Method and device for self-organizing navigation and obstacle avoidance control of unmanned swarm

Abstract

The embodiment of the application provides a kind of unmanned cluster self-organizing navigation and barrier control method and device, through the innovative design formation configuration system, through hierarchical control and position parameter, realize the effective organization of formation.Combining with the ability weight and the efficiency analysis, a reliable configuration scheme is established by constructing the situation assessment mechanism.Introduce potential field control, through multidimensional control and instruction optimization, ensure the accuracy of navigation.The method effectively solves the deficiencies of traditional technology in formation configuration, situation assessment and motion control, etc., and provides technical support for unmanned cluster cooperation.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) technology, specifically to a method and device for self-organizing navigation and obstacle avoidance control of unmanned swarms. Background Technology

[0002] Existing unmanned swarm navigation and control methods have significant shortcomings. Traditional systems perform poorly in terms of formation configuration and hierarchical control, failing to effectively achieve three-dimensional spatial cooperation and affecting formation performance.

[0003] Furthermore, existing technologies suffer from bottlenecks in situation assessment and formation optimization. Most systems lack robust capability assessment mechanisms and effectiveness analysis strategies, leading to inappropriate configuration selection.

[0004] Existing systems have technical shortcomings in motion control. The lack of in-depth analysis of multidimensional control makes it difficult to achieve accurate navigation and obstacle avoidance through potential field optimization, thus impacting system performance. Solving these problems is crucial for improving the collaborative capabilities of unmanned swarms. Summary of the Invention

[0005] To address the problems in the existing technology, this application provides a method and device for self-organized navigation and obstacle avoidance control of unmanned swarms, which can effectively solve the shortcomings of traditional technologies in formation configuration, situation assessment and motion control, and provide technical support for unmanned swarm collaboration.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] Firstly, this application provides a method for self-organizing navigation and obstacle avoidance control of an unmanned swarm, including:

[0008] The three-dimensional formation configuration is decomposed into horizontal plane formation and vertical plane formation. The unmanned cluster is divided into cross-domain control layer, group control layer and platform execution layer according to the hierarchy. The encoding sequence is generated according to the hierarchical formation rules. The encoding sequence is parsed to obtain the basic formation combination. The spatial position difference between the unmanned platform and the geometric center is calculated to generate a position parameter matrix containing axial deviation, longitudinal deviation and lateral deviation. The position parameter matrix is ​​written into the formation control unit.

[0009] A situation field model including detection capability, survivability capability, and communication capability is constructed. The capability weight coefficients are calculated using the triangular fuzzy number method. The situation field model is matched with the position parameter matrix to generate a basic formation combination scheme. The formation configuration effectiveness value is calculated based on the basic formation combination scheme and written into the navigation planning unit.

[0010] An artificial potential field function is generated based on the formation configuration performance value. A motion control law containing swarm control components, obstacle avoidance control components, and navigation control components is constructed based on the artificial potential field function. The heading angle and motion rate of the unmanned platform are read. The heading angle change rate and acceleration command are calculated based on the motion control law. The heading angle change rate and acceleration command are written into the execution control unit.

[0011] Furthermore, it also includes: decomposing the three-dimensional formation configuration into coordinate axes according to the orthogonal projection principle, calculating the projection vectors of the horizontal and vertical planes, generating a formation representation matrix containing planar configuration coordinates, distribution range, and constraint conditions based on orthogonal basis transformation, and mapping the spatial distribution position of the unmanned cluster according to the formation representation matrix;

[0012] The unmanned swarm control hierarchy is divided based on a tree structure model. Gateway nodes are assigned to the cross-domain control layer, perception nodes are assigned to the group control layer, and execution nodes are assigned to the platform execution layer. Encoding sequences are generated according to the command relationships between the levels, and the encoding sequences are used to describe the unmanned swarm formation configuration.

[0013] Furthermore, it also includes: performing decoding operations on the encoded sequence according to the hierarchical formation rules, extracting the topological structure of the basic formation units, combining the basic formation units according to the command relationship to generate the target formation, and calculating the geometric center point position of the unmanned swarm based on the target formation;

[0014] The relative distance between the unmanned platform and the geometric center point is calculated based on the vector decomposition method. Axial deviation components, longitudinal deviation components, and lateral deviation components are obtained by projection along the coordinate axes. A position parameter matrix is ​​constructed based on the deviation components, and the position parameter matrix is ​​used to describe the spatial positional relationship of the unmanned cluster.

[0015] Furthermore, it also includes: constructing a multi-dimensional situation field model based on the state vector space, mapping the detection capability index to the perception range matrix, mapping the survivability index to the threat avoidance matrix, mapping the communication capability index to the link connectivity matrix, and using the triangular fuzzy number method to normalize the matrix to generate situation field weight coefficients.

[0016] The potential energy distribution map of the unmanned swarm formation configuration is established by performing tensor product operation on the situation field weight coefficient and the position parameter matrix. The topology of the basic formation is determined according to the potential energy gradient direction. The basic formation is then combined and optimized to generate a formation transformation scheme.

[0017] Furthermore, it also includes: constructing an evaluation index system based on target evaluation theory, normalizing the detection effectiveness index, survival effectiveness index, and communication effectiveness index, performing hierarchical analysis on the basic formation combination scheme, and generating an evaluation matrix containing weight coefficients;

[0018] The evaluation matrix is ​​iteratively solved using a particle swarm optimization algorithm to calculate the comprehensive efficiency coefficients of different formation configurations. The optimal formation configuration is selected based on the comprehensive efficiency coefficients, and the efficiency value of the optimal formation configuration is mapped to a navigation reference index.

[0019] Furthermore, it also includes: constructing a spatial potential energy function based on the formation configuration effectiveness value, mapping the distance between unmanned platforms to a clustering potential field component, mapping the position of obstacles to a repulsive potential field component, mapping the position of the target point to an attractive potential field component, and superimposing the potential field components based on the gradient descent method to generate an artificial potential field function.

[0020] Motion control laws are constructed based on the negative gradient direction of the artificial potential field function. The aggregation potential field component is transformed into a cluster control term, the repulsive potential field component is transformed into an obstacle avoidance control term, and the attractive potential field component is transformed into a navigation control term. The control terms are then weighted and fused to generate a synthetic control force vector.

[0021] Furthermore, it also includes: reading the heading angle information and motion rate information of the unmanned platform, calculating the desired motion direction based on the synthetic control force vector, performing a difference calculation between the desired motion direction and the current heading angle, and generating heading angle change rate and acceleration commands;

[0022] Based on the kinematic constraints of the unmanned platform, the rate of change of the heading angle and the acceleration command are limited, the minimum turning radius is mapped to the upper limit of angular velocity, the velocity boundary is mapped to the upper limit of acceleration, and motion control commands are generated according to the limiting results.

[0023] Secondly, this application provides an unmanned swarm self-organizing navigation and obstacle avoidance control device, comprising:

[0024] The formation control module is used to decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate encoding sequence according to the hierarchical formation rules, parse the encoding sequence to obtain basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0025] The organization navigation module is used to construct a situation field model that includes detection capability, survivability capability, and communication capability. It uses the triangular fuzzy number method to calculate the capability weight coefficients, performs matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme, calculates the formation configuration effectiveness value based on the basic formation combination scheme, and writes the formation configuration effectiveness value into the navigation planning unit.

[0026] The obstacle avoidance control module is used to generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command according to the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0027] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method.

[0028] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the described unmanned swarm self-organizing navigation and obstacle avoidance control method.

[0029] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the aforementioned unmanned swarm self-organizing navigation and obstacle avoidance control method.

[0030] As can be seen from the above technical solution, this application provides a method and device for self-organized navigation and obstacle avoidance control of unmanned swarms. Through innovative design of the formation configuration system, and by using hierarchical control and position parameters, it achieves effective formation organization. A situation assessment mechanism is constructed, and a reliable configuration scheme is established by combining capability weights and effectiveness analysis. Potential field control is introduced, and navigation accuracy is ensured through multi-dimensional control and command optimization. This method effectively solves the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the unmanned swarm self-organizing navigation and obstacle avoidance control method in the embodiments of this application;

[0033] Figure 2 This is a structural diagram of the unmanned swarm self-organizing navigation and obstacle avoidance control device in the embodiments of this application;

[0034] Figure 3This is a schematic diagram of the structure of the electronic device in the embodiments of this application.

[0035] Figure label:

[0036] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0039] To address the problems existing in current technologies, this application provides a method and device for self-organized navigation and obstacle avoidance control in unmanned swarms. Through an innovative design of the formation configuration system, and by using hierarchical control and position parameters, it achieves effective formation organization. A situation assessment mechanism is constructed, and a reliable configuration scheme is established by combining capability weights and performance analysis. Potential field control is introduced, and navigation accuracy is ensured through multi-dimensional control and command optimization. This method effectively solves the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0040] To effectively address the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, and to provide technical support for unmanned swarm collaboration, this application provides an embodiment of an unmanned swarm self-organizing navigation and obstacle avoidance control method. See [link to embodiment]. Figure 1 The unmanned swarm self-organizing navigation and obstacle avoidance control method specifically includes the following:

[0041] Step S101: Decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0042] This embodiment is designed for unmanned swarms operating collaboratively across three domains: air, surface, and submarine. It is executed synchronously around the S101 phase during formation initialization and rapid reconfiguration after being threatened or disturbed.

[0043] First, the formation is decomposed at the 3D geometric level. The system uses a cross-domain unified coordinate system O-XYZ as a reference and projects the designed formation configuration P3D onto the horizontal plane ΩH (XY) and the vertical plane ΩV (XZ) using orthogonal projection, resulting in two planar formations, P_H and P_V. Considering the undulating sea surface topography and the hierarchical layering of airspace, the projection does not simply discard dimensions but retains attitude constraints and safety clearances: the expected planar spacing and connectivity constraints of the members are recorded in ΩH, and the height corridors, cross-domain coverage cones, and other constraints are recorded in ΩV. This forms the formation representation matrix M_f, which contains the target coordinates, distribution range, and non-intrusive domain of each member on the two planes. The motivation for this step is to decompose the 3D coupling problem into two constrained 2D problems, facilitating subsequent hierarchical control mapping.

[0044] At the control organization level, this embodiment divides the unmanned swarm into a cross-domain control layer L1, a group control layer L2, and a platform execution layer L3 based on mission links and physical capabilities. L1 is composed of gateway nodes with multi-domain communication and computing capabilities, responsible for cross-domain intent and global time base; L2 consists of perception and relay nodes, responsible for local formation changes and threat reporting; L3 is the specific execution platform (fixed-wing, unmanned surface vessel, AUV), which implements position and trajectory control.

[0045] To establish a clear correspondence between structure and geometry, a hierarchical encoding / decoding mechanism is introduced: the formation is numbered from top to bottom using a tree structure, with the parent node's encoding prefix indicating the upper-level formation unit and the child node's suffix indicating the arrangement within that unit. The encoded sequence E contains both hierarchical relationships (who belongs to whom) and relative positional information indices within ΩH and ΩV. The encoding order follows a "top-down, left-to-right" rule, which allows for unified semantic description across different domains and groups, reducing matching complexity during reconstruction.

[0046] When parsing the encoded sequence, this embodiment first locks down the basic formation units (such as eight primitives including linear, wedge, and rhomboid shapes) based on the upper-level command relationship, and then combines these primitives according to the local topology in E to obtain the planar arrangement of the target formation P_H and P_V. The geometric center G is defined as the "mission geometric center," that is, the centroid of the projection points of each member is taken at ΩH, and the weighted center of the altitude layer is taken at ΩV (the weight is given by the mission layer or the safe altitude band), and finally mapped back to the three-dimensional reference center. The reason for introducing weighting is to reflect the importance difference of the altitude layer in the air-surface-potential threat avoidance, for example, the weight of the underwater vehicle is lowered in strong swells to avoid unreasonable traction on the air formation center.

[0047] After obtaining the center G, the system calculates the spatial position difference Δri for each unmanned platform i and decomposes it along the coordinate axes into axial deviation dx_i, longitudinal deviation dy_i, and lateral deviation dz_i. The naming of the axial, longitudinal, and lateral directions is consistent with the mission heading: the axial direction is along the main mission heading, the longitudinal direction is orthogonal to it and located in the horizontal plane, and the lateral direction is the altitude direction. The body coordinate system is not directly used here to make the constraints clearer at the group level. The above deviations are constructed into a position parameter matrix R=[dx dy dz]N×3, where the row index corresponds to the platform identifier, and the columns correspond to the deviations in the three directions. The matrix also includes the tolerable bandwidth for each deviation (from the platform type and safety distance library). R is summarized by L2 and a consistency check is performed. If a member has a discrete value due to sensor anomaly, it will be temporarily filled by interpolation of the neighboring nodes in the same layer to avoid the global deviation due to a single point anomaly.

[0048] From a technical perspective, the role of 3D decomposition and hierarchical coding in this application scenario is to bind the three chains of "topology-geometry-control" to a single computable object. Hierarchical control loads and communication constraints are particularly critical during sea state changes: when a surface node temporarily loses contact, L2 can still locally recover its relative position slot using the coded sequence, without disrupting the overall topology; when a no-fly zone appears in the vertical direction, the ΩV formation matrix directly restricts the target domain of dz, preventing unreachable false attraction during subsequent potential field generation. For the turning radius constraint of fixed-wing aircraft, the rates of change of dx and dy in the R matrix are limited downstream and transmitted to the execution layer through attitude gain, preventing unnecessary drastic rearrangements from being introduced at the upper layer.

[0049] To illustrate the mechanism's operation, two scenarios are presented. First, when the formation traverses a narrow passage, obstacle constraints on ΩH increase. After decoding, a combination of "linearity + spacing expansion" is selected in P_H. ΩV maintains two layers of elevation coordination, the tolerance zone for dy in R shrinks, and the target difference in dz amplifies, facilitating overhead coverage for the air squadron. Second, encountering a temporary no-fly zone, L1 issues an altitude band adjustment. The decoder reduces the available layers in ΩV, flattening the original "wedge" shape in the vertical plane. Targets with dz deviations in R converge to the new layer. The L3 fixed-wing aircraft gradually descends to the new altitude slot according to the span constraint, without needing significant lateral maneuvers, maintaining communication connectivity.

[0050] In terms of integration with subsequent steps, the R matrix is ​​written into the formation control unit as the root data for the construction of the artificial potential field and the weighting of the control law. Since the clustering components of the subsequent potential field depend on the expected relative distance, R provides a deviation reference for each pair of members to G, allowing the construction of a "clustering-alignment" objective term in the potential function. The obstacle avoidance component requires an intrusive domain of ΩH / ΩV, the boundary of which has been provided by M_f. The attraction direction of the navigation component is axial, strengthening the consistency of the mission's main heading. To avoid parameter drift, the formation control unit sets a sliding window for R, increasing weighting for smoothing during intense maneuvers and allowing faster convergence during stable phases, thereby maintaining a natural transition between the decoding formation and the physical motion, without generating frequent reconstruction jitter.

[0051] Based on engineering implementation, state updates employ a two-level timescale: the structural part of the encoded sequence is updated only when the formation scheme changes, while the position parameter matrix is ​​updated at fixed intervals or triggered by events (e.g., obstacle warnings, member dropouts). The cross-domain control layer provides timestamps and version numbers, while the group control layer only accepts new version numbers not less than the locally recorded R, avoiding rollback. The formation control unit incorporates redundancy checks; if the joint position obtained from decoding ΩH and ΩV is infeasible for a certain member, it will revert to the previous feasible solution and mark that member as "pending repair" in R, handing it over to subsequent self-organizing repair strategies for processing.

[0052] In summary, from the perspective of technical issues and effects, S101 addresses the challenge of translating abstract formation intentions into executable and constrained spatial location descriptions, and aligning them with hierarchical control structures. It reduces 3D coupling complexity through orthogonal decomposition, maintains topology readability and recoverability through layered encoding, and concretizes the target deviations of each platform through center and deviation matrices. Ultimately, it provides consistent, low-conflict input for subsequent situational awareness modeling, control law decomposition, and execution layer limiting.

[0053] Step S102: Construct a situation field model that includes detection capability, survivability, and communication capability; calculate capability weight coefficients using the triangular fuzzy number method; perform matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme; calculate the formation configuration effectiveness value based on the basic formation combination scheme; and write the formation configuration effectiveness value into the navigation planning unit.

[0054] This embodiment focuses on the formation assessment and reconstruction in the three-domain coordination of air, surface and submarine at sea in S102. It follows up on the position parameter matrix R (containing the axial, longitudinal, and lateral deviations and tolerance zones of each platform relative to the geometric center) written into the formation control unit in S101. The goal is to abstract the three capabilities of detection, survival and communication into a computable situational field under the sudden threat or environmental disturbance, and integrate them into the formation selection and effectiveness assessment in a weighted uncertain manner. Finally, the executable effectiveness value is written into the navigation planning unit to provide a basis for the subsequent potential field control law.

[0055] This embodiment first constructs a situational awareness model. For detection capabilities, the sensing nodes form a sensing coverage matrix S based on sensor range, pitch / field-of-view geometry, and sea state visibility. The element Sij represents the effective detection coverage of platform i over target sector j, taking into account platform overlap loss and line-of-sight obstruction. For survivability, a threat avoidance matrix V is formed based on obstacle distribution, threat source direction, wave height, and platform vulnerability. Vij characterizes the risk cost of platform i being located in formation slot j, including safety margin for tangential wave motion and minimum turning radius limitations. For communication capabilities, a connectivity matrix C is generated based on link quality, bandwidth, and relay location. Cij describes the connectivity strength maintained by platform i in slot j with the upper / peer layer. The spatial references for all three are consistent with R, facilitating slot-by-slot matching.

[0056] A single sea state parameter is insufficient to stably represent priority; for example, the detection weight increases during sudden threats at night, but not in a leapfrog manner. This embodiment assigns triangular fuzzy numbers Ãs, Ãv, and Ãc to the weights of detection, survivability, and communication, respectively, with (l, m, u) representing pessimistic, most reliable, and optimistic values, derived from mission rules and recent observations (e.g., an increased collision warning frequency would boost the survivability weights m and u). A fuzzy judgment matrix is ​​obtained through fuzzy hierarchy comparison, and then defuzzification yields a clear weight vector w = [ws, wv, wc], where w satisfies the constraint that the sum is 1. The inherent relationship in this process is: increased risk naturally leads to a higher survivability weight; deteriorating communication links lead to a higher communication weight to stabilize formation connectivity; and during reconnaissance missions, the detection weight is increased, conforming to the common-sense principle of mission-environment dual-drive.

[0057] When performing a matching operation between the situation field and R, this embodiment treats the set of slots Q in R as the locations to be packed, and generates a slot-to-entity assignment mapping π for each candidate basic formation. The matching score consists of three parts: detection gain F_s(π)=∑i S_{i,π(i)}, survival safety F_v(π)= Let ∑i V_{i,π(i)} be the communication connectivity, F_c(π) = ∑i C_{i,π(i)} be the communication connectivity, and introduce a deviation tolerance band of R as a feasibility constraint. If an assignment causes |dx,dy,dz| to exceed the platform's tolerance range, then the mapping is marked as low feasibility or eliminated. By calculating the comprehensive score J(π) = ws·F_s + wv·F_v + wc·F_c, several high-scoring mappings are found within each candidate basic formation, and then spliced ​​at the planar combination level to form a basic formation combination scheme. To reduce combinatorial explosion, this embodiment uses the structure of the connectivity matrix C to first group the data, limiting strongly connected subsets to assignments within adjacent slot intervals, which conforms to communication routing habits and reduces computational complexity.

[0058] In the performance calculation phase, this embodiment uses a target evaluation system to verify the combined schemes. Considering the marginal differences in the three types of performance for different formations, an evaluation matrix E is constructed, consisting of normalized column vectors of detection performance, survival performance, and communication performance. Hierarchical analysis is introduced to perform pairwise comparisons between schemes and verify the consistency of fuzzy weight derivation. Subsequently, particle swarm optimization is used to search for near-optimal solutions in a continuous-discrete hybrid space of "formation selection × slot assignment × small deviation compensation". The swarm state is represented by fine-tuning within π and R as particle positions, and the fitness is J(π) minus the constraint penalty.

[0059] This embodiment provides a scoring formula for explanation:

[0060] J(π)=ws·∑i S_{i,π(i)}+wv·( ∑i V_{i,π(i)})+wc·∑i C_{i,π(i)}.

[0061] Where J(π) is the overall performance score under the current formation combination; ws, wv, and wc are the weight coefficients for detection, survival, and communication, respectively, derived from the clarity values ​​of the triangular fuzzy number defuzzification; S_{i,π(i)} is the detection coverage gain after platform i is assigned to slot π(i); V_{i,π(i)} is the corresponding risk cost, with the negative sign indicating that the higher the safety cost, the lower the performance; and C_{i,π(i)} is the connectivity strength. All parameters in this formula can be calculated or obtained from the situation field matrix and the position parameter matrix.

[0062] In the application segment, when the formation traverses fog and the sea becomes rougher, visibility decreases, causing S to decay and the threat cost V to rise. After defuzzification, wv increases and ws decreases. The particle swarm search favors a combination of "wedge-shaped + increased spacing," with slots spread out on ΩH and maintaining upper and lower layers on ΩV. The strong connectivity constraints of the communication matrix C maintain backbone stability, and the final increase in J(π) mainly comes from the decrease in the V term. When handling temporary target searches, S increases in the target sector, the system raises ws, and the combination is more biased towards "fan-shaped / diamond-shaped" to cover the sector corners. The dz tolerance band in R is used to arrange the air detachments to raise their altitude and obtain an overhead view. The C term maintains connectivity by placing several relays near the geometric center.

[0063] This embodiment writes the formation configuration effectiveness value corresponding to the optimal combination scheme into the navigation planning unit, carrying the version and weight snapshot, for subsequent setting of the intensity ratio of aggregation, repulsion and attraction terms when constructing the artificial potential field in S103. To avoid frequent switching caused by instantaneous fluctuations, the navigation planning unit introduces time window averaging and threshold hysteresis, updating only when the benefit of J(π) relative to the current scheme exceeds the threshold and the feasibility check is met; for fixed-wing platforms in the execution layer, the effectiveness value is also decomposed into reference speed bands and angle windows, facilitating alignment with the global intent even after the control law is limited. Thus, the link of situation field—weight—matching—effectiveness forms a closed loop in the event of a sudden threat, maintaining attention to the mission objective while ensuring the bottom line of survival and connectivity in adverse sea states.

[0064] Step S103: Generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command based on the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0065] This embodiment takes the formation configuration performance value and its corresponding weight snapshot output by S102 and, at the control layer, transforms the static judgment of "whether the formation is appropriate" into a continuously executable artificial potential field and motion control law. To avoid the three-domain platform making contradictory maneuvers under sudden threats, this embodiment starts with potential energy calibration in a unified coordinate system.

[0066] First, the position parameter matrix R and the optimal configuration and performance values ​​written back by the navigation planning unit are read from the formation control unit. ,Will The component weights (detection, survival, communication) are mapped to potential field strength scaling coefficients. Based on the Olfati-Saber framework, three types of potential fields are constructed: a clustering potential field U_f characterizes formation cohesion based on the consistency of expected relative distance and velocity; an obstacle avoidance potential field U_o characterizes the buffer zone based on obstacle boundaries, normals, and minimum safe distance; and a navigation potential field U_g provides attraction based on the target point and the mission's main heading. The gradients of these three fields are calculated in the ΩH and ΩV planes, and then synthesized in three dimensions to ensure that the altitude constraints and horizontal maneuvers do not overload each other.

[0067] For the specific generation of the potential field function, this embodiment calculates the deviation energy and velocity deviation energy of adjacent cluster members for each platform i, and constructs U_f using a smoothing norm and interaction function, so that it is still differentiable when the relative distance is close to zero, avoiding numerical oscillations; for U_o, obstacles are divided into two categories according to boundary type: "hyperplane" and "circle", which are solidified into the potential energy through the geometric construction of virtual node q_i. q_i moves along the boundary tangent within the platform's detection radius, generating a resultant force direction for edge-hugging sliding; U_g generates attraction with the geometric center of the mission target and the principal axis direction, with strength and... The navigation reference indicators are consistent with those in the example. The potential field parameters are not fixed. In this embodiment, the weight w of S102 is used as the outer loop adjustment of the gain. When the survival weight is increased, the gain of U_o is increased, and U_g is temporarily limited, which is in line with the logic of "safety first, mission second".

[0068] In the control law construction phase, to transform the potential field into executable trajectory commands, this embodiment uses the negative gradient method to obtain the synthetic control force F_i= (U_f+U_o+U_g) is then mapped to the controllable channel of fixed-wing / unmanned surface vessels and other platforms. The current heading angle ψ_i and ground speed v_i of platform i are read, and the desired direction φ_i and amplitude a_i^ref of the resultant force are calculated, thereby generating the desired heading difference Δψ_i and acceleration reference. Considering the kinematic differences of fixed-wing systems, this embodiment does not directly treat F_i as acceleration, but instead implements it in two steps: first, F_i is projected onto the horizontal plane to obtain the desired turning angle; second, the tangential component of the resultant force is used for acceleration or deceleration. To reduce disturbance to the communication backbone, a connectivity maintainer is also superimposed on the resultant control term, allowing for smaller heading deviations in the direction of strongly connected edges in the adjacency matrix.

[0069] Before issuing control parameters to the execution layer, a limiting and consistency check is required. The limiting involves three constraints: minimum turning radius, speed boundary, and upper limit of heading angular rate. All constraint parameters are from the platform library and can be adjusted upwards to allow for a safety margin when sea conditions worsen. The consistency check includes confirming the closed loop and altitude layer accessibility. If the dive or climb caused by U_g exceeds the allowable ΩV layer, a soft constraint penalty term is added to the potential field, allowing the resultant force to naturally avoid the inaccessible region. To express the mapping relationship, an engineering formula is given:

[0070] u_i=[dot{ψ}_i, a_i]=[k_ψ·wrap(φ_i ψ_i), k_a·||F_i||],

[0071] Where dot{ψ}_i is the rate of change of heading angle, a_i is the acceleration command; φ_i is the azimuth angle of the resultant force F_i on the horizontal plane, ψ_i is the current heading angle, and wrap(·) is the angle difference normalized to [ The operators are π,π; k_ψ and k_a are gains tuned according to the platform response characteristics. Each parameter in the formula comes from a physical quantity: a small angle difference results in a gentle turn, while a large resultant force results in strong acceleration and deceleration.

[0072] Considering the information delays and uncertainties in air-surface-submarine collaborative systems, this embodiment incorporates hysteresis and filtering into the control law. Hysteresis is used to suppress repeated policy jumps caused by rapid weight jitter, while filtering is used to smooth measurement noise in ψ_i and v_i. For individuals experiencing short-term link loss, the cluster component U_f can still be maintained by relative observations of surrounding neighbors, preventing isolated individuals from drastically deviating from the formation. If a member falls behind, the execution layer automatically weakens the navigation component based on obstacle avoidance and inter-machine collision avoidance strategies.

[0073] In a typical segment, the formation traverses two circular obstacles. The sea surface is rough, and the navigation planning unit increases the survival weight, resulting in a significant enhancement of U_o. The φ_i generated by each platform rolls along the obstacle tangent, and Δψ_i is constrained within an achievable range by the limiter. Speed ​​commands tend to decelerate to allow for maneuver margins. In another segment, during nighttime target searches, the detection weight increases, and U_g and sub-items related to sensor coverage are enhanced. The air squadron's dz target is determined by R. The climb acceleration generated by the control law is executed frame-by-frame after limitering. The cluster component maintains a wedge-shaped topology, preserving strong connectivity edges with the surface relay.

[0074] In this embodiment, dot{ψ}_i and a_i are ultimately packaged into an instruction frame recognizable by the execution control unit, with a timestamp and weighted version. The execution layer generates control surface and throttle commands based on its own dynamic model. Since the potential field and weights are derived from the performance evaluation in the previous stage, the control link achieves a natural mechanical balance between mission objectives, threat avoidance, and communication connectivity. This avoids biased maneuvering driven by a single objective and provides traceable control data for subsequent status monitoring and reassessment.

[0075] As described above, the unmanned swarm self-organizing navigation and obstacle avoidance control method provided in this application can achieve effective formation organization through innovative formation configuration system design, hierarchical control, and position parameters. It constructs a situation assessment mechanism, combining capability weights and performance analysis to establish a reliable configuration scheme. By introducing potential field control and optimizing commands through multi-dimensional control, it ensures navigation accuracy. This method effectively addresses the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0076] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0077] Step S201: Decompose the coordinate axes of the three-dimensional formation configuration according to the orthogonal projection principle, calculate the projection vectors of the horizontal and vertical planes, generate a formation representation matrix containing the planar configuration coordinates, distribution range, and constraints based on orthogonal basis transformation, and map the spatial distribution position of the unmanned cluster according to the formation representation matrix.

[0078] Step S202: Based on the tree structure model, divide the unmanned swarm control hierarchy, assign gateway nodes to the cross-domain control layer, assign perception nodes to the group control layer, assign execution nodes to the platform execution layer, generate coding sequences according to the command relationship between the hierarchy, and use the coding sequences to describe the unmanned swarm formation configuration.

[0079] This embodiment focuses on the geometric and organizational mapping of formations in the three domains of air, surface and submarine at sea. It describes the abstract three-dimensional formation in two planes around S201 to S202, and then maps it one-to-one with the control level to the coding sequence, so as to ensure that the subsequent potential field and control law have a computable and traceable geometric-organization consensus.

[0080] In this embodiment, a unified reference coordinate system O-XYZ is first determined in S201, where X is along the mission's main heading, Y is the horizontal lateral direction, and Z is the vertical direction. The three-dimensional formation configuration P3D in the design situation is decomposed into coordinate axes, and its projection vector sets in the horizontal plane ΩH (XY) and vertical plane ΩV (XZ) are calculated according to the orthogonal projection principle. The projection preserves the main axis direction and local topological relationships of the formation. Considering the physical limitations of different platforms, an orthogonal basis transformation B=[eX,eY,eZ] is used to transform the local mission basis B'=[eA,eL,eS], where eA is consistent with the mission heading (axial direction), eL is the horizontal lateral direction, and eS is the altitude direction, facilitating the expression of the desired spacing and safety zone in mission coordinates. For each member node i, its target coordinates and feasible zones on ΩH and ΩV are calculated and summarized into a formation representation matrix Mf, which includes: planar configuration coordinates (ΩH: x_i, y_i; ΩV: x_i, z_i), distribution range (anisotropic tolerance zones ΔA, ΔL, ΔS), and constraints (minimum height layer, minimum lateral spacing, non-intrusive obstacle domain, curvature constraints caused by the minimum turning radius of the fixed wing). The logic here is to decompose the 3D coupling into two constraint graphs, which are then connected by a common axis X. The two graphs jointly define the allowable spatial slots.

[0081] To use Mf for entity mapping, this embodiment constructs a slot set Q={qk}, where each slot qk has a pair of planar coordinates and a corresponding constraint segment on ΩH and ΩV, aligned using the same X-coordinate or path arc length parameter. The mapping process follows a "feasibility priority—connectivity priority—performance suboptimal" order: first, slots intersecting with obstacle domains, no-fly zones, or those with curvature inaccessibility are excluded; then, slots are clustered based on the communication connectivity matrix, assigning members with high connectivity requirements to neighboring slots within the same cluster; finally, slots for airborne squadron lifting are reserved with reference to the detection sector. This order avoids conflicts arising from prioritizing geometric aesthetics, which could render subsequent connectivity or maneuvers impossible. After mapping, Mf is written as structured data to the formation control unit, providing a direct index for the position parameter matrix R of S101 and subsequent potential field generation.

[0082] In S202, this embodiment describes the organizational structure and geometric slots together. First, based on a tree structure model, the unmanned cluster is divided into control layers: the root node is the cross-domain control layer L1, which carries the gateway and time base; its child nodes are the group control layer L2, which divides several groups according to geographical proximity and communication communities, with each group corresponding to a sub-block on ΩH and a height layer interval on ΩV; the leaf nodes are the platform execution layer L3, which is specific to each platform type and individual. When generating the encoded sequence E, a construction method of "hierarchical prefix + planar index suffix" is adopted: the prefix identifies L1 / L2 / L3 and the parent-child relationship, and the suffix is ​​composed of the slot number <kH,kV> and the formation primitive type identifier. The encoding order follows the rule of top to bottom and left to right, so that the same subgroup is continuous in the sequence, which facilitates loss recovery and local reconstruction. Encoding is more than just labels; it carries constraints and responsibilities. For example, a gateway node writes the label "cross-domain hub" in E, a perception node writes "sector coverage angle domain", and an execution node writes "mobility capability level". These labels take effect as hard / soft constraints during subsequent matching.

[0083] To ensure that the encoded sequence accurately reflects the geometric configuration, this embodiment introduces an "encoding-slot binding table" Tbind, which records the bidirectional reference between each element in E and slot qk in Mf. When the formation primitive changes from a wedge shape to a linear shape in ΩH, only the suffix of the affected subtree and Tbind are updated, while the prefix remains unchanged, thus avoiding communication storms caused by large-scale recoding.

[0084] Considering the non-stationarity of the marine environment, the granularity of the tree is not fixed: when the group size increases and the connectivity quality decreases, L2 can split into more subgroups, and the encoding prefix length increases accordingly; conversely, it merges in stationary segments to maintain a compact structure. The technical point here is that tree-based encoding solidifies the topological control relationship, geometric placeholders, and platform roles into a unified literal sequence, and any change in any dimension can be located in a limited substring modification.

[0085] From a technical perspective, S201 addresses the lossless dimensionality reduction and maneuver feasibility injection from 3D to two-plane; S202 addresses the synchronous expression and low-cost reconstruction of control links and geometric configurations. When combined, the situational field weights of S102 and the potential field control law of S103 can both be indexed by E and Mf / Tbind: for example, during weight-boosting survival, only the height layer index and slot set of the suffix need to be adjusted on the risky subtree; furthermore, if a sensing node falls behind, L2 can read the parent-child relationship of E to assign a backup member to take over its sector coverage, and Tbind can be used to map the new member to feasible positions around the original slot, resulting in a smooth transition of the R matrix and avoiding formation tearing.

[0086] Here are two examples illustrating application scenarios. First, when a formation traverses a narrow channel through reefs on either side, the feasible corridor of ΩH is significantly compressed, reducing the lateral tolerance zone ΔL in Mf. The corresponding subgroups of the tree are marked with a "narrow channel" in the encoding. E→Tbind drives these subgroups to select linear primitives and converge to the slots on the corridor centerline. The air detachment maintains two layers on ΩV, with ΔS expanded to ensure the radar's downward viewing angle. Second, a temporary search mission requires rapidly increasing detection coverage in a certain sector. L1 adjusts the group weights for that sector, and the suffix of the sensing nodes in E highlights the sector's corner region. Tbind pushes these elements to the slots in the ΩH sector expansion. ΩV moderately elevates the upper-level detachments to ensure unobstructed line of sight. Connectivity constraints ensure that relay nodes maintain stable numbering near the geometric center, without affecting the backbone link.

[0087] From an engineering implementation perspective, this embodiment manages E and Mf versions within the formation control unit, with each write including a timestamp and source. When the joint slots resolved by ΩH and ΩV conflict with the platform capability library, the mapping process will revert to the most recent feasible version and indicate the reason for the anomaly, handing it over to the navigation planning unit for processing in the next round of optimization. Through the above process, the 3D configuration is precisely, hierarchically, and constrainedly unfolded into a two-plane representation and tree-based encoding.

[0088] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0089] Step S301: Decode the encoded sequence according to the hierarchical formation rules, extract the topological structure of the basic formation units, combine the basic formation units according to the command relationship to generate the target formation, and calculate the geometric center point position of the unmanned swarm based on the target formation.

[0090] Step S302: Calculate the relative distance between the unmanned platform and the geometric center point based on the vector decomposition method, and obtain the axial deviation component, longitudinal deviation component, and lateral deviation component by projection along the coordinate axis. Construct a position parameter matrix based on the deviation components, and use the position parameter matrix to describe the spatial positional relationship of the unmanned cluster.

[0091] This embodiment focuses on S301-S302 to restore the organizational code into a geometrically executable target and form a position parameter matrix for subsequent situational field and control law calls. The inputs are the code sequence E, formation representation matrix Mf, and binding table Tbind output from S202. The core idea is: first, solve E according to the hierarchical formation rules to extract the basic formation units and their command and subordinate relationships; then, assemble the target formation using these units as block elements; and finally, calculate the three-dimensional deviations between the geometric center and each platform under a unified reference coordinate system.

[0092] In this embodiment, decoding is performed first in S301. The encoding E consists of a "hierarchical prefix + planar index suffix". The prefix identifies the parent-child relationship between the cross-domain control layer / group control layer / platform execution layer, and the suffix is ​​the slot number <kH,kV> and the basic formation primitive identifier. The decoder expands from top to bottom in a tree traversal manner: at each L2 subtree, it reads the primitive type in the suffix (such as eight types including linear, wedge, rhombus, and fan), and obtains the corresponding subgroup's slot set and constraint fragment in ΩH and ΩV from Mf and Tbind. For each subgroup, first reconstruct its internal topology (e.g., a wedge is composed of a main axis node and symmetrical nodes on the left and right wings, with the lines arranged in axial order). Then, determine the splicing method between subgroups based on the command relationship of the parent node: if the parent node is a cross-domain gateway, splicing adopts "communication backbone priority," placing high-level relay nodes in slots near the geometric center; if the parent node is a reconnaissance command node, splicing adopts "coverage priority," distributing high-detection-capability nodes in sector edge slots to form an expanded state. After assembly, the target formation is obtained. Arrangement on the two planes ΩH and ΩV.

[0093] After determining the target formation, the system calculates the geometric center point G. In the three domains of air, surface, and submersible at sea, differences in altitude and medium can affect the direct centroid due to imbalances in the number of elements between layers. Therefore, this embodiment employs a layered weighted geometric center: in ΩH, the planar centroid is calculated using the projection points of all assigned slots; in ΩV, layer weights w_z (determined by the mission phase and safe altitude zone) are introduced, and the weighted center is calculated after clustering by altitude layer. Finally, G consists of <x_c, y_c, z_c>, where x_c aligns with the mission axis to ensure the attraction direction of the subsequent navigation potential field is consistent with the mission. The rationale for weighting is that, for example, if there are many submersibles but they should tactically follow the surface backbone, their influence on the formation center needs to be suppressed; otherwise, it would drag the longitudinal layout of the air subgroup, increasing unnecessary energy consumption and communication link length.

[0094] Entering S302, this embodiment calculates the relative distance between each platform and G based on vector decomposition in the unified coordinate system O-XYZ. For platform i, the target slot coordinates p_i and the current estimated position r_i (if used for verification) are taken, and the relative vector e_i=p_i is calculated based on the target geometry. G is projected onto the task orthogonal basis B'=[eA,eL,eS], yielding the axial component dx_i=〈e_i,eA〉, the longitudinal component dy_i=〈e_i,eL〉, and the lateral component dz_i=〈e_i,eS〉. The axial direction is along the main course of the task, the longitudinal direction is horizontal, and the lateral direction is the altitude direction. This projection is consistent with the basis transformation in S201, facilitating the subsequent control law to directly map the deviation into acceleration and steering requirements. To avoid contaminating the collective description with a single abnormal slot, constraint verification is introduced in the deviation calculation: if a slot falls near the boundary of the non-intrusive domain of Mf and the curvature constraint is unreachable, then a micro-shift compensation is performed on the slot before generating the R matrix. The compensation amount is limited to the tolerance bands of ΔA, ΔL, and ΔS and recorded in the attribute field of R for the execution layer to perceive.

[0095] This embodiment condenses the above process into a position parameter matrix R, which is an N×3 matrix. Each row corresponds to a unique identifier for a platform (or slot), and the columns are dx, dy, and dz, respectively. It also includes three metadata fields: tolerance threshold, constraint flags (such as "near obstacle," "near boundary," and "curvature tight constraint"), and hierarchy labels (L1 / L2 / L3). R is written to the formation control unit and managed using versioning. The navigation planning unit and execution control unit subscribe to R based on timestamps. R not only describes the deviation itself but also defines the adjustment space of the deviation, thereby avoiding the generation of unrealizable control quantities in subsequent potential field calculations. If a platform falls behind or loses connection, the decoder fills its slot in the tree with a sibling node, and R marks this position as "pending takeover." The cluster components of the potential field will process this point with low weight.

[0096] In terms of technical principle correlation, the decoding-assembly-center-deviation chain translates the encoded semantics of S202 into geometric quantities that can be directly computed by S102 and S103. S102 requires matching the detection, survival, and communication matrices with the position, and R provides the spatial relationship between each slot and the center, facilitating the calculation of coverage overlap and risk exposure; S103 requires generating a potential field based on the deviation, and dx, dy, and dz directly enter the energy definition of aggregation and navigation terms, with dz also related to the intersection judgment of the height layer forbidden region in obstacle avoidance. Through the embedding of layered weighted G and tolerance zone, the subsequent control law will not exhibit phenomena such as excessively rapid center drift and frequent formation reconfiguration during sea state fluctuations.

[0097] To illustrate the computational logic more intuitively, a vector decomposition formula is provided for explanation:

[0098] e_i=p_i G, dx_i=〈e_i,eA〉, dy_i=〈e_i,eL〉, dz_i=〈e_i,eS〉.

[0099] Where e_i is the displacement vector of platform i relative to the geometric center; p_i is its target slot coordinate; G is the geometric center coordinate; eA, eL, and eS are the unit basis vectors in the axial, horizontal, and height directions, respectively; 〈·,·〉 represent the vector dot product operation. dx_i, dy_i, and dz_i are scalar deviations in the three directions, which will be directly used by the control law for the decomposition calculation of heading and acceleration. The physical meaning of the parameters is clear and consistent with the mission's main heading.

[0100] In a segment from a maritime application, when traversing narrow channels, the decoded linear subgroup is centrally arranged, with G close to the corridor centerline on ΩH. The dy component in R converges overall, and dz expands at safe inter-layer distances on ΩV. When searching for targets at night, the fan-shaped subgroup expands on ΩH, and the air squadron is assigned to a larger dz to obtain a top-down view. The dz tolerance zone recorded by R provides an upper limit for the climb amplitude of the execution layer. In another segment, a surface relay loses contact, and the tree subgroup automatically fills the slots with backup nodes. G shifts slightly, and R makes limited fine adjustments to the relevant dx and dy. The potential field generated by S103 does not trigger a large-scale rearrangement of the entire squadron, thus preserving connectivity.

[0101] From an engineering integration perspective, this embodiment stably delivers the outputs of S301 to S302 to subsequent modules in the form of a data structure. This preserves command relationships and role information while providing clear deviations and constraints at the geometric level. Through version and anomaly annotations, the navigation and execution layers can determine when to adopt a conservative control strategy and when to allow faster convergence to the target formation. At this point, the encoded semantics have been fully "landed" as spatial quantities, facilitating rapid assessment and the issuance of consistent maneuver commands under sudden threats.

[0102] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0103] Step S401: Construct a multi-dimensional situational field model based on the state vector space, map the detection capability index to the perception range matrix, the survivability index to the threat avoidance matrix, and the communication capability index to the link connectivity matrix, and use the triangular fuzzy number method to normalize the matrix to generate situational field weight coefficients.

[0104] Step S402: Perform tensor product operation on the situation field weight coefficient and the position parameter matrix to establish the potential energy distribution map of the unmanned swarm formation configuration, determine the topology of the basic formation according to the potential energy gradient direction, and perform combination optimization on the basic formation to generate a formation transformation scheme.

[0105] This embodiment targets a coordinated maritime air-surface-submarine environment. Around S401-S402, it unifies the "measurable capability indicators—spatial positional relationships" into a situational awareness framework, allowing formation selection and transformation to be naturally derived from potential energy distribution and gradient direction, rather than relying on empirical thresholds. Inputs include the position parameter matrix R (dx, dy, dz and tolerance band) generated by S301-S302, the formation representation matrix Mf generated by S201, and link quality, threat, and visibility data continuously reported by the communication and sensing subsystems.

[0106] In S401, this embodiment first constructs a state vector space X. The state vector xi of each platform i includes platform type, load sensing parameters (field of view, range, pitch), structural safety parameters (minimum turning radius, permissible overload), communication parameters (transmit power, available bandwidth, current link SNR), and spatial deviation bound to R. Three types of capability matrices are mapped based on X. First, the sensing range matrix S: the task sector is discretized into an angle-distance grid, calculating the visibility coverage of platform i in slot q, considering the reduction of effective range due to attitude and sea state visibility, and avoiding false high-altitude coverage through overlap subtraction. Second, the threat avoidance matrix V: constructed from obstacle geometry (hyperplane, circle) and threat source direction and wave height level, converging the risk cost with minimum safe distance, curvature attainability, and tangential passage margin; excessively large dz or insufficient curvature will increase the cost. Third, the link connectivity matrix C: Based on RSSI / SNR, link hop count, and relay topology, it evaluates the connectivity strength of platform i at slot q to the upper layer and neighboring areas, and introduces a traversal attenuation term for the near-boundary region. The indices of all three matrices are consistent with the slot set, ensuring element-wise alignment with R in subsequent iterations.

[0107] To address multi-source uncertainty, this embodiment employs triangular fuzzy numbers for weight normalization. Triangular fuzzy numbers Ãs, Ãv, and Ãc are assigned for detection, survival, and communication, respectively. The parameters (l, m, u) are derived from joint assessments of mission phases and environmental observations. For example, in heavy fog and with increased collision warning frequency, the m and u values ​​of Ãv are increased; in the search phase, Ãs is correspondingly increased. After constructing a fuzzy comparison matrix, consistency checks and defuzzification are performed to obtain the clear weight w = [ws, wv, wc], which is then normalized to a sum of 1. This weight reflects natural causality: increased threat → survival priority, improved line-of-sight → amplified detection gain, weakened backbone link → increased communication weight to stabilize connectivity.

[0108] In step S402, this embodiment performs a tensor product of w and R to establish a potential energy distribution map. For each slot q, the detection gain Fs(q), survivability Fv(q), and connectivity gain Fc(q) are calculated. The slots are then embedded into the spatial grid using the dx, dy, and dz values ​​provided by R, forming a scalar field J(q) = ws·Fs(q) + wc·Fc(q). wv·V(q). The potential energy map is interpolated at ΩH and ΩV respectively, and then synthesized into a three-dimensional field. The inaccessible region is set as a high-potential barrier, and a soft penalty term is added to the curvature-inaccessible region. Then, the negative gradient is calculated. ( J) serves as the preferred direction field, upon which the topology of the basic formation is determined: if the gradient is concentrated near the main heading and the lateral component is weak, it is biased towards a linear shape; if the gradient diverges outward in a sector and the boundary potential energy is low, it is biased towards a sector / wedge shape; if the central potential energy pit is deep and the periphery is gentle, it is biased towards a rhombus shape to expand around the backbone.

[0109] In the combinatorial optimization phase, this embodiment uses basic formation primitives as building blocks, which are binned onto the potential energy map. The objective function is a weighted sum of maximizing the field value integral and minimizing the constraint penalty. The constraints include: the tolerance band of R, the height layer and obstacle edge of Mf, strong connectivity maintenance of the communication community, and curvature reachability of the fixed wing. The search process does not rely on black-box learning, but uses a hybrid of heuristic neighborhood exchange and small-scale particle swarm optimization: neighborhood exchange is responsible for discrete slot swapping, while particle swarm optimization is responsible for continuous fine-tuning of small deviations. The two are updated alternately to balance convergence and feasibility. The output is the formation transformation scheme, containing "primitive type - slot set - hierarchy label - transition trajectory", along with a version stamp and trigger reason, which is then passed to the navigation planning unit.

[0110] To better understand the meaning of potential energy and gradient, here is an explanatory expression:

[0111] J(q) = ws·Fs(q) + wc·Fc(q) wv·V(q),

[0112] Where q is the slot index; Fs(q) is the detection coverage gain corresponding to the slot (obtained by weighting the target sector using the S matrix); Fc(q) is the link connectivity gain (obtained by summing the upper layer and neighboring connectivity strengths using the C matrix); V(q) is the risk cost (obtained by comprehensively considering factors such as obstacles, curvature, and wave height using the V matrix); ws, wc, and wv are weighting coefficients. A larger J(q) indicates a higher overall efficiency for deploying members in that slot.

[0113] In a typical segment, the formation needs to traverse a narrow corridor between two reefs. V rises sharply near the boundary to form a "potential barrier," and the equipotential lines of J are lengthened along the central axis of the corridor. The gradient is more consistent in the axial direction. The system selects a linear shape as the main feature and inserts short wedges locally to maintain detection coverage. In another segment, the target sector is searched at night. Visibility fluctuates, but the backbone link is stable. S rises on the target sector, and J forms a "potential valley" at the edge of the sector. The topology discrimination tends to unfold in a fan-shaped / diamond shape. The air detachment rises to the low-risk layer on ΩV, and the surface relay maintains strong connectivity around the geometric center.

[0114] From a technical and effectiveness perspective, S401 compresses multi-source uncertainties into a unified potential energy framework, avoiding instability caused by manual threshold switching. S402 uses tensor fusion and gradient guidance to transform the abstract concept of "good or bad" into concrete "shape and position," outputting a verifiable and traceable formation transformation scheme without introducing a black-box model. This scheme seamlessly integrates with S102 weight calculation and S103 potential field control law: changes in weights are immediately reflected in the potential energy diagram shape, and the control law generates the resultant force direction according to the negative gradient. By limiting turning and speed at the execution layer, a smooth transition can be achieved, adapting to the rapid self-organizing reconstruction needs under sudden threats.

[0115] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0116] Step S501: Based on the target evaluation theory, construct an evaluation index system, normalize the detection efficiency index, survival efficiency index, and communication efficiency index, perform hierarchical analysis on the basic formation combination scheme, and generate an evaluation matrix containing weight coefficients.

[0117] Step S502: The evaluation matrix is ​​iteratively solved using the particle swarm optimization algorithm to calculate the comprehensive efficiency coefficient of different formation configurations. The optimal formation configuration is selected based on the comprehensive efficiency coefficient, and the efficiency value of the optimal formation configuration is mapped to a navigation reference index.

[0118] This embodiment focuses on formation assessment and decision-making in the air-surface-submarine three-domain coordination at sea, unfolding steps S501 to S502. It builds upon the basic formation combination schemes and potential energy distribution maps generated in steps S401 to S402, maintaining consistency with the situational field weights in step S102. The problem to be solved is to transform the comparison of the advantages and disadvantages of three heterogeneous indicators—detection, survivability, and communication—under different formations into a calculable and traceable comprehensive performance ranking. Furthermore, the performance of the optimal formation is quantified into reference indicators that can be used by the navigation planning unit, avoiding manual thresholds and experience-based decisions.

[0119] This embodiment constructs an evaluation index system in S501. The index selection follows the objective evaluation theory, with the top-level objective being "comprehensive formation effectiveness," and the criterion layer including detection effectiveness, survivability effectiveness, and communication effectiveness. When necessary, constraint layers (such as maneuverability and inter-layer connectivity redundancy) are introduced as penalty terms. For candidate basic formation combinations, the detection effectiveness Ds is obtained by summarizing the effective detection coverage area and overlap penalty from the S matrix; the survivability effectiveness Sv is obtained by statistically integrating the risk cost in space from the V matrix and taking the negative index; and the communication effectiveness Cm is obtained by evaluating the composite value of backbone connectivity strength and average hop count from the C matrix. Since the three original dimensions are different, this embodiment performs unified normalization processing, adopts interval scaling, and introduces robust quantile truncation to avoid extreme value bias. Regarding weights, instead of directly reusing fixed values, a weight vector w=[ws,wv,wc] is generated based on the defuzzification result of the triangular fuzzy number of S401, and a hierarchical analysis consistency check is performed: a comparison matrix based on task state and environment state is constructed, and after verifying consistency, an evaluation matrix E containing weight coefficients is output. The rows of E correspond to the scheme, and the columns correspond to the weighted component values ​​of the three types of effectiveness.

[0120] In implementing analytic hierarchy process, this embodiment avoids excessive subjectivity. Prior weights are derived from fuzzy numbers (reflecting uncertainty), and consistency is judged by the ratio of the matrix's largest eigenvalue to the consistency ratio. If the threshold is exceeded, the fuzzy triples are adjusted back to the situation source data until stability is achieved. The evaluation matrix also includes a constraint mask: if a formation has a hard conflict with curvature reachability or height layer prohibitions, a penalty is marked in E to prevent it from being misjudged as high-efficiency in subsequent searches.

[0121] In S502, a particle swarm optimization algorithm is used to iteratively solve the evaluation matrix. The particle position is encoded as a hybrid vector of "formation primitive selection × slot assignment × micro-bias compensation," and the velocity is the corresponding step size. The fitness function is defined as the overall performance coefficient J = ws·Ds + wv·Sv + wc·Cm λ·P, where P is the constraint penalty (unreachability, connectivity disruption, overcrowding), and λ is the penalty weight. To ensure physical rationality, the velocity update uses a hierarchical inertia setting: dimensions related to the communication backbone use smaller inertia to prevent frequent jumps from disrupting connectivity; dimensions related to probe deployment allow for larger exploration to cover sectors. Population initialization is based on the local optimal neighborhood of the S402 potential energy graph to avoid random starts and taking long paths. The convergence criterion uses a dual condition of fitness improvement threshold and algebraic upper limit.

[0122] The input / output of the particle swarm optimization is defined according to the scenario: the input is the evaluation matrix E, weights w, constraint set, and slot set; the output is the sequence of comprehensive performance coefficients for each candidate formation configuration and the parameter vector corresponding to the optimal solution. The algorithm does not introduce black-box learning; instead, it uses explicit metrics and penalties to constitute the objective function. The correlation between metrics and results follows natural rules: the more comprehensive the detection coverage, the lower the risk cost, and the more robust the connectivity, the higher the comprehensive performance. To avoid local optima, this embodiment sets a small number of "global exploration particles" and periodically resamples them according to the gradient troughs of the potential energy map; when sudden scene changes cause rapid weight changes, inertial decay is increased to accelerate reconvergence.

[0123] To enable navigation planning to directly consume optimal results, this embodiment maps the performance value of the optimal formation configuration to a navigation reference index set Rnav, which includes three levels: first, the global reference strength Knav, which serves as the baseline gain of the artificial potential field U_g in S103; second, the topology preservation coefficient Kf, used for the trade-off of the swarm component U_f, ensuring that the high-performance formation is not easily broken during the transition period; and third, the risk buffer coefficient Ko, which acts on the obstacle avoidance component U_o, reflecting the outer-loop adjustment of the survival weight. The mapping follows a monotonic relationship to avoid overweighting: for example, if Sv is high, Ko increases, allowing the guiding control law to reserve a larger safety margin; if Cm is low, connected sensitive slots are marked in Rnav, and subsequent control imposes stricter limits on the maneuver amplitude of these slots.

[0124] Two examples illustrate the mechanism's operation. When traversing a narrow channel, the V field increases dramatically on both sides, Sv dominates in E, the particle swarm converges to a predominantly linear solution, Knav decreases and Ko increases, and the navigation reference guides the formation to converge through the corridor, keeping the relay near the geometric center. During nighttime sector searches, the S field increases in the target sector, the weight ws rises, the particle swarm tends towards sector / rhombus combinations, Knav in Rnav increases to encourage expansion towards the sector, and Kf maintains the local topology to prevent the air swarm from becoming disconnected from the surface relay.

[0125] From a technical perspective, S501 provides a clearly sourced and verifiable evaluation matrix, making the evaluation of abstract formations verifiable. S502, through particle swarm optimization in a hybrid space, integrates potential energy, weights, and constraints into a comprehensive performance coefficient, and translates the results into navigation reference indicators for direct use by the control law. Under conditions of sudden threats and sea state fluctuations, this link can stably output executable optimal formations and associated parameters, reducing human intervention and handover jitter, and ensuring a balance between formation mission objectives and survivability requirements.

[0126] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0127] Step S601: Construct a spatial potential energy function based on the formation configuration effectiveness value, map the distance between unmanned platforms as a clustering potential field component, map the position of obstacles as a repulsive potential field component, map the position of the target point as an attractive potential field component, and perform superposition calculation on the potential field components based on the gradient descent method to generate an artificial potential field function.

[0128] Step S602: Construct a motion control law based on the negative gradient direction of the artificial potential field function, transform the aggregation potential field component into a cluster control term, the repulsive potential field component into an obstacle avoidance control term, and the attractive potential field component into a navigation control term, and perform weighted fusion on the control terms to generate a synthetic control force vector.

[0129] This embodiment builds upon the potential energy distribution and optimal formation transformation scheme generated in S401-S402. In S601-S602, the conclusion of "effectiveness assessment - geometric constraints" is pushed down to a continuous artificial potential field and control terms. The goal is to achieve both steady-state formation and continuous maneuverability to avoid sudden threats in a maritime air-surface-submarine coordinated environment. Inputs include: formation configuration effectiveness values ​​and their weighted snapshots w=[ws,wv,wc], position parameter matrix R (dx, dy, dz and tolerance zone), obstacle domain and altitude layer constraints in the formation representation matrix, target point and mission main heading information, and adjacency and curvature reachability constraints.

[0130] In S601, this embodiment first sets the gain of each component of the potential field according to the formation configuration effectiveness value. The aggregation potential field U_f is used to maintain the consistency of the desired relative distance and velocity. Its core comes from the Olfati-Saber framework: for each platform i and its neighbor j, a deviation energy based on the smoothing norm is constructed to make the distance approach the desired d and the velocity aligned, avoiding the divergence of the derivative when the distance is close to zero; the obstacle avoidance potential field U_o maps environmental obstacles and threats as repulsive forces. The obstacle geometry is modeled in two types - the hyperplane boundary is defined by the normal and the passing point, and the circular obstacle is defined by the center and radius. A virtual node q_i is constructed for each platform to fall on the edge of the nearest obstacle. q_i moves along the tangential direction, so that a natural resultant force of "skimming along the edge" is formed after the potential fields are superimposed; the navigation potential field U_g uses the mission target and the main heading as the attraction source. Combined with the formation transformation scheme output by S402, the target geometric center and axial priority of the current stage are determined. To incorporate performance values ​​into potential field calibration, this embodiment maps the survival weight in w to the gain of U_o. The detection and communication weights jointly adjust the proportions of U_f and U_g, emphasizing repulsion and cohesion during the danger period and enhancing attraction and coverage during the search period.

[0131] The potential field components must be constructed separately in two planes, ΩH (horizontal) and ΩV (vertical), and then synthesized in three dimensions. This is because the reachability curvature and altitude-level no-entry constraints of fixed-wing aircraft and surface vessels are not symmetrical. The calculation process is as follows: Based on the deviation and tolerance zone given by R, construct U_f^H on ΩH using adjacency sets, and construct U_f^V on ΩV using altitude layers and minimum intervals. For U_o, establish distance fields in both planes based on obstacle projections, setting the no-entry region as a high-potential barrier, and adding soft penalties to the near-boundary region to avoid "being sucked into the wall." For U_g, set axial attraction in ΩH according to the mission's main heading, and set altitude-level target attraction in ΩV. After weighted synthesis of the potential fields in the two planes, the three-dimensional U = U_f + U_o + U_g is obtained. To preserve differentiability, all distances use a smoothing norm instead of the Euclidean norm; and the potential energy slope is gradually increased as the tolerance zone boundary approaches, forming a buffer zone to avoid drastic fluctuations in the control quantity.

[0132] Entering S602, this embodiment uses the negative gradient of the artificial potential field function to generate the control direction. The composite control force F_i is calculated for platform i. U(r_i) is further decomposed into three types of control terms: the clustering control term F_i^flock, derived from U_f, reflects clustering and alignment, with its direction towards the desired adjacent configuration; the obstacle avoidance control term F_i^obs, derived from U_o, moves away from the normal risk along the tangent of the nearest obstacle; and the navigation control term F_i^nav, derived from U_g, points towards the task's geometric center and the next stage slot reference. These three terms are not simply added together, but rather weighted and fused based on w and the platform state, with additional constraints on yaw amplitude in the connecting backbone direction to avoid relay breakage. For ease of engineering implementation, this embodiment provides an explanatory formula for the control quantity mapping:

[0133] u_i=[dot{ψ}_i, a_i]=[k_ψ·wrap(arg(F_i^H) ψ_i), k_a·||F_i||],

[0134] Where dot{ψ}_i is the rate of change of the heading angle, a_i is the acceleration command; F_i^H is the projection of F_i onto the horizontal plane; arg(·) outputs the horizontal azimuth angle; ψ_i is the current heading angle; wrap(·) normalizes the angle difference to [ [π,π]; k_ψ and k_a are the gains corresponding to the platform response characteristics. Each parameter in the formula has a clear physical meaning: the greater the horizontal azimuth difference, the higher the turning rate; the greater the resultant force amplitude, the stronger the acceleration, but it will be kinematically limited subsequently.

[0135] The inherent correlation between weights and control terms follows natural laws: when sea conditions worsen or threats approach, wv increases after S401 defuzzification, U_o gain is increased, and F_i^obs weight increases. The control law, in the same scenario, tends to maintain distance and avoid threats. When performing search subtasks, ws increases, U_g and sub-terms associated with detection coverage are enhanced, F_i^nav weight increases, and air units are guided to higher dz slots to gain a top-down view. However, the tolerance zone in R suppresses excessively rapid ascents, ensuring that curvature and energy constraints do not exceed limits. When the communication weight wc increases, swarm control terms are given higher priority on strongly connected edges, and the control law reduces lateral maneuvers that conflict with the backbone, maintaining network connectivity.

[0136] This embodiment performs triple limiting and consistency checks after generating u_i. Limiting includes mapping the minimum turning radius to the upper limit of |dot{ψ}_i|, the velocity boundary to the upper limit of |a_i|, and the height layer boundary to the upper limit of dz variation. Consistency checks include obstacle intrusion domain verification and formation confirmation closed-loop verification. If F_i is detected, it will guide the platform out of bounds. The system temporarily increases the repulsive force weight of that domain in the potential field or projects and truncates F_i to ensure the physical reachability of the control quantity. Considering sensor noise and link delay, u_i uses first-order filtering and hysteresis thresholding before being sent to avoid rapid weight fluctuations causing repeated switching of maneuver direction.

[0137] From a technical and effectiveness perspective, the artificial potential field established by S601 projects the potential energy evaluation of S402 into a continuous mechanical quantity, avoiding artificial threshold switching. S602 strictly maps the three types of potential field components to the "swarming-obstacle avoidance-navigation" control terms and fuses them with weights and physical constraints to form a synthetic control force that is both interpretable and achievable. The control commands are ultimately written into the execution control unit, where the lower-level controllers of the fixed-wing, unmanned surface vessel, and AUV generate control surfaces and thrust. Because the upstream R matrix and weights change slowly with the environmental state, the synthetic control force is smooth in time, and drastic rearrangement is less likely to occur during formation changes, meeting the requirements for self-organized navigation and obstacle avoidance under sudden maritime threats.

[0138] In one embodiment of the unmanned swarm self-organizing navigation and obstacle avoidance control method of this application, it may further include the following:

[0139] Step S701: Read the heading angle and motion rate information of the unmanned platform, calculate the desired motion direction based on the synthetic control force vector, perform a difference calculation between the desired motion direction and the current heading angle, and generate heading angle change rate and acceleration commands;

[0140] Step S702: Based on the kinematic constraints of the unmanned platform, the rate of change of the heading angle and the acceleration command are subjected to amplitude limiting processing, the minimum turning radius is mapped to the upper limit of angular velocity, the velocity boundary is mapped to the upper limit of acceleration, and motion control commands are generated according to the amplitude limiting results.

[0141] This embodiment takes the synthetic control force vector F_i output from S601 to S602 and the three-dimensional deviation and tolerance band in the R matrix. Around S701 to S702, it translates the "continuous potential field control quantity" into heading and acceleration commands executable by various platforms. It then limits the amplitude under different kinematic constraints for fixed-wing aircraft, surface vessels, and AUVs, ultimately generating motion control commands that are written to the execution control unit. The processing link must ensure two points: first, the control direction must be consistent with the three objectives of mission main heading, formation maintenance, and obstacle avoidance; second, the command amplitude must not exceed the platform's minimum turning radius, velocity boundaries, and acceleration / deceleration capability boundaries to avoid aerodynamic / hydrodynamic instability.

[0142] In this embodiment, state reading and direction calculation are performed first in S701. The execution control unit periodically reads the current heading angle ψ_i and ground speed v_i of platform i, and performs circumangular normalization on ψ_i to avoid sudden jumps caused by angle reversal. F_i is projected onto the horizontal plane ΩH to obtain F_i^H, and then its azimuth angle φ_i=arg(F_i^H) is calculated. This azimuth angle represents the "desired trajectory orientation" under the current potential field synthesis. To take into account altitude layer commands, the potential field components related to dz are not directly converted into heading, but are processed by the longitudinal velocity / heave channel (for fixed wings, the climb rate constraint is used; for underwater AUVs, the heave rudder / ballast is used). Then, the difference between the desired direction and the current heading Δψ_i=wrap(φ_i) is calculated. The resultant force amplitude ||F_i|| is mapped to the target acceleration reference a_i^ref, and the raw control quantities are generated through the platform response gains k_ψ and k_a: dot{ψ}_i^ref=k_ψ·Δψ_i, a_i^raw=k_a·||F_i||. In humid sea conditions or during periods of link jitter, the system makes gradual adjustments to k_ψ and k_a based on the tolerance band of R and the weighted snapshot to suppress connectivity breaks and energy waste caused by high-frequency maneuvers.

[0143] To improve robustness, this embodiment performs two consistency processing steps before direction calculation. First, adjacency consistency: The strongly connected edge set in the adjacency matrix is ​​read, and the constraints on these edge directions are limited by an angle window. When Δψ_i causes significant stretching of the backbone edge, k_ψ is dynamically reduced, and a small-angle projection is made onto φ_i to maintain the steady state of the communication backbone. Second, curvature reachability pre-check: The achievable instantaneous turning radius range is estimated based on the platform speed v_i and checked against the "curvature tight constraint" slot marked in R. If it is about to enter the tight constraint region, a_i^raw is reduced in advance to release the turning margin. These processing steps avoid excessive saturation clipping during the limiting phase of S702, maintaining the interpretability and smoothness of the control.

[0144] In this embodiment, amplitude limiting and command generation are performed in S702, following the sequence of "geometric reachability first, dynamic reachability second, and safety redundancy third." The geometric reachability stage handles the constraint of the minimum turning radius R_min: for fixed-wing aircraft and surface vessels, the upper limit of angular velocity is given by ω_max = v_i / R_min; for AUVs, ω_max can be increased at low speeds, but is constrained by the attitude stabilizer. The dynamic reachability stage handles velocity and acceleration boundaries: for airspace platforms, the lower limit of velocity is determined by the stall margin, and the upper limit is given by the mission speed band and airspace constraints; the upper limit of acceleration is related to the platform thrust / controller efficiency, and is reduced by a safety factor in severe sea conditions. The safety redundancy stage converts the altitude layer of ΩV (forbidden / reserved layers) into soft and hard thresholds for dz variation. If the tangential component of a_i^raw induces excessively rapid ascent / descent, the corresponding component of a_i^raw is reduced, extending the longitudinal transition time. After amplitude limiting is completed, the command pair dot{ψ}_i and a_i is obtained.

[0145] This embodiment provides a constraint mapping formula for engineering use:

[0146] ω_max=v_i / R_min,u_i=[dot{ψ}_i,a_i]=[clip(k_ψ·Δψ_i, ω_max,ω_max),clip(k_a·||F_i||, a_max, a_max)).

[0147] Where ω_max is the upper limit of angular velocity, which is related to the current velocity v_i and the platform's minimum turning radius R_min; clip(·) is the amplitude limiting operator; a_max is the upper limit of the acceleration amplitude allowed by the platform in the current state; Δψ_i is the difference between the expected and current heading angles; k_ψ and k_a are the response gains. Each parameter in this formula has a clear physical meaning and can be obtained from the platform's capability library and sensor estimation, reflecting that "the higher the speed, the larger the turning radius, and the smaller the allowed angular velocity."

[0148] Considering the uncertainties in perception and communication, this embodiment performs low-pass filtering on ψ_i, v_i, and F_i before amplitude limiting, with the filtering time constant adjusted according to sea state and link stability. After amplitude limiting, a hysteresis threshold is introduced for u_i, preventing small directional fluctuations from triggering command updates and avoiding control surface chattering. For individuals that fall behind or temporarily lose contact, the execution control unit reads the "pending takeover" flag in R, downgrades the swarm weight, retains only the minimum requirements for obstacle avoidance and navigation, and further lowers the amplitude limits for angular velocity and acceleration, ensuring that it returns to the communicable area with a conservative strategy before restoring full weight.

[0149] This embodiment illustrates the process details in two segments. During the narrow channel traversal phase, speed is limited and R_min is large, while ω_max is small. The calculated Δψ_i is close to the corridor's main axis. After limiting, dot{ψ}_i maintains a small, continuous turn, and a_i is reduced to release turning margin. The platform traverses with a "slow turn and slow move," and adjacency consistency keeps the relay near the geometric center. During the nighttime sector search phase, the navigation terms of the airborne squadron F_i are raised, and Δψ_i shifts towards the sector. Due to the upward movement of targets at higher altitudes, the upper limit of dz variation in the longitudinal channel takes effect. The system releases the vertical component of acceleration in stages to avoid short-term overload. Surface relays, due to their higher communication weight, obtain a stricter angular velocity upper limit in the limiter, reducing lateral maneuvering.

[0150] In connection with the preceding steps, S701 directly consumes the F_i and R matrices from S602 and generates a reference value for the controllable channel according to the preferred direction given by the negative gradient of the potential field. S702 then projects the reference value into the reachable domain of the platform to form the actual execution command. The generated command frame contains dot{ψ}_i, a_i, validity period and version stamp, and includes the trigger reason (such as "weight switching" or "obstacle approaching"). The execution control unit uses this to generate control surface / throttle / thrust allocation.

[0151] To effectively address the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, and to provide technical support for unmanned swarm collaboration, this application provides an embodiment of an unmanned swarm self-organizing navigation and obstacle avoidance control device for implementing all or part of the aforementioned unmanned swarm self-organizing navigation and obstacle avoidance control method. See [link to embodiment]. Figure 2 The unmanned swarm self-organizing navigation and obstacle avoidance control device specifically includes the following components:

[0152] The formation control module 10 is used to decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0153] The organization navigation module 20 is used to construct a situation field model that includes detection capability, survivability capability, and communication capability. It uses the triangular fuzzy number method to calculate the capability weight coefficients, performs matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme, calculates the formation configuration effectiveness value based on the basic formation combination scheme, and writes the formation configuration effectiveness value into the navigation planning unit.

[0154] The obstacle avoidance control module 30 is used to generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command according to the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0155] As described above, the unmanned swarm self-organizing navigation and obstacle avoidance control device provided in this application embodiment can achieve effective formation organization through innovative formation configuration system design, hierarchical control, and position parameters. It constructs a situation assessment mechanism, combining capability weights and performance analysis to establish a reliable configuration scheme. By introducing potential field control and optimizing commands through multi-dimensional control, it ensures navigation accuracy. This method effectively solves the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0156] From a hardware perspective, in order to effectively address the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, and to provide technical support for unmanned swarm collaboration, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned unmanned swarm self-organizing navigation and obstacle avoidance control method. The electronic device specifically includes the following components:

[0157] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the unmanned swarm self-organizing navigation and obstacle avoidance control device and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the unmanned swarm self-organizing navigation and obstacle avoidance control method and the unmanned swarm self-organizing navigation and obstacle avoidance control device in the embodiments, the content of which is incorporated herein, and repeated details will not be described again.

[0158] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0159] In practical applications, parts of the unmanned swarm self-organizing navigation and obstacle avoidance control method can be executed on the electronic device side as described above, or all operations can be completed in the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0160] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0161] Figure 3 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 3 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 3 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0162] In one embodiment, the unmanned swarm self-organizing navigation and obstacle avoidance control method function can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0163] Step S101: Decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0164] Step S102: Construct a situation field model that includes detection capability, survivability, and communication capability; calculate capability weight coefficients using the triangular fuzzy number method; perform matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme; calculate the formation configuration effectiveness value based on the basic formation combination scheme; and write the formation configuration effectiveness value into the navigation planning unit.

[0165] Step S103: Generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command based on the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0166] As described above, the electronic device provided in this application, through an innovative design of a formation configuration system, achieves effective formation organization through hierarchical control and position parameters. A situation assessment mechanism is constructed, and a reliable configuration scheme is established by combining capability weights and performance analysis. Potential field control is introduced, and navigation accuracy is ensured through multi-dimensional control and command optimization. This method effectively solves the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0167] In another embodiment, the unmanned swarm self-organizing navigation and obstacle avoidance control device can be configured separately from the central processing unit 9100. For example, the unmanned swarm self-organizing navigation and obstacle avoidance control device can be configured as a chip connected to the central processing unit 9100, and the unmanned swarm self-organizing navigation and obstacle avoidance control method function can be realized through the control of the central processing unit.

[0168] like Figure 3As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 3 All components shown; in addition, the electronic device 9600 may also include Figure 3 For components not shown, please refer to existing technologies.

[0169] like Figure 3 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0170] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0171] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0172] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0173] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device for communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0174] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0175] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 (transmitter / receiver) is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0176] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method with a server or client as the execution subject in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0177] Step S101: Decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0178] Step S102: Construct a situation field model that includes detection capability, survivability, and communication capability; calculate capability weight coefficients using the triangular fuzzy number method; perform matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme; calculate the formation configuration effectiveness value based on the basic formation combination scheme; and write the formation configuration effectiveness value into the navigation planning unit.

[0179] Step S103: Generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command based on the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0180] As described above, the computer-readable storage medium provided in this application embodiment achieves effective formation organization through an innovatively designed formation configuration system and hierarchical control and position parameters. It constructs a situation assessment mechanism, combining capability weights and performance analysis to establish a reliable configuration scheme. Furthermore, it introduces potential field control, ensuring navigation accuracy through multi-dimensional control and command optimization. This method effectively addresses the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0181] Embodiments of this application also provide a computer program product capable of implementing all steps in the unmanned swarm self-organizing navigation and obstacle avoidance control method in the above embodiments, where the execution subject is a server or client. When executed by a processor, this computer program / instruction implements the steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method. For example, the computer program / instruction implements the following steps:

[0182] Step S101: Decompose the three-dimensional formation configuration into horizontal plane formation and vertical plane formation, divide the unmanned cluster into cross-domain control layer, group control layer and platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation and lateral deviation, and write the position parameter matrix into the formation control unit;

[0183] Step S102: Construct a situation field model that includes detection capability, survivability, and communication capability; calculate capability weight coefficients using the triangular fuzzy number method; perform matching operations between the situation field model and the position parameter matrix to generate a basic formation combination scheme; calculate the formation configuration effectiveness value based on the basic formation combination scheme; and write the formation configuration effectiveness value into the navigation planning unit.

[0184] Step S103: Generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command based on the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

[0185] As described above, the computer program product provided in this application, through an innovative design of a formation configuration system, achieves effective formation organization through hierarchical control and position parameters. It constructs a situation assessment mechanism, combining capability weights and performance analysis to establish a reliable configuration scheme. By introducing potential field control and optimizing commands through multi-dimensional control, it ensures navigation accuracy. This method effectively addresses the shortcomings of traditional technologies in formation configuration, situation assessment, and motion control, providing technical support for unmanned swarm collaboration.

[0186] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0187] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for self-organizing navigation and obstacle avoidance control of an unmanned swarm, characterized in that, The method includes: The three-dimensional formation configuration is decomposed into horizontal plane formation and vertical plane formation. The unmanned swarm is divided into cross-domain control layer, group control layer and platform execution layer according to the hierarchy. The encoding sequence is generated according to the hierarchical formation rules. The encoding sequence is parsed to obtain the basic formation combination. The spatial position difference between the unmanned platform and the geometric center is calculated to generate a position parameter matrix containing axial deviation, longitudinal deviation and lateral deviation. The position parameter matrix is ​​smoothed by setting a sliding window. The position parameter matrix is ​​written into the formation control unit. A multi-dimensional situational awareness model is constructed based on the state vector space. Detection capability indicators are mapped to a perception range matrix, survivability indicators to a threat avoidance matrix, and communication capability indicators to a link connectivity matrix. The matrices are normalized using the triangular fuzzy number method to generate situational awareness weight coefficients. These weight coefficients are then multiplied by the position parameter matrix using tensors to establish a potential energy distribution map of the unmanned swarm formation configuration. The topology of the basic formation is determined based on the potential energy gradient direction. The basic formation is then optimized to generate a combination scheme, which is then used for target assessment. A theoretical evaluation index system is constructed, and the detection efficiency index, survival efficiency index, and communication efficiency index are normalized. A hierarchical analysis is performed on the basic formation combination scheme to generate an evaluation matrix containing weight coefficients. A particle swarm optimization algorithm is used to iteratively solve the evaluation matrix, calculating the comprehensive efficiency coefficient of different formation configurations. Based on the comprehensive efficiency coefficient, the optimal formation configuration is selected, and the efficiency value of the optimal formation configuration is used as the formation configuration efficiency value. This efficiency value is then mapped to a navigation reference index, which is written into the navigation planning unit. An artificial potential field function is generated based on the formation configuration performance value. A motion control law containing swarm control components, obstacle avoidance control components, and navigation control components is constructed based on the artificial potential field function. The heading angle and motion rate of the unmanned platform are read. The heading angle change rate and acceleration command are calculated based on the motion control law. The heading angle change rate and acceleration command are written into the execution control unit.

2. The unmanned swarm self-organizing navigation and obstacle avoidance control method according to claim 1, characterized in that, The process involves decomposing the 3D formation configuration into horizontal and vertical planar formations, dividing the unmanned swarm into a cross-domain control layer, a group control layer, and a platform execution layer, and generating a coding sequence based on hierarchical formation rules, including: The coordinate axis of the three-dimensional formation configuration is decomposed according to the orthogonal projection principle, the projection vectors of the horizontal and vertical planes are calculated, and a formation representation matrix containing the planar configuration coordinates, distribution range and constraints is generated based on the orthogonal basis transformation. The spatial distribution position of the unmanned cluster is mapped according to the formation representation matrix. The unmanned swarm control hierarchy is divided based on a tree structure model. Gateway nodes are assigned to the cross-domain control layer, perception nodes are assigned to the group control layer, and execution nodes are assigned to the platform execution layer. Encoding sequences are generated according to the command relationships between the levels, and the encoding sequences are used to describe the unmanned swarm formation configuration.

3. The unmanned swarm self-organizing navigation and obstacle avoidance control method according to claim 1, characterized in that, The process of parsing the encoded sequence to obtain the basic formation combination, calculating the spatial position difference between the unmanned platform and the geometric center, generating a position parameter matrix including axial deviation, longitudinal deviation, and lateral deviation, and writing the position parameter matrix into the formation control unit includes: The encoded sequence is decoded according to the hierarchical formation rules, the topological structure of the basic formation unit is extracted, the basic formation unit is combined according to the command relationship to generate the target formation, and the geometric center point of the unmanned swarm is calculated according to the target formation. The relative distance between the unmanned platform and the geometric center point is calculated based on the vector decomposition method. Axial deviation components, longitudinal deviation components, and lateral deviation components are obtained by projection along the coordinate axes. A position parameter matrix is ​​constructed based on the deviation components, and the position parameter matrix is ​​used to describe the spatial positional relationship of the unmanned cluster.

4. The unmanned swarm self-organizing navigation and obstacle avoidance control method according to claim 1, characterized in that, The step of generating an artificial potential field function based on the formation configuration performance value, and constructing a motion control law based on the artificial potential field function that includes swarm control components, obstacle avoidance control components, and navigation control components, includes: Based on the formation configuration effectiveness value, a spatial potential energy function is constructed, the distance between unmanned platforms is mapped to a clustering potential field component, the position of obstacles is mapped to a repulsive potential field component, the position of the target point is mapped to an attractive potential field component, and the potential field components are superimposed and calculated based on the gradient descent method to generate an artificial potential field function. Motion control laws are constructed based on the negative gradient direction of the artificial potential field function. The aggregation potential field component is transformed into a cluster control term, the repulsive potential field component is transformed into an obstacle avoidance control term, and the attractive potential field component is transformed into a navigation control term. The control terms are then weighted and fused to generate a synthetic control force vector.

5. The unmanned swarm self-organizing navigation and obstacle avoidance control method according to claim 4, characterized in that, The process of reading the unmanned platform's heading angle and velocity, calculating the heading angle change rate and acceleration command based on the motion control law, and writing the heading angle change rate and acceleration command into the execution control unit includes: Read the heading angle and motion rate information of the unmanned platform, calculate the desired motion direction based on the synthetic control force vector, perform a difference calculation between the desired motion direction and the current heading angle, and generate heading angle change rate and acceleration commands; Based on the kinematic constraints of the unmanned platform, the rate of change of the heading angle and the acceleration command are limited, the minimum turning radius is mapped to the upper limit of angular velocity, the velocity boundary is mapped to the upper limit of acceleration, and motion control commands are generated based on the limiting results.

6. A self-organizing navigation and obstacle avoidance control device for unmanned swarms, characterized in that, The device includes: The formation control module is used to decompose the three-dimensional formation configuration into horizontal and vertical planar formations, divide the unmanned swarm into a cross-domain control layer, a group control layer, and a platform execution layer according to the hierarchy, generate a coding sequence according to the hierarchical formation rules, parse the coding sequence to obtain the basic formation combination, calculate the spatial position difference between the unmanned platform and the geometric center, generate a position parameter matrix including axial deviation, longitudinal deviation, and lateral deviation, smooth the position parameter matrix by setting a sliding window, and write the position parameter matrix into the formation control unit. The organization and navigation module is used to construct a multi-dimensional situational field model based on the state vector space. It maps detection capability indicators to a perception range matrix, survivability indicators to a threat avoidance matrix, and communication capability indicators to a link connectivity matrix. The matrix is ​​normalized using a triangular fuzzy number method to generate situational field weight coefficients. These weight coefficients are then multiplied by the position parameter matrix using tensors to establish a potential energy distribution map of the unmanned swarm formation configuration. The topology of the basic formation is determined based on the potential energy gradient direction, and the basic formation is combined and optimized to generate a basic formation combination. The scheme constructs an evaluation index system based on target evaluation theory, normalizes detection efficiency, survival efficiency, and communication efficiency indicators, performs hierarchical analysis on the basic formation combination scheme, and generates an evaluation matrix containing weight coefficients. A particle swarm optimization algorithm is used to iteratively solve the evaluation matrix, calculates the comprehensive efficiency coefficient of different formation configurations, selects the optimal formation configuration based on the comprehensive efficiency coefficient, uses the efficiency value of the optimal formation configuration as the formation configuration efficiency value, and maps the formation configuration efficiency value of the optimal formation configuration to a navigation reference index. The obstacle avoidance control module is used to generate an artificial potential field function based on the formation configuration performance value, construct a motion control law containing swarm control components, obstacle avoidance control components, and navigation control components based on the artificial potential field function, read the heading angle and motion rate of the unmanned platform, calculate the heading angle change rate and acceleration command according to the motion control law, and write the heading angle change rate and acceleration command into the execution control unit.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the unmanned swarm self-organizing navigation and obstacle avoidance control method according to any one of claims 1 to 5.

Images