A method of collaborative assignment-observation-control integration design and operation for many-to-many interception

Abstract

This invention discloses an integrated design method for collaborative allocation, observation, and guidance control in many-to-many interception scenarios, applicable to multi-target engagement scenarios involving heterogeneous interceptor clusters equipped with detectors. The target allocation module performs initial target allocation based on hard constraints, which are determined according to the feasibility of the collaborative observation and guidance control process. The collaborative observation module performs joint collaborative observation combining distributed collaborative observation and leader-based local observation based on the target allocation results, obtaining collaborative observation results. The collaborative guidance control module performs collaborative guidance control based on the collaborative observation results. During the interception process, hard constraint detection is continuously performed; hard constraint repair is performed on target interception groups that do not meet the hard constraints, and dynamic optimization is performed on target interception groups that meet the hard constraints but experience a decrease in effectiveness. This invention can solve the online target allocation problem with partial detector constraints in many-to-many interception, ensuring the feasibility of the target allocation scheme during the collaborative observation and guidance control process.

Description

Technical Field

[0001] This invention belongs to the field of multi-agent cooperative control technology, specifically relating to a cooperative allocation-observation-guided control integrated design and working method for many-to-many interception. Background Technology

[0002] With the continuous advancement of urbanization and the increasing demands for public safety, multiple high-speed, highly mobile potential dynamic targets frequently emerge in civilian fields such as outdoor security for large-scale events, comprehensive urban security, and airspace protection for airports and core areas, requiring coordinated tracking and response. The increasing number of targets, the complexity of their trajectories, and the growing environmental uncertainty render traditional single-platform response methods ineffective. To achieve timely and effective coordinated handling of multiple dynamic targets within a limited time and space window, the parallel deployment and collaborative operation of multiple intelligent agents (especially interceptor swarms such as drones) has become an important technological direction. Through resource coordination, information coordination, and action coordination among interceptors, the overall success rate and mission robustness of the system in responding to target groups can be significantly improved.

[0003] In missions involving multiple interceptor clusters working together to deal with multiple dynamic targets, the system needs to address the following core issues: First, the dynamic task allocation problem, i.e., how to rationally and efficiently allocate multiple interceptors to different targets based on the real-time motion status and threat level of the targets, forming several collaborative task groups; second, the collaborative observation and state estimation problem, i.e., during mission execution, especially under constraints such as field of view, accuracy, or communication range of the detectors carried by some interceptors, how to achieve real-time, accurate, and collaborative perception of the target's motion status through information interaction and fusion within the cluster, providing support for dynamic allocation and precise guidance; third, the integrated collaborative guidance problem, i.e., within each task group, achieving coordinated action of interceptors in time and space, ensuring the synchronization and accuracy of response actions to cope with target maneuvers and environmental disturbances.

[0004] Existing solutions mostly employ a sequential "allocate first, then execute" process, typically with offline or pre-planned missions by a central control station. Each interceptor then acts independently, relying solely on limited local detector information, or engages in simple, fixed information exchange within a group. This approach struggles to adapt to real-time dynamic changes in targets, lacking flexibility. Furthermore, existing technologies generally rely on the idealized assumption that "all executing interceptors possess complete and homogeneous sensing and communication capabilities." However, in practical civilian systems, due to considerations of cost, payload, or functional division, interceptor clusters often consist of a hybrid of heterogeneous platforms, including reconnaissance, communication relay, and response types. Some platforms may lack or not fully possess independent target detection capabilities. This reality makes the assumption that "the target's state is completely known to all platforms" untenable, severely limiting the overall effectiveness and practicality of the system in complex real-world scenarios.

[0005] However, existing solutions still have significant shortcomings in complex adversarial environments. First, existing many-to-many target allocation methods mostly rely on centralized computing or offline pre-planning, often requiring global situational information. When target maneuvers increase or interceptor states change rapidly, the allocation results may become invalid, lacking online rapid adjustment and reliability guarantees. Second, although some solutions introduce distributed allocation mechanisms, they usually decouple allocation from guidance and control, failing to fully consider the impact of allocation results on subsequent observation conditions, communication topology connectivity, and the feasibility of coordinated guidance and control. This easily leads to problems such as "allocation is feasible but guidance and control is not feasible" or "guidance and control are feasible but the overall interception performance of the cluster is poor."

[0006] Furthermore, existing collaborative guidance and control and observation sharing schemes generally implicitly assume that "all members possess detectors or complete measurement capabilities." However, in engineering practice, detectors are critical components that account for a high proportion of cost, size, and power consumption. To achieve cost reduction and efficiency improvement, interceptor clusters need to present a heterogeneous structure where "some interceptors possess detectors, while the rest only possess communication and execution capabilities." Under this partial detector constraint, interceptors lacking detectors cannot directly obtain key end-point guidance and control information such as line-of-sight angle and relative distance, and can only rely on a small number of interceptors with detectors for information sharing or estimation compensation. Existing virtual target guidance and slave-following-leader modes can alleviate measurement gaps to some extent, but they generally suffer from problems such as over-reliance on leader guidance and control, insufficient robustness, and limited adaptability to target maneuvers and dynamic environments. Once key measurement nodes fail or multiple targets maneuver, the overall interception effect degrades significantly.

[0007] Furthermore, in many-to-many cooperative interception, time coordination and precise interception often need to be satisfied simultaneously. Due to high-speed interception, the interception window is very limited, necessitating rapid convergence of the control law. Many existing cooperative guidance and control methods rely excessively on initial interception conditions, linear approximations, and specific motion assumptions. When facing high-speed, highly maneuverable targets, complex geometric relationships, and parameter uncertainties, they are prone to risks such as accumulated estimation biases, overly aggressive control commands, or unstable cooperative effects. Meanwhile, existing research often conducts task allocation, distributed target state estimation, and cooperative guidance and control separately, lacking a scheme for coupled modeling and systematic design of these three elements within a unified framework. This makes it difficult to form a closed-loop guarantee under the more realistic conditions of "multiple targets, partial detectors, communication constraints, online adjustment, and cooperative interception."

[0008] In summary, existing technologies are insufficient to simultaneously address online target allocation, consistent target state acquisition, and pre-set time-based coordinated interception under multi-to-multi interception constraints with partial detectors. Therefore, there is an urgent need to develop an integrated allocation-observation-guidance control design method for intercepting maneuvering target groups using heterogeneous interceptor clusters with partial detectors, thereby improving the overall efficiency and robustness of multi-target coordinated interception under communication constraints and resource limitations. Summary of the Invention

[0009] In view of this, the present invention provides an integrated design and working method for collaborative allocation-observation-guidance control in many-to-many interception. This method designs hard constraints based on the collaborative observation and guidance control process, and performs initial target allocation and dynamic optimization of online target allocation based on the hard constraints. It can solve the problem of online target allocation under partial detector constraints in many-to-many interception, ensure the feasibility of the target allocation scheme in the collaborative observation and guidance control process, and avoid problems such as "allocation is feasible but guidance control is not feasible" or "guidance control is feasible but the overall interception efficiency of the cluster is poor".

[0010] To solve the above-mentioned technical problems, the present invention is implemented as follows.

[0011] A collaborative allocation-observation-guidance control integrated design and operating method for many-to-many interception, applicable to heterogeneous interceptor clusters with some detectors; the method includes:

[0012] Step 1: The target allocation module performs initial target allocation based on hard constraints. The hard constraints are determined based on the feasibility of the collaborative observation and guidance control process, including: Constraint 1, the leader of the target interception group assigned to the target must have a detector; Constraint 2, all members in the target interception group maintain connectivity; Constraint 3, the members in each interceptor group do not exceed the capacity limit and conform to the smallest group size. Step 2: The collaborative observation module performs joint collaborative observation based on the target allocation results, combining distributed collaborative observation with local observation by the lead aircraft, to obtain collaborative observation results; Step 3: The collaborative guidance and control module performs collaborative guidance and control based on the collaborative observation results, and implements interception; Step 4: The target allocation module performs dynamic target allocation during the interception process: continuously performs hard constraint detection; performs hard constraint repair for target interception groups that do not meet the hard constraints; performs utility degradation detection for target interception groups that meet the hard constraints, and for target interception groups that trigger utility degradation conditions, under the condition of meeting the hard constraints, selects interceptors with utility gain effects from other target interception groups to add to this target interception group.

[0013] Preferably, in step 1, the initial target allocation based on hard constraints is as follows: Step 11: Leader allocation: The interception capability is represented by the utility score U of the interceptor's interception of the target. The distributed maximum consensus algorithm is adopted to enable all interceptors to reach a consensus on the interception capability through communication. Each target is assigned a unique interceptor with a detector as the leader to satisfy constraint 1 in the hard constraints. Step 12: Slave allocation: The interceptors other than the already allocated leader are treated as slaves. Based on the utility score U, the target is allocated slaves using a price auction algorithm, and the allocation is adjusted using the hard constraints.

[0014] Preferably, the aircraft allocation in step 11 further includes aircraft allocation conflict resolution processing to ensure that one aircraft allocation corresponds to only one target: All targets are sorted in descending order of their utility score U relative to the leader. Leader conflict is determined for each target in sequence. If the leader initially assigned to the current target is not occupied by other targets, the current leader assignment is confirmed and marked as "assigned". If the leader initially assigned to the current target is occupied by other targets, the unassigned interceptor with the highest utility score U relative to the current target is selected from the remaining unassigned interceptors equipped with detectors and designated as the leader, and the new leader is marked as "assigned".

[0015] Preferably, the slave allocation in step 12 specifically includes: Step 121: Ideal Load Calculation: For each target Calculate the target from all slave devices. The utility scores U are averaged to obtain the average comprehensive utility avg_utility. j ); the average combined utility of all objectives avg_utility( j Further, the average value is calculated to obtain the total average utility (total_utility); the proportion of the average comprehensive utility in the total average utility is multiplied by the total number of slaves, and this is used as the target. ideal_load( j ), and limit the ideal load (ideal_load). j The total number of interceptors and the lead aircraft does not exceed the target interception group's capacity limit; Step 122: Price auction algorithm based on ideal load: ideal_load( j The ideal number of slave devices is given, and the ideal load is given by ideal_load( j Under the guidance of [unclear], a price auction method is used to allocate slave machines to the target; in each iteration of the price auction, the slave machines are used... and target Utility scores between and target The current auction price is π_j, and the target value is calculated. Assigned slave net utility util_all( j )= -π_j, and adjust based on the target status and the deviation from the ideal load number of the target interception group; when assigning targets to slaves, the target with the highest net utility is selected as the best choice; calculate the price change based on the difference between the number of slaves assigned to the target and the ideal load number, and update the price π_j accordingly; Step 123: Guaranteed Allocation: If there are still unallocated slaves after the price auction, a greedy strategy is used to force allocation, ensuring that all slaves are allocated; specifically: for each unallocated slave... Calculate all targets relative to slaves The overall utility score, i.e., score(j) = (net utility + target value × kill probability); the slave... Prioritize assigning to the target with the highest score(j) and remaining capacity; if all targets are full, select the target with the lowest current load; if all targets are dead, default to assigning to the designated target. Step 124: Perform constraint repair based on the hard constraints: (a) Local repair: Capacity restoration: If a target interception group exceeds its maximum capacity, remove redundant members, prioritizing the removal of slave devices without detectors; Minimum group size repair: If the number of members in a target interception group is insufficient, members are borrowed from other groups, with priority given to slaves without detectors, while ensuring that the borrowed target interception group still meets the hard constraints; Leader constraint repair: If a target interception group has no leader, a slave with a detector is borrowed from another group, and the borrowed target interception group still satisfies the hard constraints. (b) Connectivity restoration: for each target The Breadth-First Search (BFS) algorithm is used to check if the nodes within the target interception group form a connected subgraph. If they are connected, the original assignments remain unchanged. If they are not connected, BFS is used to find all connected components. A principal component is selected, prioritizing components containing the leader; if no leader is found, the largest component is selected. The principal component is then retained in the original target. Using BFS distance, starting from the root of each target, other isolated components are redistributed to the nearest target; (c) Constraint check: Verify whether all hard constraints are satisfied after the repair; Step 125: Generate initial allocation results.

[0016] Preferably, the price auction algorithm based on ideal load guidance described in step 122 specifically includes: Step 1221: Capacity Update: For each target Recalculate the remaining capacity cap for each target interception group based on the current allocation status. j ); Step 1222: Slave Selection Stage: Slave The competitive objectives are divided into two scenarios, A and B, and will be handled accordingly. Situation A: Slave If a target (current_assign_ii) has been assigned, determine whether to switch to a better target. First, check if the current assignment is still valid. If the assigned target is dead or overloaded, clear the invalid assignment, execute the reselection logic, and jump to case B. If the assigned target is alive and not overloaded, execute the following steps: Step (a1): Calculate the net utility of all objectives using util_all( j First, calculate the slave device. For each objective The basic net utility, i.e., util_all( j )= -π_j; then for util_all( j Adjustments are made based on the difference between the target current load and the ideal load, determining either an overload penalty or an underload reward to adjust the net utility (util_all). j ) value, to optimize the priority of assigning it to slaves; for dead targets, util_all( j Adjust to negative infinity; for non-allocated targets current_assign_ii, if the capacity is full, then in net utility util_all( j Imposing severe penalties on it and lowering its priority for assigning slave machines; Step (a2): Get the net utility of the current allocation util_all(current_assign_ii); if the capacity of the allocated target current_assign_ii is full, adjust the net utility of the current allocation util_all(current_assign_ii) to negative infinity; Step (a3): Switch judgment: Select net utility util_all( j The biggest goal is to serve as the slave machine. The best alternative target is best_j; if the capacity of the currently assigned target current_assign_ii is full, a forced switch is determined and step (a4) is executed; if the currently assigned target is not full, and the net utility of the best alternative target best_j is higher than the net utility of the currently assigned target current_assign_ii, and the difference exceeds the threshold, it is determined that a switch is worthwhile, and step (a4) is executed. Step (a4): Perform the switch: Replace the currently assigned target current_assign_ii with the best alternative target best_j and assign it to the slave. ; Situation B: Slave If the allocation is unallocated or has been cleared, proceed with the following steps: Step (b1): Calculate the net utility of all objectives using util_all( j ); Step (b2): Select the target with the highest net utility as the slave. The best target choice is best_j; Step (b3): ​​Submit request: Slave Submit an allocation request to the best target selection (best_j) and wait for the target approval phase to process it. Step 1223: Goal Approval Phase: For each goal If the number of allocation requests (demand) does not exceed the remaining capacity (remaining_cap), all requests are approved; if the number of allocation requests exceeds the remaining capacity, requesters are sorted in descending order of net utility, the requests at the top are approved, the remaining requests are rejected, and the rejected slaves are reselected in the next round. Step 1224: Price Update Phase: Update the target price based on supply and demand to guide slaves to shift towards low-load targets. Specifically, this includes: First, determining the approved actual load (current_load_after); determining the excess demand (excess_demand) based on the remaining capacity (remaining_cap) and the number of allocation requests (demand); if the actual load (current_load_after) does not exceed the ideal load, then the load is insufficient. Calculate the price reduction factor based on the insufficient load ratio, and then calculate the price change (price_change) based on the price reduction factor (price_reduction_factor) and the excess demand (excess_demand); update the auction price (π_j) using the price change (price_change); if the load is sufficient or overloaded, update the auction price (π_j) using the excess demand (excess_demand). Step 1225: Convergence Judgment: Determine whether the price auction has converged and decide whether to exit the loop.

[0017] Preferably, the utility score U of the interceptor is calculated as follows: Targeting the interceptor The design utility function is :

[0018] in, These are non-negative weighting coefficients, used to measure the importance of target value, threat level, engagement geometry, time advantage, and role preference, respectively. and These are the target's value and its threat level, respectively. , , , These represent kill probability, engagement geometry score, time advantage bonus, and character preference value, respectively. Probability of lethality Based on combat geometric quality score Detector dependence coefficient and motor vehicle penalty items Sure:

[0019] Among them, the detector dependence coefficient For interceptors equipped with detectors, Take the detector coefficient of the detector interceptor For interceptors that do not have detectors, Take the detector coefficient of the detectorless interceptor ; Motor penalty items Designed to punish situations where the target's maneuverability exceeds the interceptor's maneuverability, as follows:

[0020] in, It is the design coefficient. The upper bound of the estimate represents the target normal acceleration. It is the maximum permissible normal acceleration of the interceptor.

[0021] Preferably, step 4 includes four stages: Phase 1: Distributed Hard Constraint Detection Each interceptor The system detects the constraint violations of its own target group, uses the maximum consensus algorithm to propagate the constraint violation information in the communication network, and after convergence, all interceptor nodes reach a consensus on which assigned target interception groups have violated the hard constraints; if the hard constraints are satisfied, it proceeds to stage 3; if the hard constraints are not satisfied, it proceeds to stage 2. Phase 2: Hard Constraint Repair: For target interception groups that violate hard constraints, based on the utility score U gain and the feasibility of leaving the original group as defined by the hard constraints, each interceptor independently calculates whether it can become a target. The candidate is then determined using the maximum consensus algorithm, and the target allocation result is updated. Then, leader conflict resolution is performed to ensure that one leader corresponds to only one target. If hard constraint repair fails, find the leader root for each target and use the BFS algorithm to reassign all interceptor nodes to the nearest root target to ensure that all hard constraints are satisfied. Phase 3: Utility Decline Detection: Utility decline detection is performed only on target interception groups that meet hard constraints. Specifically, the utility scores U of each interceptor and target in the current target interception group are calculated and summed to obtain the current group utility current_util. The current group utility current_util is compared with the group utility calculated based on the initial allocation to obtain the utility decline ratio. When the utility decline ratio is greater than a set value, the maximum consensus algorithm is triggered to confirm the triggering of utility optimization within the group. If a majority of nodes in the group confirm the triggering of utility optimization, a reassignment is triggered. Phase 4: Utility Optimization: For each objective that triggers utility optimization... Candidate interceptors are discovered from the communicable neighbors of other groups. The candidate condition is that the candidate interceptor joins the target. Target interception group against target The utility gain is greater than zero; at the same time, for a candidate interceptor to leave the original group, it must be a leaf node in the original group, and the original group still satisfies the hard constraints after leaving the original group. When multiple candidate interceptors are available, priority is given to candidates with high utility gain and small original group size. If multiple targets compete for the same candidate interceptor, the target with the highest priority is selected as the candidate interceptor, and the maximum consensus algorithm is used to reach a consensus in the communication network.

[0022] Preferably, the joint collaborative observation adopted by the collaborative observation module is as follows: All interceptors in the target interception group use a distributed observer to estimate the target motion state; the consistency estimation error in the distributed observer includes the estimation difference error term between itself and its neighbors and the local target estimation error term that is only valid for the leader aircraft; The leader directly measures and estimates the local state of the target through its own detector, and uses the local state of the target to construct the local target estimation error term.

[0023] Preferably, all interceptors use a preset time-distributed observer; the leader uses a preset time-extended state observer (PTESO). The preset time-distributed observer is:

[0024]

[0025]

[0026] in, , and Interceptors The target estimated by the upper observer Estimates of the line-of-sight distance vector, velocity vector, and acceleration vector; For interceptors The velocity vector; It is the gain coefficient of the distributed observer; For the sign function; Consistency estimation error The function of consensus estimation error Consistency estimation errors including line-of-sight distance vector, velocity vector, and acceleration vector , , ; Defined as:

[0027] in, , It is an adjustable parameter, and the values ​​are taken separately in different distributed observers. , This is the preset time for the distributed observer; The consistency estimation error of the above line-of-sight distance vector, velocity vector, and acceleration vector is defined as follows:

[0028]

[0029]

[0030] in, Indicate target The number of interceptors in the target interception group; For the goal The elements in the communication topology adjacency matrix corresponding to the target interception group. Representative interceptor With interceptors They can directly exchange information based on the communication network. Representative interceptor With interceptors They cannot communicate directly; Representative interceptor With interceptors The relative line-of-sight distance vector between them; coefficient Representative interceptor With or without detectors; if detectors are equipped, then... ,otherwise ; , , These are the local estimates of the target line-of-sight distance vector, velocity vector, and acceleration vector obtained by the interceptors equipped with detectors in the target interception group based on the detector measurement information and using a preset time-dilation state observer (PTESO). The preset time-dilation state observer PTESO is designed as follows:

[0031]

[0032]

[0033] in, , The target line-of-sight distance vector obtained by the detector; It is the gain coefficient of the preset time-dilation state observer; Defined as:

[0034] in, , It is an adjustable parameter, and can be selected in different preset time-dilation state observers. , It is the preset time of the preset time expansion state observer; .

[0035] Beneficial effects: 1. An integrated design of allocation, observation, and guidance control addresses the feasibility and effectiveness issues caused by decoupled designs. Existing technologies decouple allocation from guidance control, failing to fully consider the impact of allocation results on subsequent observation conditions, communication topology connectivity, and the feasibility of coordinated guidance control. This application adopts an integrated design: In the initial allocation phase, connectivity repair ensures communication connectivity among groups; in further optimization, constraint repair ensures that each group meets the minimum group size and leader constraints, providing a feasible basis for subsequent observation and guidance control; in the observation phase, a dual-layer observer design fully utilizes the leader's detector measurements and the slave's communication capabilities, achieving consistent target state estimation through distributed observers; in the guidance control phase, coordinated guidance control commands are calculated based on the target state estimated by the observers, and time consistency is achieved within the same target group through preset time synchronization. This integrated design ensures that the allocation results are implementable in the observation and guidance control phases, avoiding the problem of "allocation being feasible but guidance control not being implementable," thus improving overall interception effectiveness.

[0036] 2. A distributed online allocation and dynamic reallocation mechanism addresses the limitations of centralized computing and offline planning. Existing technologies often employ centralized computing or offline planning, relying on global situational awareness and making rapid online adjustments difficult. This invention uses a distributed maximum consensus selection of the leader, where leaders achieve global optimal consensus through a communication network, eliminating the need for central coordination. It employs a switching incremental price auction for slave allocation targets, with prices dynamically adjusted based on supply and demand to achieve load balancing. During task execution, distributed hard constraint detection and utility degradation detection trigger dynamic reallocation, locally correcting constraint violations or optimizing performance to avoid global relocation. This mechanism enables the system to adjust rapidly online even when target maneuverability increases or interceptor states change quickly, improving the real-time performance and adaptability of allocation results.

[0037] 3. A dual-layer observer and preset-time collaborative guidance control under partial detector constraints addresses the robustness and fast convergence issues of heterogeneous platforms. Existing technologies implicitly assume "all personnel have detectors," resulting in insufficient robustness under partial detector constraints. Collaborative guidance control methods overly rely on initial conditions and linear approximations, leading to instability when facing high-speed, highly maneuvering targets. This application addresses the heterogeneous structure with partial detectors: the first-layer preset-time extended state observer (PTESO) is executed only by the lead aircraft with detectors, using detector measurements to quickly estimate the target state; the second-layer distributed observers share and reach a consensus on the target state estimate through a communication network, enabling slave aircraft without detectors to also obtain the target state; preset-time control ensures the observers converge within a specified time, independent of initial conditions; the collaborative control law, based on the target state estimated by the observers, uses a preset-time synchronization mechanism to ensure that the remaining flight time of each interceptor in the group reaches a consensus within a specified time, achieving fast convergence. This design can still guarantee the accuracy of target state estimation and the fast convergence of collaborative guidance control under partial detector constraints, improving adaptability to target maneuvers and dynamic environments, and enhancing system robustness.

[0038] 4. Load-balanced price auctions and minimal-modification connectivity repair enhance allocation quality and system stability. Existing distributed allocation mechanisms do not adequately consider load balancing, leading to uneven allocation. This application introduces ideal load calculation based on comprehensive utility into the price auction, more aggressively lowering prices for low-load targets during price updates to accelerate the transfer of slaves to low-load targets; a load balancing term is added to the net utility calculation to achieve a balance between load balancing and utility maximization; connectivity repair employs a minimal-modification strategy, repairing only disconnected groups while maintaining the allocation of connected groups, thus preserving the optimization results of the price auction to the greatest extent. This mechanism improves system stability while ensuring allocation quality and avoids frequent large-scale adjustments.

[0039] 5. Join / Leave rules and a hard constraint-priority redistribution mechanism ensure constraint satisfaction during dynamic adjustment. Existing technologies lack effective dynamic redistribution mechanisms, making it difficult to simultaneously guarantee multiple constraints such as communication connectivity, leader constraints, and capacity limitations in tasks involving cluster interception of mobile target groups. This application proposes Join / Leave rules: The Join rule requires that when an interceptor joins a target group, at least one neighbor must already be in the target group or the target group must be empty, ensuring communication connectivity within the group after joining; the Leave rule requires that when an interceptor leaves its original group, it must be a leaf node within the group, and the original group must still satisfy hard constraints after leaving, ensuring continued connectivity within the original group; a four-stage redistribution process prioritizing hard constraints is adopted to ensure the satisfaction of hard constraints and reduce the impact of dynamic redistribution on overall stability. This mechanism ensures that each group always satisfies hard constraints such as communication connectivity, minimum group size, and leader constraints during dynamic redistribution, improving system reliability and robustness. Attached Figure Description

[0040] Figure 1 This is a schematic diagram illustrating the relative motion between the interceptor and the target. Figure 2 A schematic diagram of the algorithm framework for target assignment; Figure 3 This is a schematic diagram of a closed-loop architecture for many-to-many interception: initial allocation, observation sharing, collaborative guidance and control, and dynamic reallocation. Detailed Implementation

[0041] This invention provides an integrated design and operating method for collaborative allocation, observation, guidance, and control of many-to-many interception, applicable to heterogeneous interceptor clusters with some detectors. The method includes: Step 1: The target allocation module performs initial target allocation based on hard constraints. The hard constraints are determined based on the feasibility of the collaborative observation and guidance control process, including: Constraint 1, the leader of the target interception group assigned to the target must have a detector; Constraint 2, all members in the target interception group maintain connectivity; Constraint 3, the members in each interceptor group do not exceed the capacity limit and conform to the smallest group size.

[0042] Step 2: The collaborative observation module performs joint collaborative observation based on the target allocation results, combining distributed collaborative observation with local observation by the lead aircraft, to obtain collaborative observation results.

[0043] Step 3: The collaborative guidance and control module performs collaborative guidance and control based on the collaborative observation results, and implements interception.

[0044] Step 4: The target allocation module performs dynamic target allocation during the interception process: continuously performs hard constraint detection; performs hard constraint repair for target interception groups that do not meet the hard constraints; performs utility degradation detection for target interception groups that meet the hard constraints, and for target interception groups that trigger utility degradation conditions, under the condition of meeting the hard constraints, selects interceptors with utility gain effects from other target interception groups to add to this target interception group.

[0045] As can be seen, this invention designs hard constraints based on the collaborative observation and guidance control process, and performs initial target allocation and online target allocation dynamic optimization based on the hard constraints. This can solve the problem of online target allocation under the constraints of some detectors in many-to-many interception, ensure the feasibility of the target allocation scheme in the collaborative observation and guidance control process, and avoid problems such as "allocation is feasible but guidance control is not feasible" or "guidance control is feasible but the overall interception efficiency of the cluster is poor".

[0046] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0047] Assume a scenario where m interceptors attack n moving targets, where the number of detector interceptors in the interceptor cluster is greater than or equal to the number of targets. Let the i-th interceptor and the j-th target be represented as follows: and The initial communication topology between interceptors is defined as follows: , is an undirected connected graph, and the corresponding adjacency matrix is ​​. , If the elements in the adjacency matrix are... Able to Communication, ,otherwise ,in addition, The velocity vectors of the interceptor and the target are respectively expressed as... and The inertial coordinate system and line-of-sight coordinate system in which the two intercepts occur are respectively determined by... and Indicates. Interceptor The acceleration vector is ,in, , , They are exist Components along the three axes in the coordinate system. Target. The acceleration vector is ,in , , They are exist Components on the next three axes.

[0048] Interceptor With the goal The line-of-sight distance vector between them is The elevation angle and azimuth angle of the line of sight are respectively and , Interceptor With the goal Line-of-sight distance between them. A diagram illustrating the relative motion between the interceptor and the target is attached. Figure 1 As shown. For a given interceptor target pair The equations of relative motion in the line-of-sight coordinate system can be described in the following form: (1) in, (2) This technical solution mainly comprises three modules: a target allocation module, a collaborative observation module, and a collaborative guidance and control module. These three modules are designed and operate collaboratively. Specifically, the steps include: Step 1: The target allocation module performs initial target allocation based on hard constraints.

[0049] This step is performed by the target allocation module. Target allocation consists of two phases: "initial allocation" and "dynamic reallocation." This step performs the initial allocation, then the collaborative observation module performs collaborative observation according to the initial allocation scheme, and the collaborative guidance and control module performs collaborative guidance and control based on the collaborative observation results. During the collaborative observation and collaborative guidance and control processes, the target allocation module performs real-time monitoring and performs "dynamic reallocation" operations on the allocation scheme to achieve better observation and tracking.

[0050] First, regarding the interceptor target pair The design utility function is : (3) in, These are non-negative weighting coefficients used to measure the importance of factors such as target value, threat level, interception geometry, time advantage, and role preference. and These are the target's value and threat level, respectively, and are generally given constant values ​​based on experience. , , , These represent kill probability, interception geometric quality score, time advantage gain, and role preference value, respectively.

[0051] 1) Interception geometric quality score The design is shown below: (4) in, It belongs to the weighting coefficient and satisfies The appropriate value can be selected based on the actual situation. , , These represent the interception distance score, heading match score, and approach speed score, respectively, and can be calculated using the following formulas: (5) in, Indicates the distance decay threshold. and These represent the unit velocity vector of the interceptor and the unit vector of the line-of-sight direction, respectively. This represents the coefficient of the rate of change of distance. This indicates the rate of change in line-of-sight distance.

[0052] 2) Interceptor kill probability According to the present invention, a special design is made for a heterogeneous interceptor cluster of some detectors, based on the engagement geometry quality score. Detector dependence coefficient and motor vehicle penalty items It is confirmed that, because the difference between the two levels of the three data is relatively large, it can be designed as the product of the three, as shown below: (6) Among them, the detector dependence coefficient Calculated by the following formula: (7) Among them, coefficient This indicates whether the interceptor has a detector. If it is equipped with a detector, then... ,otherwise . and These represent the detector coefficients for interceptors without detectors and those with detectors, respectively.

[0053] Motor penalty items Designed to punish situations where the target's maneuverability exceeds the interceptor's maneuverability, as follows: (8) in, It is the design coefficient. The upper bound of the estimate represents the target normal acceleration. It is the maximum permissible normal acceleration of the interceptor.

[0054] 3) Time advantage gain Designed as follows: (9) in, Represents the interception time threshold. Represents the remaining flight time, which can generally be determined by... Calculated.

[0055] 4) Role Preference Value Designed as follows: (10) in, This represents the base score of the interceptor itself. This represents the degree to which the interceptor matches this type of target; both are generally given constant values ​​based on experience.

[0056] Based on the aforementioned utility function U design, initial allocation is performed. Considering constraints on some detectors, the initial allocation first determines the allocation result for the leader, and then determines the allocation result for the slaves. Specifically, each target is first assigned a unique leader (using a distributed maximum consensus algorithm to allow all interceptors to reach a consensus on the "optimal leader" through communication), and then targets are assigned to the slaves (using an improved price auction algorithm to introduce load balancing, allowing slaves to move to targets with "high utility and low load," and also allowing slaves to switch targets when the price changes significantly). After allocation, constraint repair (capacity, smallest group, communication connectivity) is performed to ensure that the allocation result can support subsequent observation and guidance control.

[0057] The initial allocation consists of two main steps: Step 11: Leader Assignment: The interception capability is represented by the utility score U of the interceptor's intercepted targets. A distributed maximum consensus algorithm is used to ensure that all interceptors reach a consensus on their interception capabilities through communication. Each target is then assigned a unique interceptor with a detector as the leader. This step satisfies constraint 1 in the hard constraints.

[0058] Step 12: Slave Assignment: Interceptors other than the already assigned leader are designated as slaves. Based on the utility score U, a price auction algorithm is used to assign slaves to the target, and the hard constraints are applied for allocation adjustment. Constraint repair is also included after slave assignment.

[0059] The following details the specific processes of assigning master and slave machines.

[0060] For aircraft registration and allocation, a maximum consistency allocation method is designed. Therefore, step 11, aircraft registration and allocation, specifically includes the following steps.

[0061] Step 111: Initialization.

[0062] Suppose there are L interceptors equipped with detectors. For these interceptors... ( i =1,2,...,m L For each target Maintain a pair of values ​​(score( i,j ), owner( i,j Initialized to score( i,j ) = owner( i,j ) = i score( i,j It is updated in the loop so that it records for each target. Maximum score, owner ( i,j Record which interceptor the highest score came from.

[0063] Step 112: Maximum Consistency Iteration. The maximum number of iteration rounds is set to the diameter of the interceptor cluster communication topology graph. In each iteration round, for each interceptor... (Including all interceptors, not limited to those with detectors), find and The set of neighboring interceptors that can communicate is nei = { k | >0};For each target Find the maximum score value in the neighbor set nei, i.e., best_s = max{score( k,j ) | k ∈ { i} ∪ nei}, then update: score( i,j ) = best_s, owner( i,j =The corresponding owner value. Maximum consistency iteration ensures that each interceptor reaches a consensus on its interception capability against the target.

[0064] Step 113: Extract results after convergence. Using the state of the first node as the benchmark (due to maximum consistency, all node states are the same), extract results for each target... Calculate: score_of( j ) = score(1, j ), leader_of( j = owner(1, j Among them, leader_of( j Record to the target Initially assigned interceptors, score_of( j (Target after initial allocation) The utility score between the assigned interceptor and the target interceptor.

[0065] The interceptor equipped with a detector assigned to a target is called the lead interceptor.

[0066] Step 114: Resolve aircraft lead conflicts to ensure that one lead aircraft corresponds to only one target.

[0067] In this step, all targets are sorted in descending order of their utility score U relative to the leader, and a leader conflict judgment is performed on each target in sequence. If the leader initially assigned to the current target is not occupied by other targets, the current leader assignment is confirmed and the leader is marked as "assigned". If the leader initially assigned to the current target is occupied by other targets, the unassigned interceptor with the highest utility score U relative to the current target is selected from the remaining unassigned interceptors equipped with detectors and designated as the leader, and the new leader is marked as "assigned".

[0068] Specifically, it includes: Step 1141: Check if multiple objectives have the same leader_of value, meaning multiple objectives have been assigned the same leader. If multiple objectives have selected the same leader, process them in descending order of objective utility score.

[0069] Step 1142: For each target (In descending order of utility score), if the initially selected leader_of( j If the leader is not occupied by other targets, then the confirmation operation is executed directly: leader_of_final( j ) = leader_of( j The leader has been assigned: leader_assigned(leader_of( j )) = j If the initially selected leader is already occupied by another target, then select a target from the remaining unassigned leader. The leader with the highest utility is assigned a new leader, and `leader_of_final(j)` is set to `new_selected_leader`. This marks the leader as already assigned, i.e., `leader_assigned(new_leader)` = `new_leader`. j。 Then, `leader_of = leader_of_final` is updated to ensure that each target has a unique leader, and a leader can only be assigned to one target. This step only logically determines the leader and has not yet been applied to the assignment vector `assign`. The target assignment module and the collaborative observation module use the assignment vector `assign` for information transmission.

[0070] Step 115: Confirm aircraft allocation.

[0071] The above is the logical allocation result. In actual engineering, it is not used directly. The preferred solution is to set up a separate actual allocation step for the leader machine. The core reason is that the result at the logical level only completes the "correspondence confirmation" between the leader machine and the target, but does not fall into the allocation execution system of the entire interceptor cluster, nor does it build an operable parameter foundation for the subsequent allocation of slave machines.

[0072] Specifically, this step first initializes the allocation vector assign = zeros(n,1) and initializes the capacity cap = the maximum number of interceptors allowed to be allocated to each target × ones(m,1). Here, zeros(n,1) represents an n×1 zero matrix, and ones(m,1) represents an m×1 1 matrix.

[0073] For each target If leader_of( jIf )>0, then the actual allocation operation is assign(leader_of( j )) = j， Update the target at the same time. The remaining capacity is cap( j ) = cap( j -1. At this point, the leader has been actually assigned to the corresponding target, and the assignment vector assign and capacity cap have been updated, in preparation for the subsequent slave price auction.

[0074] The following is a slave-based improved price auction algorithm. Therefore, step 12 mainly includes the following key steps.

[0075] Step 121: Ideal Load Calculation: For each target Calculate the target from all slave devices. The utility scores U are averaged to obtain the average comprehensive utility avg_utility. j ); the average combined utility of all objectives avg_utility( j Further, the average value is calculated to obtain the total average utility (total_utility); the proportion of the average comprehensive utility in the total average utility is multiplied by the total number of slaves, and this is used as the target. ideal_load( j ), and limit the ideal load (ideal_load). j The total number of interceptors and the lead aircraft does not exceed the upper limit of the target interception group's capacity.

[0076] Step 122: Price auction algorithm based on ideal load: ideal_load( j The ideal number of slave devices is given, and the ideal load is given by ideal_load( j Under the guidance of [unclear], a price auction method is used to allocate slave machines to the target; in each iteration of the price auction, the slave machines are used... and target Utility scores between and target The current auction price is π_j, and the target value is calculated. Assigned slave net utility util_all( j )= -π_j, and adjust based on the target state and the deviation from the ideal load number of the target interception group; when assigning targets to slaves, the target with the highest net utility is selected as the best choice; calculate the price change based on the difference between the number of slaves assigned to the target and the ideal load number, and update the price π_j.

[0077] Step 123: Guaranteed Allocation: If there are still unallocated slaves after the price auction, a greedy strategy is used to force allocation, ensuring that all slaves are allocated; specifically: for each unallocated slave... Calculate all targets relative to slaves The overall utility score, i.e., score(j) = (net utility + target value × kill probability); the slave... Prioritize assigning to the target with the highest score(j) and remaining capacity; if all targets are full, select the target with the lowest current load; if all targets are dead, assign to the designated target by default.

[0078] Step 124: Perform constraint repair according to the hard constraints to ensure compliance with the hard constraints.

[0079] Step 125: Generate initial allocation results.

[0080] The following is a more detailed description of the implementation process for steps 121-125 above. It includes the following steps: Step 120: Initialization.

[0081] Initialize the price, i.e., π_j = 0 ( j =1,2,...,n), all targets have an initial price of zero. Determine the slave set, i.e., F = {i|i {leader_of( j ) | j =1,2,...,n}}, excluding all interceptors after the assigned server has been assigned. Note that the load balancing weight is represented by balance_weight, the switching threshold by switching_threshold, and the price update step size by alpha0.

[0082] Step 121: Ideal Load Calculation. Calculate the average overall utility for each objective (considering only slaves), i.e., the utility for each objective. Collect all slave pairs Utility value: follower_utils = { | i ∈ F, If follower_utils is not empty, then execute avg_utility( j = mean(follower_utils), otherwise execute avg_utility( j )= , for The target value. Then, allocate slaves according to the overall utility ratio, that is, calculate the total average utility of all surviving targets: total_utility = Σ avg_utility( j ), calculate the ideal load (ideal_load) for each target. j =(mn) × avg_utility( j The ideal load needs to take into account capacity limitations, subtracting the allocated host machines, i.e., ideal_load( j ) = min(ideal_load( j ), maxGroupSizePerTarget-1), while ensuring that the ideal load is non-negative, i.e., ideal_load( j = max(ideal_load( j ), 0). maxGroupSizePerTarget is the maximum number of members in the target intercept group.

[0083] Step 122: Improve the price auction algorithm. This auction algorithm consists of the following five sub-steps.

[0084] Sub-step (1): Capacity update. For each target Recalculate the remaining capacity cap for each target interception group based on the current allocation status. j ).

[0085] Specifically, for each target The remaining capacity is recalculated based on the current allocation status, i.e., cap( j ) =maxGroupSizePerTarget -sum(assign == j When a slave device switches targets, the capacities of the old and new targets need to be updated, so this must be recalculated at the start of each iteration.

[0086] Sub-step (2): Slave selection stage.

[0087] For each slave machine , i ∈ F, consider cases A and B.

[0088] First, there's scenario A, which involves the slave machine. If an application has been assigned to a target (current_assign_ii), determine whether to switch to a better target.

[0089] First, check if the current allocation is still valid. If the allocated current_assign_ii target is dead or overloaded, clear the invalid allocation, assign(i) = 0, and continue with the reselection logic (jump to case B). If the allocated current_assign_ii target is alive and not overloaded, perform the following steps: (a1) Calculate the net utility of all objectives using util_all( j First, calculate the slave device. For each objective The basic net utility, i.e., util_all( j ) = - π_j; then for util_all( j Adjustments will be made: • Based on the gap between the target current load and the ideal load, determine the overload penalty or underload reward to adjust the net utility util_all( j This value is used to optimize the priority of assigning slave devices. Specifically: That is, for each target Calculate the current load: current_load_j = sum(assign == j (Currently assigned to target) The number of interceptors), if overloaded, i.e., current_load_j > ideal_load( j If the load is too high, then the overload penalty is calculated as: overload_penalty = balance_weight × (current_load_j - ideal_load( j )) / max(ideal_load( j ), 1), and update net utility util_all( j = util_all( j - overload_penalty; if the load is insufficient, i.e., current_load_j <ideal_load( j If the load is insufficient, the bonus is calculated as: underload_bonus = balance_weight × (ideal_load( j )-current_load_j) / max(ideal_load( j ), 1), and update net utility util_all( j = util_all( j) + underload_bonus.

[0090] • Adjust util_all(j) to negative infinity for dead targets: execute util_all(j) = -inf.

[0091] • For non-assigned targets current_assign_ii, if the capacity is full, then in net utility util_all( j Apply a heavy penalty to each target (excluding the currently assigned target) and reduce its priority for assigning slave machines: that is, for each target If cap( j )≤ 0 and j If it is not the current_assign_ii, then util_all( j = util_all( j ) – 1000.

[0092] • Handling the special case where the current allocation target is full: Get the net utility of the current allocation util_all(current_assign_ii); if the capacity of the allocated target current_assign_ii is full, i.e., cap(current_assign_ii) ≤ 0, then adjust the net utility of the current allocation util_all(current_assign_ii) to negative infinity, i.e., execute util_all(current_assign_ii) = -inf, which means that when the current allocation target is full, its net utility is set to negative infinity, forcibly allowing switching.

[0093] (a2) Get the net utility of the current allocation util_all(current_assign_ii); set current_net_util = util_all(current_assign_ii); if the capacity of the allocated target current_assign_ii is full, adjust the net utility of the current allocation util_all(current_assign_ii) to negative infinity, that is, execute current_net_util = -inf.

[0094] (a3) Switching judgment: Determine the slave device Should we switch targets, or keep the current assignment?

[0095] Select net utility using util_all( j The biggest goal is to serve as the slave machine. The best alternative target is best_j; if the capacity of the currently assigned target current_assign_ii is full, a forced switch is determined and step (a4) is executed; if the currently assigned target is not full, and the net utility of the best alternative target best_j is higher than the net utility of the currently assigned target current_assign_ii, and the difference exceeds the threshold, it is determined that a switch is worthwhile, and step (a4) is executed.

[0096] Specifically, it checks if the current allocation target is full. If the current allocation target is full, i.e., cap(current_assign_ii) ≤ 0, then a forced switch is performed, and the assign( i If `util_all_for_max = util_all`, clear the current assignment. If the current assignment target is not full, find the best alternative target, create a candidate list `util_all_for_max = util_all`, and find the target with the highest net utility among all the optional targets (excluding the current assignment target if it is full). Check if there are any optional targets. If there are no optional targets, that is, all targets have `util_all_for_max` of -inf, then the current assignment must be maintained. A switching condition is determined: if the net utility of the best alternative target is significantly higher than the current assignment (the difference exceeds the switching threshold), it is worth switching, i.e., `best_j ≠ current_assign_ii` and `max_net_util > current_net_util + switching_threshold`, then the switch is executed, `assign(i) = 0`, the current assignment is cleared, and the process jumps to case B to continue the reselection logic. If the net utility of the best alternative target is not high enough (the difference does not exceed the switching threshold), it is not worth switching, then the current assignment is maintained, i.e., `req{current_assign_ii}(end+1) = 0`. i Continue to the next slave machine.

[0097] (a4) Perform a switchover: Replace the currently assigned target current_assign_ii with the best alternative target best_j and assign it to the slave. .

[0098] Then comes scenario B: slave machine If the allocation is unallocated or has been cleared, perform the following steps: (b1) Calculate the net utility of all objectives using util_all( j This is exactly the same as step (a1) in case A, so I will not repeat it here.

[0099] (b2) Select the objective with the highest net utility as the slave. The best target is chosen as best_j; calculate best_j = argmax{util_all( j ) | j =1,2,...,n}.

[0100] (b3) Submit request: slave machine Submit an allocation request to the best target (best_j) and wait for the target approval phase, i.e., execute the operation req{best_j}(end+1) = i .

[0101] Step 123: Target Approval Phase.

[0102] For each objective If the number of allocation requests (demand) does not exceed the remaining capacity (remaining_cap), all requests are approved; if the number of allocation requests exceeds the remaining capacity, requesters are sorted in descending order of net utility, the requests at the top are approved, the remaining requests are rejected, and the rejected slaves are reselected in the next round.

[0103] Specifically: for each target The number of requests is demand=length(req{ j}), the remaining capacity is remaining_cap=cap( j If the number of requests does not exceed the capacity (demand ≤ remaining_cap), then all requests are approved, meaning for each... i ∈req{ j} Execute assign( i )= j If the number of requests exceeds the capacity, requesters are sorted in descending order of net utility, and the first `remaining_cap` requests are approved. i ,assign( i )= j If the request is rejected, the slave device can reselect in the next round.

[0104] Step 124: Price Update Phase. Update the target price based on supply and demand to guide slave machines to shift towards low-load targets.

[0105] First, determine the approved actual load (current_load_after). Based on the remaining capacity (remaining_cap) and the number of allocation requests (demand), determine the demand exceeding the capacity, denoted as excess demand (excess_demand). If the actual load (current_load_after) does not exceed the ideal load, the load is insufficient. Calculate the price reduction factor based on the insufficient load ratio, and then calculate the price change (price_change) based on the price reduction factor (price_reduction_factor) and excess demand (excess_demand). Update the auction price (π_j) using the price change (price_change). If the load is sufficient or overloaded, update the auction price (π_j) using excess demand (excess_demand).

[0106] Specifically, this step applies to each target. The current load, i.e., the actual load after approval, is current_load_after = sum(assign == j The requirement is demand=length(req{ j The current round of requests is represented by `excess_demand = demand - max(remaining_cap, 0)`, where `excess_demand` represents the demand exceeding capacity. If the load is insufficient, the current load after `excess_demand` represents the excess demand. <ideal_load( j Then calculate the underload ratio, i.e., underload_ratio = (ideal_load( j )-current_load_after) / max(ideal_load( j ), 1), and then calculate the price reduction factor as price_reduction_factor = 1.0 + 1.0 × underload_ratio. Next, calculate the price change, i.e., price_change = alpha0 × excess_demand × (1 + price_reduction_factor × underload_ratio). Finally, update the price, i.e., π_j = max(0, π_j + price_change). If the load is sufficient or overloaded, update the price π_j = max(0, π_j + alpha0 × excess_demand).

[0107] Step 125: Convergence Judgment: Determine whether the price auction has converged and decide whether to exit the loop.

[0108] Specifically, this step calculates the number of unassigned slaves, i.e., unassigned_count = sum(assign(F) == 0). If all slaves have been assigned, i.e., unassigned_count == 0, then the loop exits; if the number of unassigned slaves does not decrease for three consecutive rounds, then the loop also exits. Otherwise, prev_unassigned_count = unassigned_count is updated, and the loop continues.

[0109] Step 126: Guarantee allocation.

[0110] If there are still unassigned slaves after the price auction, a greedy strategy is used to force allocation, ensuring that all slaves are allocated. That is, after the main loop of the price auction ends, if unassigned_count > 0, then for each unassigned slave... Calculate the overall score for all objectives, i.e., score(j) = ( - π_j)+ × (Net utility + target value × kill probability); then prioritize the target with remaining capacity. That is, if there is a target with remaining capacity, select the target with the highest score. If all targets are full, select the target with the lowest current load (temporarily relax capacity limit). If all targets are dead, then the target is assigned to target 1 by default.

[0111] Step 127: Constraint Repair. Repair any constraint violations that may arise from price auctions and guaranteed allocations, ensuring that the allocation results satisfy all hard constraints. This step mainly includes three aspects: (a) Local Repair. First, capacity repair is performed: if a target group exceeds the maximum capacity (maxGroupSizePerTarget), excess members are removed, prioritizing slave machines. Then, minimum group size repair is performed: if a target group has fewer members than (minGroupSizePerTarget), members are borrowed from other groups, prioritizing slave machines. Finally, if a target group has no leader machine, a leader machine is borrowed from other groups, while ensuring the original group still meets the constraints.

[0112] (b) Connectivity Repair. Repair violations of communication connectivity constraints while preserving the optimization results of the price auction to the greatest extent possible. For each objective... The Breadth-First Search (BFS) algorithm is used to check if the nodes within a group form a connected subgraph. If they are connected, the original assignments are maintained. If they are not connected, BFS is used to find all connected components. A principal component is selected, prioritizing components containing the leader node; if none of these have a leader node, the largest component is selected. The principal component is then retained in the original target. Using BFS distance, starting from the leader root of each target, other isolated components are redistributed to the nearest target.

[0113] (c) Constraint Check. Verify that the allocation results after constraint repair meet all hard constraints, namely, ensure that each group has at least one leader, ensure that each group meets the minimum group size minGroupSizePerTarget, and ensure that each group is in communication mode.

[0114] Step 128: After the initial allocation is completed, the results of the initial allocation can be generated.

[0115] Then, during task execution, a dynamic reallocation algorithm is triggered when a hard constraint violation or utility decline is detected. The dynamic reallocation algorithm adopts a four-stage process: "hard constraint detection, hard constraint repair, utility decline detection, and utility optimization." This process will be described in step 4.

[0116] Step 2: The collaborative observation module performs joint collaborative observation based on the target allocation results, combining distributed collaborative observation with local observation by the lead aircraft, to obtain collaborative observation results.

[0117] The joint collaborative observation module adopts the following method: all interceptors in the target interception group use distributed observers to estimate the target motion state; the consistency estimation difference in the distributed observers includes the estimation difference error term between itself and its neighbors and the local target estimation error term that is only valid for the leader aircraft; the leader aircraft directly measures and estimates the local state of the target through its own detectors, and uses the local state of the target to construct the local target estimation error term.

[0118] To ensure simultaneous convergence, all interceptors use a preset time-distributed observer, while the leader uses a preset time-extended state observer (PTESO).

[0119] Consider a pre-assigned target group, named The target group contains targets and several interceptors , The role of the collaborative observation module is to design distributed observers for each interceptor within the target group, and collaboratively estimate the target's... line-of-sight distance vector Velocity vector With acceleration vector Their estimates were respectively used , and express.

[0120] To enable each interceptor to quickly estimate the target's motion state, the preset time-distributed observer is:

[0121]

[0122]

[0123] in, , and Interceptors The target estimated by the upper observer Estimates of the line-of-sight distance vector, velocity vector, and acceleration vector; For interceptors The velocity vector; It is the gain coefficient of the distributed observer; For the sign function; Consistency estimation error The function of consensus estimation error Consistency estimation errors including line-of-sight distance vector, velocity vector, and acceleration vector , , .

[0124] Defined as:

[0125] in, , It is an adjustable parameter, and the values ​​are taken separately in different distributed observers. , This is the preset time for the distributed observer.

[0126] The consistency estimation error of the above line-of-sight distance vector, velocity vector, and acceleration vector is defined as follows:

[0127]

[0128]

[0129] The consistency estimation error described above differs from existing technologies in that it includes a second term; whether this term is included depends on whether the interceptor has a detector, i.e. If detectors are present, the consensus error includes an error estimate based on the detection results, which improves accuracy. Through a distributed observation process, this error estimate based on the detection results is gradually propagated to the error estimates of other interceptors, thereby improving the overall estimation accuracy of the swarm.

[0130] In the above formula, Indicate target The number of interceptors in the target interception group; For the j-th target The elements in the adjacency matrix of the communication topology corresponding to the target interception group, since this invention uses an undirected communication topology, therefore, Representative interceptor With interceptors They can directly exchange information based on the communication network. Representative interceptor With interceptors They cannot communicate directly with each other. Representative interceptor With interceptors The relative line-of-sight distance vector between them; coefficient Representative interceptor With or without detectors; if detectors are equipped, then... ,otherwise .

[0131] , , These are the local estimates of the target's line-of-sight distance vector, velocity vector, and acceleration vector obtained by the interceptor equipped with a detector, based on the detector's measurement information, using a Predefined-Time Extended State Observer (PTESO).

[0132] The preset time-dilation state observer PTESO is designed as follows: Define the local line-of-sight distance vector error as:

[0133] in, The target line-of-sight distance vector obtained by the detector.

[0134] The PTESO for each pilot aircraft can then be designed as follows:

[0135]

[0136]

[0137] in, This is the design gain of the PTESO. Note that the sign(·) in this scheme refers specifically to the sign function in mathematics.

[0138] Defined as:

[0139] in, , It is an adjustable parameter, and can be selected in different preset time-dilation state observers. , It is the preset time of the preset time expansion state observer; This causes the leader to converge first.

[0140] After obtaining the target motion state information provided by the cooperative observation module, it can be input into the cooperative guidance and control module to guide the interceptor to hit the target.

[0141] Step 3: The collaborative guidance and control module performs collaborative guidance and control based on the collaborative observation results, and implements interception.

[0142] The design of the collaborative bootstrapping control module is described below. It should be noted that all variables in this module are marked with a superscript "". "These all indicate that the variable was calculated using the estimated value provided by the collaborative observation module. Such variables will not be explained further below."

[0143] Define the remaining time error for consistency for:

[0144] To ensure that the remaining time consistency error converges within the preset time, the interceptor acceleration command in the line-of-sight coordinate system is designed as follows:

[0145] in, It is the estimated value of the first component in formula (2). It is the projection component of the target acceleration vector estimate onto the line-of-sight direction, i.e.

[0146] in, It is the unit vector of the line-of-sight direction calculated using estimated values. Feedback item. Designed as follows:

[0147] in, , It is an adjustable parameter. This is the preset convergence time.

[0148] To ensure the interceptor accurately hits the target, two acceleration commands in the normal direction of the line-of-sight coordinate system need to be designed. First, the sliding mode variable related to the line-of-sight angular rate is defined as:

[0149]

[0150] Then, the two normal acceleration commands are designed as follows:

[0151]

[0152] in,

[0153]

[0154] in, and These represent the estimated values ​​of the second and third components in formula (2), respectively. and These are the two projection components of the target estimated acceleration vector in the line-of-sight coordinate system, representing the pitch and azimuth channels. , , , and It is the preset convergence time for adjusting the line-of-sight angular rate.

[0155] Finally, the interceptor The command acceleration vector in the inertial coordinate system can be expressed as:

[0156] in, , , It is the representation of the unit vectors of the three axes of the line-of-sight coordinate system in the inertial coordinate system.

[0157] Step 4: The target allocation module performs dynamic target allocation during the interception process.

[0158] This step continuously performs hard constraint detection; hard constraint repair is performed on target interception groups that do not meet the hard constraints; utility degradation detection is performed on target interception groups that meet the hard constraints, and for target interception groups that trigger the utility degradation condition, interceptors with utility gain effects are selected from other target interception groups and added to this target interception group, provided that the hard constraints are met.

[0159] This process consists of four stages, as follows: Phase 1: Distributed hard constraint detection.

[0160] Each interceptor Detect the constraint violations in the target group where it is located, and use the maximum consensus algorithm to propagate the constraint violation information in the communication network. After convergence, all interceptor nodes reach a consensus on which assigned target interception groups violate the hard constraints; if the hard constraints are satisfied, enter Phase 3, and if the hard constraints are not satisfied, enter Phase 2.

[0161] Specifically, this step mainly detects violations of the minimum group size, i.e., the number of people in the group < minGroupSizePerTarget, the leader constraint, i.e., no leader in the group, and the communication connectivity, i.e., the nodes in the group do not form a connected subgraph, etc. Each interceptor Detect the constraint violations in the target group where it is located, and use the maximum consensus algorithm to propagate the constraint violation information in the communication network. After convergence, all interceptor nodes can know which assigned target groups violate the hard constraints.

[0162] Phase 2: Hard constraint repair.

[0163] For the target groups that violate the hard constraints, use the distributed candidate discovery and conflict resolution mechanism. Specifically, considering the utility gain and the feasibility of leaving the original group, each interceptor independently calculates whether it can be a candidate for the target This step is consistent with some details in the utility optimization step, that is, according to whether the utility gain of the interceptor for this target group is greater than zero, it is decided whether to be a candidate condition. In addition, the candidate interceptor must also meet the Join rule and the Leave rule; then select the best candidate through maximum consensus, apply candidate assignment, and update the target assignment result.

[0164] If the hard constraint repair fails, find the leader root for each target, and use the BFS algorithm to reassign all interceptor nodes to the nearest root target to ensure that all hard constraints are satisfied.

[0165] Phase 3: Utility decline detection.

[0166] Only perform utility decline detection on the target interception groups that satisfy the hard constraints. Specifically, calculate the utility scores U between each interceptor and the target in the current target interception group and sum them to obtain the current group utility current_util; compare the current group utility current_util with the group utility calculated based on the initial assignment to obtain the utility decline ratio. When the utility decline ratio is greater than the set value, trigger the use of the maximum consensus algorithm to confirm the trigger of utility optimization within the group; if the majority of nodes in the group confirm the trigger of utility optimization, trigger reallocation.

[0167] Specifically, calculate the current group utility current_util = Σ_{ i ∈group_j} Calculate the utility drop ratio util_drop_ratio = (baseline_util( j ) - current_util) / baseline_util( j If the utility drop ratio is greater than the trigger threshold (util_drop_ratio > utility_drop_threshold), then the maximum consensus algorithm is used to confirm the trigger within the group. If a majority of nodes in the group confirm the trigger, then a redistribution is triggered.

[0168] Phase 4: Utility optimization.

[0169] For each target of triggering utility optimization Candidate interceptors are discovered from the communicable neighbors of other groups. The candidate condition is that the candidate interceptor joins the target. Target interception group against target The utility gain must be greater than zero; simultaneously, for a candidate interceptor to leave the original group, it must be a leaf node within the original group, and the original group must still satisfy the hard constraints after leaving. When multiple candidate interceptors exist, priority is given to candidates with large utility gains and small original group sizes; if multiple targets compete for the same candidate interceptor, the target with the highest priority is selected for the candidate interceptor, and the maximum consensus algorithm is used to reach consensus in the communication network.

[0170] Specifically, for each triggering target Candidate interceptors are discovered from other groups, and the candidate condition is that the candidate interceptor is effective against the target. The utility gain is greater than zero. First, calculate the group utility after inclusion, i.e., future_group_util = Σ_{ i ∈{group_j ∪ candidate}} Calculate the utility of the current group, i.e., current_group_util = Σ_{ i ∈group_j} The utility gain is calculated as `utility_gain = future_group_util - current_group_util`. Only interceptors with a utility gain greater than zero (i.e., `utility_gain > 0`) can be considered as candidates. Furthermore, candidate interceptors must simultaneously satisfy both the Join and Leave rules. The Join rule means that candidate interceptors must be added to the target group. For a group to be successful, at least one neighbor must already be in the target group. In the group, or the target The group is empty. The Leave rule states that for a candidate interceptor to leave the original group, it must be a leaf node within the original group (i.e., it has at most one neighbor in the same group), and the original group must still satisfy the hard constraints after leaving.

[0171] Then, the candidate priority is calculated, i.e., priority = utility_gain / (donor_group_size + 1), prioritizing candidates with high utility gain and small original group size. If multiple targets compete for the same candidate, the target with the highest priority is selected, and the maximum consensus algorithm is used to reach consensus in the communication network. A schematic diagram of the overall framework of the target allocation module algorithm is attached. Figure 2 As shown.

[0172] This concludes the technical process.

[0173] The target allocation, collaborative observation, and collaborative guidance control modules have all been designed. The integrated system adopts a closed-loop architecture of "initial allocation - observation sharing - collaborative guidance control - dynamic reallocation". At the start of the simulation, the optimal leader is selected for each target using a distributed maximum consensus algorithm, and then targets are allocated to slave targets through an improved price auction, completing the initial allocation. During mission execution, at each time step, the distributed observers are updated first. Leaders with detectors use PTESO to estimate the target state, and all interceptors share and reach a consensus on the target state estimate through the communication network. Based on this estimate, each interceptor calculates the collaborative guidance control command. Within the same target group, a preset time synchronization mechanism keeps the remaining flight time consistent, achieving collaborative interception. Simultaneously, the system continuously detects violations of hard constraints (such as insufficient personnel in the group, loss of a leader, or communication failure) and utility degradation (such as group utility degradation exceeding a threshold). Once triggered, distributed reallocation is initiated, and the allocation relationship is locally adjusted through mechanisms such as candidate discovery and conflict resolution to repair constraints or optimize performance. The reallocation results may be applied with a delay to ensure system stability. The entire process forms the following closed loop: allocation determines the targets for observation and guidance control; observation provides the state information required for guidance control; guidance control executes the interception task; and reallocation dynamically optimizes the allocation based on the execution status, achieving coordinated interception of multiple maneuvering targets by multiple interceptors. A schematic diagram of this integrated system architecture is attached. Figure 3 As shown.

[0174] The characteristics of this proposed solution include the following aspects: 1. Distributed initial allocation mechanism with multi-factor utility fusion: This invention integrates multi-factor utility calculations of target value, threat level, geometric factors, time energy, and platform role matching; it adopts a distributed maximum consensus algorithm to select the optimal leader for each target; and it adopts a switching incremental improved price auction algorithm to allocate targets to slaves, allowing allocated slaves to switch targets when the price changes sufficiently.

[0175] 2. Load balancing price auction mechanism based on comprehensive utility: The ideal load of each target is calculated based on the comprehensive utility of the slave; when updating the price, the price of low-load targets is reduced more aggressively to accelerate the transfer of slaves to low-load targets; a load balancing term is added to the net utility calculation to achieve a balance between load balancing and utility maximization.

[0176] 3. Minimal Modification Communication Connectivity Repair Strategy: Based on the price auction results, only disconnected target groups are repaired, while the allocation of connected groups remains unchanged; a breadth-first search algorithm is used to find connected components, retaining the principal components in the original targets and redistributing isolated components to the nearest targets; thus preserving the optimization results of the price auction to the greatest extent possible.

[0177] 4. Design of Distributed Preset Time Observer for Heterogeneous Interceptor Cluster: For interceptors with detectors, PTESO is designed to achieve local estimation of target motion. Then, the distributed preset time observer achieves consistent state estimation of target motion for each interceptor through observation sharing. At the same time, preset time control is adopted to ensure that the observation error converges within a specified time.

[0178] 5. Preset time coordination control law for maneuvering targets: The guidance control command is calculated based on the target state estimated by the observer, and the target acceleration compensation is considered to improve the guidance control accuracy; within the same target group, the remaining flight time of each interceptor is kept consistent through a preset time synchronization mechanism; and time consistency control is achieved by using preset time control.

[0179] 6. Join / Leave rules for distributed dynamic reallocation: The Join rule requires that when an interceptor joins the target group, at least one neighbor is already in the target group or the target group is empty; the Leave rule requires that when an interceptor leaves the original group, it must be a leaf node in the group (at most one neighbor in the same group), and the original group still satisfies the hard constraints after leaving; ensuring that each group always satisfies the communication connectivity constraints during the reallocation process.

[0180] 7. Four-stage redistribution process prioritizing hard constraints: Stage 1 performs distributed hard constraint detection, Stage 2 performs hard constraint repair, Stage 3 performs utility degradation detection, and Stage 4 performs utility optimization.

[0181] 8. Candidate Priority Calculation Strategy: The priority calculation formula is utility gain divided by the original group size plus one, giving priority to candidates with smaller original group sizes; while improving the utility of the target group, the impact on the original group is minimized, and the system stability is maintained.

[0182] 9. Reuse mechanism for initial allocation and dynamic reallocation: Dynamic reallocation reuses the utility calculation, maximum consistency algorithm, communication topology and configuration parameters in the initial allocation; it adopts local repair rather than global reallocation to improve efficiency and maintain system stability.

[0183] 10. Closed-loop integrated architecture: Initial allocation, observation sharing, collaborative guidance and control, and dynamic reallocation form a closed loop; allocation determines the objectives of observation and guidance and control, observation provides the state information required by guidance and control, guidance and control executes interception tasks, and reallocation dynamically optimizes allocation based on the execution situation.

[0184] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A collaborative allocation-observation-guidance control integrated design and operating method for many-to-many interception, applicable to heterogeneous interceptor clusters with some detectors; characterized in that, The method includes: Step 1: The target allocation module performs initial target allocation based on hard constraints. The hard constraints are determined based on the feasibility of the collaborative observation and guidance control process, including: Constraint 1, the leader of the target interception group assigned to the target must have a detector; Constraint 2, all members in the target interception group maintain connectivity; Constraint 3, the members in each interceptor group do not exceed the capacity limit and conform to the smallest group size. Step 2: The collaborative observation module performs joint collaborative observation based on the target allocation results, combining distributed collaborative observation with local observation by the lead aircraft, to obtain collaborative observation results; Step 3: The collaborative guidance and control module performs collaborative guidance and control based on the collaborative observation results, and implements interception; Step 4: The target allocation module performs dynamic target allocation during the interception process: continuously performs hard constraint detection; performs hard constraint repair for target interception groups that do not meet the hard constraints; performs utility degradation detection for target interception groups that meet the hard constraints, and for target interception groups that trigger utility degradation conditions, under the condition of meeting the hard constraints, selects interceptors with utility gain effects from other target interception groups to add to this target interception group.

2. The method as described in claim 1, characterized in that, In step 1, the initial target allocation based on hard constraints is as follows: Step 11: Leader allocation: The interception capability is represented by the utility score U of the interceptor's interception of the target. The distributed maximum consensus algorithm is adopted to enable all interceptors to reach a consensus on the interception capability through communication. Each target is assigned a unique interceptor with a detector as the leader to satisfy constraint 1 in the hard constraints. Step 12: Slave allocation: The interceptors other than the already allocated leader are treated as slaves. Based on the utility score U, the target is allocated slaves using a price auction algorithm, and the allocation is adjusted using the hard constraints.

3. The method as described in claim 2, characterized in that, Step 11, which assigns aircraft to a designated location, further includes conflict resolution to ensure that one designated location corresponds to only one target. All targets are sorted in descending order of their utility score U relative to the leader. Leader conflict is determined for each target in sequence. If the leader initially assigned to the current target is not occupied by other targets, the current leader assignment is confirmed and marked as "assigned". If the leader initially assigned to the current target is occupied by other targets, the unassigned interceptor with the highest utility score U relative to the current target is selected from the remaining unassigned interceptors equipped with detectors and designated as the leader, and the new leader is marked as "assigned".

4. The method as described in claim 2, characterized in that, The slave allocation in step 12 specifically includes: Step 121: Ideal Load Calculation: For each target Calculate the target from all slave devices. The utility scores U of all targets are averaged to obtain the average total utility avg_utility(j); the average total utility avg_utility(j) of all targets is further averaged to obtain the total average utility total_utility; the proportion of the average total utility in the total average utility is multiplied by the total number of slaves to obtain the target utility. The ideal load is determined by ideal_load(j), and the sum of ideal_load(j) and the leader's load is limited to not exceeding the upper limit of the target intercept group capacity. Step 122: Price Auction Algorithm Based on Ideal Load: The ideal load `ideal_load(j)` gives the ideal number of slaves. Under the guidance of `ideal_load(j)`, a price auction method is used to allocate slaves to the target. In each iteration of the price auction, the number of slaves is used... and target Utility scores between and target The current auction price is π_j, and the target value is calculated. Assigned slave net utility util_all( j )= -π_j, and adjust based on the target status and the deviation from the ideal load number of the target interception group; when assigning targets to slaves, the target with the highest net utility is selected as the best choice; calculate the price change based on the difference between the number of slaves assigned to the target and the ideal load number, and update the price π_j accordingly; Step 123: Guaranteed Allocation: If there are still unallocated slaves after the price auction, a greedy strategy is used to force allocation, ensuring that all slaves are allocated; specifically: for each unallocated slave... Calculate all targets relative to slaves The overall utility score, i.e., score(j) = (net utility + target value × kill probability); the slave... Prioritize assigning to the target with the highest score(j) and remaining capacity; if all targets are full, select the target with the lowest current load; if all targets are dead, default to assigning to the designated target. Step 124: Perform constraint repair based on the hard constraints: (a) Local repair: Capacity restoration: If a target interception group exceeds its maximum capacity, remove redundant members, prioritizing the removal of slave devices without detectors; Minimum group size repair: If the number of members in a target interception group is insufficient, members are borrowed from other groups, with priority given to slaves without detectors, while ensuring that the borrowed target interception group still meets the hard constraints; Leader constraint repair: If a target interception group has no leader, a slave with a detector is borrowed from another group, and the borrowed target interception group still satisfies the hard constraints. (b) Connectivity restoration: for each target The Breadth-First Search (BFS) algorithm is used to check if the nodes within the target interception group form a connected subgraph. If they are connected, the original assignments remain unchanged. If they are not connected, BFS is used to find all connected components. A principal component is selected, prioritizing components containing the leader; if no leader is found, the largest component is selected. The principal component is then retained in the original target. Using BFS distance, starting from the root of each target, other isolated components are redistributed to the nearest target; (c) Constraint check: Verify whether all hard constraints are satisfied after the repair; Step 125: Generate initial allocation results.

5. The method as described in claim 4, characterized in that, The price auction algorithm based on ideal load guidance described in step 122 specifically includes: Step 1221: Capacity Update: For each target The remaining capacity cap(j) of each target interception group is recalculated based on the current allocation status. Step 1222: Slave Selection Stage: Slave The competitive objectives are divided into two scenarios, A and B, and will be handled accordingly. Situation A: Slave If a target (current_assign_ii) has been assigned, determine whether to switch to a better target. First, check if the current assignment is still valid. If the assigned target is dead or overloaded, clear the invalid assignment, execute the reselection logic, and jump to case B. If the assigned target is alive and not overloaded, execute the following steps: Step (a1): Calculate the net utility of all objectives util_all(j): First calculate the slave... For each objective The basic net utility, i.e., util_all(j) = -π_j; then adjust util_all(j): determine overload penalty or underload reward based on the difference between the target's current load and ideal load, and use this to adjust the net utility util_all(j) value to optimize the priority of assigning slaves to it; for dead targets, adjust util_all(j) to negative infinity; for non-assigned targets current_assign_ii, if the capacity is full, apply a large penalty to the net utility util_all(j) and reduce the priority of assigning slaves to it; Step (a2): Get the net utility of the current allocation util_all(current_assign_ii); if the capacity of the allocated target current_assign_ii is full, adjust the net utility of the current allocation util_all(current_assign_ii) to negative infinity; Step (a3): Switching judgment: Select the target with the largest net utility util_all(j) as the slave. The best alternative target is best_j; if the capacity of the currently assigned target current_assign_ii is full, a forced switch is determined and step (a4) is executed; if the currently assigned target is not full, and the net utility of the best alternative target best_j is higher than the net utility of the currently assigned target current_assign_ii, and the difference exceeds the threshold, it is determined that a switch is worthwhile, and step (a4) is executed. Step (a4): Perform the switch: Replace the currently assigned target current_assign_ii with the best alternative target best_j and assign it to the slave. ; Situation B: Slave If the allocation is unallocated or has been cleared, proceed with the following steps: Step (b1): Calculate the net utility of all objectives util_all(j); Step (b2): Select the target with the highest net utility as the slave. The best target choice is best_j; Step (b3): ​​Submit request: Slave Submit an allocation request to the best target selection (best_j) and wait for the target approval phase to process it. Step 1223: Goal Approval Phase: For each goal If the number of allocation requests (demand) does not exceed the remaining capacity (remaining_cap), all requests are approved; if the number of allocation requests exceeds the remaining capacity, requesters are sorted in descending order of net utility, the requests at the top are approved, the remaining requests are rejected, and the rejected slaves are reselected in the next round. Step 1224: Price Update Phase: Update the target price based on supply and demand to guide slaves to shift towards low-load targets. Specifically, this includes: First, determining the approved actual load (current_load_after); determining the excess demand (excess_demand) based on the remaining capacity (remaining_cap) and the number of allocation requests (demand); if the actual load (current_load_after) does not exceed the ideal load, then the load is insufficient. Calculate the price reduction factor based on the insufficient load ratio, and then calculate the price change (price_change) based on the price reduction factor (price_reduction_factor) and the excess demand (excess_demand); update the auction price (π_j) using the price change (price_change); if the load is sufficient or overloaded, update the auction price (π_j) using the excess demand (excess_demand). Step 1225: Convergence Judgment: Determine whether the price auction has converged and decide whether to exit the loop.

6. The method as described in claim 2, characterized in that, The utility score U of the interceptor is calculated as follows: Targeting the interceptor The design utility function is : in, These are non-negative weighting coefficients, used to measure the importance of target value, threat level, engagement geometry, time advantage, and role preference, respectively. and These are the target's value and its threat level, respectively. , , , These represent kill probability, engagement geometry score, time advantage bonus, and character preference value, respectively. Probability of lethality Based on combat geometric quality score Detector dependence coefficient and motor vehicle penalty items Sure: Among them, the detector dependence coefficient For interceptors equipped with detectors, Take the detector coefficient of the detector interceptor For interceptors that do not have detectors, Take the detector coefficient of the detectorless interceptor ; Motor penalty items Designed to punish situations where the target's maneuverability exceeds the interceptor's maneuverability, as follows: in, It is the design coefficient. The upper bound of the estimate represents the target normal acceleration. It is the maximum permissible normal acceleration of the interceptor.

7. The method as described in claim 1, characterized in that, Step 4 includes four stages: Phase 1: Distributed Hard Constraint Detection Each interceptor The system detects the constraint violations of its own target group, uses the maximum consensus algorithm to propagate the constraint violation information in the communication network, and after convergence, all interceptor nodes reach a consensus on which assigned target interception groups have violated the hard constraints; if the hard constraints are satisfied, it proceeds to stage 3; if the hard constraints are not satisfied, it proceeds to stage 2. Phase 2: Hard Constraint Repair: For target interception groups that violate hard constraints, based on the utility score U gain and the feasibility of leaving the original group as defined by the hard constraints, each interceptor independently calculates whether it can become a target. The candidate is then determined using the maximum consensus algorithm, and the target allocation result is updated. Then, leader conflict resolution is performed to ensure that one leader corresponds to only one target. If hard constraint repair fails, find the leader root for each target and use the BFS algorithm to reassign all interceptor nodes to the nearest root target to ensure that all hard constraints are satisfied. Phase 3: Utility Decline Detection: Utility decline detection is performed only on target interception groups that meet hard constraints. Specifically, the utility scores U of each interceptor and target in the current target interception group are calculated and summed to obtain the current group utility current_util. The current group utility current_util is compared with the group utility calculated based on the initial allocation to obtain the utility decline ratio. When the utility decline ratio is greater than a set value, the maximum consensus algorithm is triggered to confirm the triggering of utility optimization within the group. If a majority of nodes in the group confirm the triggering of utility optimization, a reassignment is triggered. Phase 4: Utility Optimization: For each objective that triggers utility optimization... Candidate interceptors are discovered from the communicable neighbors of other groups. The candidate condition is that the candidate interceptor joins the target. Target interception group against target The utility gain is greater than zero; at the same time, for a candidate interceptor to leave the original group, it must be a leaf node in the original group, and the original group still satisfies the hard constraints after leaving the original group. When multiple candidate interceptors are available, priority is given to candidates with high utility gain and small original group size. If multiple targets compete for the same candidate interceptor, the target with the highest priority is selected as the candidate interceptor, and the maximum consensus algorithm is used to reach a consensus in the communication network.

8. The method as described in claim 1, characterized in that, The joint collaborative observation adopted by the collaborative observation module is as follows: All interceptors in the target interception group use a distributed observer to estimate the target motion state; the consistency estimation error in the distributed observer includes the estimation difference error term between itself and its neighbors and the local target estimation error term that is only valid for the leader aircraft; The leader directly measures and estimates the local state of the target through its own detector, and uses the local state of the target to construct the local target estimation error term.

9. The method as described in claim 8, characterized in that, All interceptors use a preset time-based distributed observer; the leader uses a preset time-extended state observer (PTESO). The preset time-distributed observer is: in, , and Interceptors The target estimated by the upper observer Estimates of the line-of-sight distance vector, velocity vector, and acceleration vector; For interceptors The velocity vector; It is the gain coefficient of the distributed observer; For the sign function; Consistency estimation error The function of consensus estimation error Consistency estimation errors including line-of-sight distance vector, velocity vector, and acceleration vector , , ; Defined as: in, , It is an adjustable parameter, and the values ​​are taken separately in different distributed observers. , This is the preset time for the distributed observer; The consistency estimation error of the above line-of-sight distance vector, velocity vector, and acceleration vector is defined as follows: in, Indicate target The number of interceptors in the target interception group; For the goal The elements in the communication topology adjacency matrix corresponding to the target interception group. Representative interceptor With interceptors They can directly exchange information based on the communication network. Representative interceptor With interceptors They cannot communicate directly; Representative interceptor With interceptors The relative line-of-sight distance vector between them; coefficient Representative interceptor With or without detectors; if detectors are equipped, then... ,otherwise ; , , These are the local estimates of the target line-of-sight distance vector, velocity vector, and acceleration vector obtained by the interceptors equipped with detectors in the target interception group based on the detector measurement information and using a preset time-dilation state observer (PTESO). The preset time-dilation state observer PTESO is designed as follows: in, , The target line-of-sight distance vector obtained by the detector; It is the gain coefficient of the preset time-dilation state observer; Defined as: in, , It is an adjustable parameter, and can be selected in different preset time-dilation state observers. , It is the preset time of the preset time expansion state observer; .

Images