to be initialized collect data packet information, accumulate TOA measurements, and calculate the initial relative position using the iterative weighted least squares method; Nodes to be initialized The broadcast information is paired with the TOA measurements of the corresponding epochs to complete clock compensation and metric distance conversion. Using the closed-loop height difference as a known parameter, the three-dimensional slant range observation is transformed into a positioning problem with only the two-dimensional position correction as unknown. Within a short window, the correction is approximated as constant, and the zero two-dimensional correction corresponding to the inertial navigation position is used as the initial value to perform Gauss-Newton iterations.
[0067] in, For the number of iterations, Let be the distance residual vector. For Jacobian matrices, Under this approximate positioning, It only includes the TOA ranging error covariance. Is with Small quantities with the same dimensions are stable. It is a second-order identity matrix; this small quantity is not considered as new positioning information.
[0068] When the level information matrix is fully rank and has converged effectively without stabilizing terms, calculate the local covariance of the two-dimensional correction:
[0069] This represents the final effective iteration count. (In the output epoch) Directly generate two-dimensional positioning results:
[0070] Positioning corrections, auxiliary common coordinates, and their covariances are all kept in two dimensions and are not combined with the height channel. Measurements continue to be collected when convergence is not achieved or geometric information is insufficient; valid results are used for dynamic network pseudo-measurement updates.
[0071] Step 5: Construct dynamic network pseudo-measurement equations, execute pseudo-measurement updates for self-machines and pseudo-measurement updates for other machines, and realize the propagation of the original cluster state and the establishment of correlation with members to be initialized.
[0072] Pseudo-measurements for both the self-machine and other-machine systems are constructed using 2D positioning correction values and 2D broadcast coordinates, respectively. Height is not estimated during the positioning phase; during the pseudo-measurement phase, the sensitivity of the 2D measurement to the remaining height error is included in the covariance. New nodes are added. The self-propelled pseudo-measurement is:
[0073] This is for the correction of errors in two-dimensional positioning. Node The modeled height error is retained in the state term; the unmodeled height error of the reference member is incorporated into the noise. Under the approximation where the two errors are uncorrelated:
[0074] The remaining height error variance for the reference member. Receiving node. For broadcast members Construct a pseudo-measurement from another machine; when , and hour:
[0075] This represents the two-dimensional broadcast coordinate error. The noise covariance is formed by adding the two-dimensional coordinate covariance and the covariance of the unmodeled height term, preserving... Dimensionality. When a reference member acts as a receiving node, its modeled height term is retained in the state; the auxiliary coordinates of the reference member itself are not used as independent position pseudo-measurements.
[0076] Newly added nodes first receive the original cluster position state, then perform their own update, and broadcast the initialization results to the original cluster. Under the condition that pseudo-measurement noise is uncorrelated with prior estimation error, Kalman update is performed according to the measurement model and noise covariance, and the complete covariance is updated in Joseph form to establish the cross-covariance between the old and new states.
[0077] Step 6: After the dynamic networking is completed, release the isolation between the members to be initialized and the original cluster for ordinary measurements.
[0078] After a newly added node completes effective positioning and self-update, it broadcasts the results. Receiving nodes set a modeling completion flag after completing updates on other machines. Ordinary measurement isolation is lifted based on the completion status; those not yet completed continue initialization. Only when a member leaves the network, relevant statuses and measurement caches are cleared before recursion continues.
[0079] Step 7: Construct time update and measurement update equations; Based on the adjusted membership set, continuously predict the joint error state and its covariance:
[0080]
[0081] The left and right sides of the conditional notation represent the estimated epoch and the cutoff epoch of the utilized measurements, respectively; the two equations predict the state and covariance, respectively. Time updates are performed throughout the initialization process; after receiving the allowed RTT, altimeter, direct TOA, and shared TOA measurements, the corresponding measurement updates are performed.
[0082] Step 8: Perform error feedback correction, use the corrected absolute state information to calculate the optimal relative state estimate between cluster nodes, and output the relative navigation results.
[0083] Each node feeds back the navigation error estimated by the overall relative navigation filter to the nominal value of the inertial navigation system, where latitude and longitude correction is performed using:
[0084] The left side shows the corrected position, and the right side shows the difference between the nominal inertial navigation position and the estimation error. Velocity and attitude are corrected according to the corresponding error definitions, and the navigation estimation result is output by the overall relative navigation filter and continues to be calculated recursively.
[0085] Example 1: Initialization method for relative navigation consistency of multi-center data link in UAV swarm for node network entry scenario This embodiment considers that multiple nodes in the original cluster have already completed intra-cluster relative navigation, and the nodes to be initialized... The original cluster is added during task execution. The original cluster member set is denoted as... Public reference members are recorded as ,and The network access process in this embodiment maintains common reference members. constant.
[0086] Each node is equipped with an inertial measurement unit, a data link device with communication and ranging capabilities, and an altimeter. The altimeter measurement is used for altitude-assisted closed-loop control, enabling the initialization to transform three-dimensional slant range information into a two-dimensional horizontal position problem. The data link employs a round-robin broadcast mechanism, with each node maintaining its own navigation error state and that of the other nodes it models. No new azimuth correction is added during initialization; the altitude-related states in the conventional navigation filter are retained according to their original structure.
[0087] In the following text, This indicates the node that performs the filtering calculation. Represents any actual member number. This indicates that other members are in the node. The local state index in the filter is mapped to the local state index through a numbering system. correspond; Indicates the filtering epoch. This indicates the sequence number within the initialization window. This indicates the number of Gauss-Newton iterations.
[0088] Step 1: Establish a multi-center relative navigation full error state model (1.1) Joint error state and covariance node The joint error state is defined as
[0089] in, For nodes Its own 11-dimensional navigation error state, The model for it The 9-dimensional navigation error state of each of the other members, The number of other members being modeled.
[0090] Its own state is taken as
[0091] Other member states are set to
[0092] in, , and These are the errors in longitude, latitude, and altitude, respectively. For the redefined velocity error; This refers to the angle of attitude misalignment. and These represent the clock phase and frequency errors of the JTIDS receiver, respectively.
[0093] In this embodiment, latitude and longitude errors and attitude misalignment angles are expressed in radians, altitude and position are expressed in meters, speed is expressed in meters per second, clock deviation is expressed in seconds, and clock deviation change rate is expressed in seconds per second.
[0094] The covariance maintained by the filter is defined as
[0095] The off-diagonal sub-blocks between different members are used to describe the cross-correlation of member state estimation errors and are not omitted due to member block modeling.
[0096] (1.2) Definition of speed error The speed error in this embodiment is defined as follows:
[0097] in, The direction cosine matrix from the real machine system to the navigation system. The cosine matrix of the navigation system direction indicated by the inertial navigation system. and These are the inertial navigation indicated speed and the actual speed, respectively.
[0098] This formula is used to define the error state; when the algorithm is executed, the nominal state is used to build the propagation model, and the real state is not input.
[0099] (1.3) Discrete-time propagation model node The joint error state satisfies
[0100] in, The joint state transition matrix, For discrete noise driving matrix, For covariance equals Discrete process noise.
[0101] The joint state transition matrix is constructed by its members:
[0102] Step 2: Identify network entry events and augment filter state (2.1) Identify changes in the membership set Let the sets of stable members before and after the change be respectively and The sets of newly added members and the sets of members who have left the network are respectively
[0103] In this embodiment, the original cluster node detects a new member. Add it to the modeling member table and record the common reference member number used in this initialization. .
[0104] (2.2) Set temporary augmentation state Arrange the retained states first and the newly added states last. Write the temporary augmented states and covariance as follows:
[0105] in, and To preserve the original estimates and covariance of the members, The temporary covariance is set according to the initial uncertainty of the new member.
[0106] The temporary covariance corresponding to the newly added state is taken as a positive definite diagonal matrix, which is used for temporary storage and calculation during initialization.
[0107] (2.3) Enable ordinary measurement isolation New members The modeling completion flag is set to incomplete. Any direct TOA or shared TOA measurement involving members that have not yet been initialized will not be updated in the normal relative navigation measurement update.
[0108] During isolation, inertial propagation, high-level assistance, clock synchronization, and measurement updates within the original cluster that do not involve uninitialized members continue to operate.
[0109] Step 3: Unify auxiliary common coordinates and share initialization data (3.1) Coordinate and position errors by Representing the Earth-fixed coordinate system, with Member The local navigation coordinate system, with Indicates public reference member The local navigation coordinate system. All nodes use consistent coordinate transformation parameters within the initialization window. The nominal and true positions of the inertial navigation system satisfy:
[0110] Geographic location error and its latitude and longitude error subvectors are defined as follows:
[0111] superscript This is used only to identify latitude and longitude error subvectors, excluding altitude errors. The transformation matrix from latitude, longitude, and altitude errors to local metric position errors is:
[0112] , These represent the radii of curvature for the tropospheric and meridional circles, respectively, and the matrix coefficients are calculated from the corresponding nominal navigation values. The nominal values for position and velocity still use (~), while the state error subvector does not use this notation.
[0113] (3.2) Auxiliary common coordinates and associated error quantities After the altimeter is closed-loop, initialization only processes the two-dimensional horizontal position. The following... This indicates that the east and north components of the common navigation system are used, without introducing additional state or projection matrices. The two-dimensional auxiliary common coordinates and the nominal horizontal position of the inertial navigation system are uniformly defined as follows:
[0114] Both vectors are metric two-dimensional quantities. Define the error coefficient matrix and its latitude and longitude error columns:
[0115]
[0116] Subscript The two rows used to identify the east and north of the coefficient matrix above are actually used in the two-dimensional equation:
[0117] The two-dimensional position correction is defined as:
[0118] This approximation ignores the influence of residual altitude in the positioning calculation. To separately represent altitude uncertainty in pseudo-measurements, the two-dimensional row coefficients are further written as follows, with latitude and longitude columns and altitude column:
[0119] , for , , for , , for When the public navigation system is a member When using a local navigation system, ,and The corresponding height term contributes zero to the horizontal pseudo-measurement.
[0120] The estimation errors and covariance of auxiliary common coordinates and position corrections are denoted as:
[0121] Both of the above error vectors are two-dimensional. and All The difference in auxiliary common coordinates among members satisfies:
[0122] Therefore, step four uses the two-dimensional coordinate difference and the known closed-loop height difference to construct positioning constraints, and step five uses the same coordinate and error convention to construct two-dimensional pseudo-measurement.
[0123] (3.3) Broadcast initialization information Original cluster members broadcast their own navigation nominal values, closed-loop altitudes, and covariances according to preset time slots, while simultaneously sharing common reference member IDs and two-dimensional auxiliary common coordinate estimates. Two-dimensional covariance The data includes valid coordinate markers and timestamps. Receiving nodes cache information based on member number, common reference number, and epochal correspondence. The data shared by all nodes is shown in the table below.
[0124]
[0125] Step 4: Perform Gauss-Newton iterations using asynchronous TOA measurements to determine the initial relative position. Nodes to be initialized Collect raw cluster broadcast information and corresponding TOA measurements. After receiving all the information required in this round, form a valid data set according to epochs. Assume that it contains... Measurement, No. The broadcast members of the measurement are The corresponding epoch is The TOA was converted to metric distance after clock compensation. The known height difference is calculated using closed-loop height estimation. Under the local navigation approximation, the three-dimensional slant range observation is transformed into a distance model that depends only on the two-dimensional correction:
[0126] The TOA ranging error includes clock compensation errors that have not yet been eliminated. This positioning model uses the converged broadcast coordinates as a given input and ignores the residual effects after altitude closure; therefore, the additive error only retains... The height difference is a known parameter and is not considered a state to be estimated. Within the short window, the two-dimensional correction is approximated as a constant, and each measurement still uses the inertial navigation horizontal position, broadcast two-dimensional coordinates, and closed-loop height of its respective epoch.
[0127] The inertial navigation position is used as the initial value, i.e. . No. In the next iteration, the two-dimensional position difference and predicted distance are calculated:
[0128] The Jacobian matrix corresponding to the distance residual and the two-dimensional position correction is:
[0129] Sequencing numbers will be used for and Stacked as , This embodiment uses an uncorrelated approximation for different TOA ranging errors, and its error covariance and weight matrix are as follows:
[0130] in, For the first The variance of the TOA ranging error. The positioning weight matrix no longer includes broadcast coordinate error terms or zenith correction error terms. Gauss-Newton updates with numerically stable small quantities are used.
[0131] in, Used to improve the numerical conditions of the normal equation matrix, with the same dimensions as the information matrix; It is a second-order identity matrix. This small quantity only performs stable iterative calculations, does not supplement the actual distance measurement information, and does not replace geometric validity checks.
[0132] After each update, the distance residual and Jacobian matrix are recalculated. Stop iteration when This is the metric convergence threshold. If convergence is not achieved after reaching the maximum number of iterations, or if the rank of the information matrix without a stabilizing term is insufficient, the current result will not be used, and the system will remain in the uninitialized state.
[0133] Let the final effective number of iterations be . The local covariance of the two-dimensional positioning correction is calculated from the effective ranging information:
[0134] This covariance is a local first-order approximation when the broadcast coordinates and closed-loop height are given as inputs; numerically stable small quantities are not included in the covariance as prior information. It directly outputs the two-dimensional correction and auxiliary common coordinates.
[0135] in, To output the epoch, It is a two-dimensional auxiliary common coordinate estimation covariance. , ,correspond The covariance, common reference member number, and epoch are passed to step five without increasing the celestial correction or the level-celestial cross-covariance block.
[0136] Step 5: Construct pseudo-measurements for the self and other machines, and update the initialization state and joint covariance. This step uses the two-dimensional auxiliary common coordinates, two-dimensional position correction, and matrix partitioning defined in step three. The positioning output is always two-dimensional; to retain the handling method for the unmodeled height error in the original pseudo-measurement model, only the covariance after horizontal sensitivity transformation is added to the two-dimensional positioning covariance, without expanding the dimension of the position correction. The following data all use coordinates, nominal values, and corresponding covariances of the same epoch.
[0137] First, nodes Constructing a self-generated pseudo-measurement using two-dimensional corrections:
[0138] The altitude error of this machine remains within the original 11-dimensional navigation state. This only describes its impact on the 2D position pseudo-measurement, without adding or solving for the astronomical position correction. When the unmodeled altitude error of the reference member and the 2D positioning error are treated as uncorrelated approximations, the auto-device pseudo-measurement covariance is:
[0139] Reference member remaining height error variance. Auto-machine pseudo-measurement matrix. Enter the latitude and longitude error column for this machine. In the members Enter the latitude and longitude error column Set the remaining columns to zero.
[0140] Subsequently, the receiving node Utilizing members The broadcast uses two-dimensional auxiliary public coordinates to construct pseudo-measurements for other machines. When , and hour:
[0141] member and The height error is not in the 9-dimensional other machine state block of the receiving node, and therefore is treated as residual height noise. Assuming that the height error and broadcast coordinate error are uncorrelated, the 2D other machine pseudo-measurement covariance is:
[0142] in, Other machine pseudo-measurement matrix according to Determine the status column, in the members Enter the latitude and longitude error column In the members Enter the latitude and longitude error column All other columns are set to zero. All covariances above are sums of two-dimensional covariances, without three-dimensional block stitching.
[0143] when At that time, the machine height has been modeled, and Retain it in the local height error column and remove it from the noise; if the two-dimensional coefficient is zero, its covariance contribution is also zero. When At that time, the difference between the broadcast auxiliary coordinates and the nominal horizontal position of the inertial navigation system is an identity constraint and is not regarded as an independent position pseudo-measurement.
[0144] Other machine pseudo-measurement is performed in two directions: node take over The broadcast transmits the original cluster positional relationships into its own filter; nodes After completing the self-update, broadcast the two-dimensional result. and The original cluster nodes are Update its newly added status.
[0145] For each update, the corresponding measurement, measurement matrix, and noise covariance are denoted as follows: , and Under the condition that pseudo-measurement noise is unrelated to the current prior estimation error, uniform execution is performed.
[0146] Update the complete joint covariance using the Joseph form:
[0147] superscript " "and" "Represents before and after the update," The identity matrix has the same dimension as the joint state. Each pseudo-measurement and its noise covariance are two-dimensional; the Kalman update still operates on the complete joint state and covariance, simultaneously maintaining the cross-covariance between the old and new states, without artificially setting the updated cross-covariance to zero. The pseudo-measurement noise being uncorrelated with prior estimation errors is a condition for the above update to be applicable. After propagation is complete, the initialization completion determination is performed.
[0148] For cases where multiple members or a local cluster are merged, the above location and pseudo-measurement process is performed on each member to be initialized, and their respective measurement sources and modeling flags are maintained; the effective internal covariance of the original local cluster should not be cleared to zero simply because of state reordering.
[0149] Step 6: Determine if initialization is complete and remove isolation. node After completing the effective least squares solution and dynamic network pseudo-measurement update, the initialization results can be broadcast; after the relevant receiving nodes complete their corresponding pseudo-measurement updates, they will then broadcast their own results to the nodes. The modeling flag is set to complete to avoid mutual waiting caused by waiting for the entire network to complete before broadcasting.
[0150] Step 7: Perform regular time updates and measurement updates (7.1) Time update Each node executes based on the current set of members.
[0151]
[0152] Time updates are performed continuously during measurement acquisition and initialization.
[0153] (7.2) Conventional measurement model Altitude measurement adopts
[0154] The relative clock difference measurement obtained from RTT clock synchronization processing is adopted
[0155] For members and The TOA distance between them was measured using...
[0156] Shared TOA measurements use the same physical model, only changing the member number and data source corresponding to the measurement.
[0157] Step 8: Feedback correction and output of relative navigation results Each node uses the navigation error estimated by its local relative navigation filter for feedback correction of the corresponding nominal inertial navigation value. Position correction employs...
[0158] The velocity and attitude feedback should be consistent with the redefined velocity and attitude errors used in step one. After correction, the overall relative navigation filter outputs navigation estimation results based on the current modeling member set and continues normal recursion. This step does not repeat the initial position solution or dynamic network pseudo-measurement update.
[0159] Example 2: A Multi-Center Data Link Relative Navigation Processing Method for UAV Swarms in Off-Network Scenarios This embodiment uses the same sensor configuration, error state definition, and conventional navigation recursive model as Embodiment 1. The difference is that after some members in the current cooperative cluster leave, the state of these members is no longer retained, and it is not necessary to repeat position initialization for members that have already been retained and are valid.
[0160] Step 1: Determine off-network members and retained members Determine based on changes in the stable membership set
[0161]
[0162] Generate a state preservation matrix based on the unique member ID. The matrix selects only the state components of the retained members and arranges them according to the new modeling order.
[0163] Step 2: Delete offline states and retain edge distribution For nodes State estimation and covariance execution
[0164]
[0165] The operation synchronously deletes the state blocks and covariance rows and columns corresponding to the off-network members, while retaining the existing cross-covariance among the remaining members.
[0166] Simultaneously, the measurements to be merged, member number mappings, and initialization flags related to off-network members are deleted, and the state transition matrix and process noise model are reorganized according to the new member set.
[0167] Step 3: Restore information sharing and navigation recursion for the remaining members Each node updates its broadcast schedule table based on the reserved member set and sends the new member set, reference member number, and navigation information.
[0168] After the reference number and measurement object are matched, the time update shown in Example 1, as well as the altimeter, RTT, direct TOA and shared TOA measurement updates are performed on the remaining members.
[0169] Step 4: Output the relative navigation results under the new grouping Calculate the relative position and covariance based on the new reference members and the current retained state, and output the relative navigation results under the new grouping.
[0170] When a node leaves the network while other nodes join the network, the state deletion operation of this embodiment is performed on the left-off member first, and then the augmentation, positioning and pseudo-measurement update process in embodiment one is performed on the newly added member.
[0171] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for initializing relative navigation consistency in a multi-center data link of an unmanned aerial vehicle (UAV) swarm, characterized in that, include: The steps to establish a multi-center relative navigation joint error state model for each node in a UAV swarm are as follows: construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains; and establish the corresponding state covariance based on the correlation between the states of each node. The steps include detecting changes in the members of the drone swarm, determining new and off-network members based on the member set before and after the change, deleting the status information corresponding to off-network members, augmenting the status information corresponding to new members, and isolating the relative measurement information involving new members to form the temporary joint navigation state corresponding to the members to be initialized. The steps involve establishing a unified auxiliary common coordinate system based on preset common reference members, converting the navigation status information of each node to the unified auxiliary common coordinate system, and sharing the navigation status information, auxiliary common coordinate estimation information, and corresponding covariance information of each node through a data link to form shared data for initializing new members. The steps include: acquiring shared data and distance measurement information between new members; establishing positional constraints for new members based on the shared data and distance measurement information; and determining the initial positional correction and corresponding covariance of new members relative to the common reference members based on the positional constraints. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic networking pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic networking pseudo-measurement relationship, and the cross-correlation between the status of the new member and the status of the original member is established. The steps include: releasing the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster relative navigation status update, and performing status prediction, measurement update, and error feedback correction based on the updated joint navigation status, and outputting the relative navigation results among the UAV cluster members.
2. The method for initializing relative navigation consistency of a multi-center data link in a UAV swarm according to claim 1, characterized in that, The establishment of a multi-center relative navigation joint error state model for each node in the UAV cluster includes constructing a joint error state vector based on the navigation error state of each node and the navigation error states of other nodes it maintains, and constructing a state covariance matrix based on the error correlation between different node states. The state covariance matrix is used to characterize the error correlation between the navigation states of each node.
3. The method for initializing relative navigation consistency of a multi-center data link in a UAV swarm according to claim 1, characterized in that, The detection of changes in the drone swarm members includes determining the set of newly added members and the set of members that have left the network based on the stable set of members before and after the change; deleting the corresponding state variables for the set of members that have left the network, while retaining the covariance relationship between the states of the remaining members; adding the corresponding state variables to the original joint navigation state for the set of newly added members, and setting temporary states and temporary covariances for the initialization process.
4. The method for initializing relative navigation consistency of a multi-center data link in a UAV swarm according to claim 1, characterized in that, The establishment of unified auxiliary common coordinates includes using a preset common reference member as a coordinate reference benchmark, generating auxiliary common coordinate estimation information based on the positional relationship between the navigation status information of each node and the common reference member, and sending the auxiliary common coordinate estimation information and the corresponding covariance information to the member to be initialized via a data link.
5. The method for initializing relative navigation consistency of a multi-center data link in a UAV swarm according to claim 1, characterized in that, The step of determining the initial position correction of the new member relative to the common reference member includes establishing a distance constraint relationship based on the received shared data of multiple members and the corresponding distance measurement information, taking the position correction of the new member as a parameter to be determined, iteratively solving the position correction based on the distance constraint relationship and the weight information corresponding to the measurement error, and determining the initial position correction and the corresponding covariance based on the solution result.
6. The method for initializing relative navigation consistency of a multi-center data link in a UAV swarm according to claim 1, characterized in that, The construction of dynamic network pseudo-measurement relationships includes constructing a pseudo-measurement relationship for the self based on the initial position correction of the new member and the state information of the common reference member, so as to constrain the positional relationship between the state of the new member and the state of the common reference member; and constructing a pseudo-measurement relationship for the other machine based on the navigation state information shared by the original members, so as to jointly update the state information of the original members and the state information of the new members.
7. A device for initializing relative navigation consistency of a multi-center data link in an unmanned aerial vehicle (UAV) swarm, characterized in that, include: Establish a multi-center relative navigation joint error state model for each node in a UAV cluster. Construct a joint error state vector based on the navigation state information of each node and the navigation state information of other nodes it maintains. Establish a corresponding state covariance module based on the correlation between the states of each node. The module detects changes in the members of the drone swarm, determines the new and off-network members based on the member set before and after the change, deletes the status information corresponding to the off-network members, augments the status information corresponding to the new members, and isolates the relative measurement information involving the new members to form a module for the temporary joint navigation status of the members to be initialized. A unified auxiliary common coordinate is established based on preset common reference members. The navigation status information of each node is converted to the unified auxiliary common coordinate. The navigation status information, auxiliary common coordinate estimation information and corresponding covariance information of each node are shared through the data link to form a module for shared data initialization of new members. The module acquires shared data and distance measurement information between new members, establishes positional constraints for new members based on the shared data and distance measurement information, and determines the initial positional correction and corresponding covariance of new members relative to the common reference members based on the positional constraints. Based on the initial position correction, the auxiliary common coordinate estimation information, and the corresponding covariance, a dynamic network pseudo-measurement relationship is constructed. The joint navigation status of the new member and the original member is updated through the dynamic network pseudo-measurement relationship, and a module for establishing the cross-correlation between the status of the new member and the status of the original member is established. This module releases the corresponding measurement isolation based on the initialization completion status of the new member, enabling the new member to participate in the cluster's relative navigation status update. Based on the updated joint navigation status, it performs status prediction, measurement updates, and error feedback correction, and outputs the relative navigation results among the UAV cluster members.
8. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method of claim 1.
9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1.
10. A computer program product, as a computer program, is characterized by: When the computer program is executed, it implements the method of claim 1.
