Multi-unmanned aerial vehicle cooperative task planning method based on dynamic contract and consensus auction

Abstract

The invention discloses a multi-unmanned aerial vehicle cooperative task planning method based on a dynamic contract and consensus auction. The method comprises the following steps: firstly, generating a dynamic contract containing a basic value, a commitment bonus and a penalty for default linked with a system re-planning frequency, and coding a system target as economic incentive; the central controller performs initial task allocation according to the capability of the unmanned aerial vehicle, then the unmanned aerial vehicle can initiate local consensus auction based on a contract to compete for a new task, and neighbor unmanned aerial vehicles cooperatively determine the optimal attribution of the task through weighted voting. The system monitors the re-planning frequency at the same time, and dynamically adjusts the penalty of default in the contract to form a negative feedback closed loop, thereby adaptively maintaining the stability of the system. According to the method, individual rational decision making is guided through economic incentive, efficient cooperation is achieved through distributed consensus auction, and the task efficiency, the resource utilization rate and the overall robustness of the multi-unmanned-aerial-vehicle system in the dynamic environment are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology for intelligent unmanned systems, and in particular to a multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions. Background Technology

[0002] Multi-UAV collaborative task planning is a core technology in UAV swarm applications. Its goal is to rationally allocate multiple tasks to UAVs in a swarm and coordinate them to complete the tasks efficiently and reliably in a dynamic and uncertain environment. Traditional task planning methods are mainly divided into centralized and distributed approaches, but both have significant limitations.

[0003] Centralized approaches typically rely on a central controller (such as a ground control station) to collect global information and perform unified planning and allocation. While these approaches can theoretically seek globally optimal solutions, they face significant challenges in practical applications: First, the central controller needs to maintain high-bandwidth, low-latency communication with all UAVs, resulting in a heavy communication burden and susceptibility to interference. Second, as the number of UAVs and the complexity of missions increase, the computational complexity of solving the global optimization problem grows exponentially, making real-time performance difficult to guarantee. More critically, centralized architectures are susceptible to single points of failure; if the central controller fails, the entire system will be paralyzed.

[0004] Distributed methods decentralize decision-making power to individual drones, allocating tasks through local communication and negotiation. These methods offer advantages such as high scalability and robustness. However, purely distributed negotiation often lacks an effective global coordination mechanism, easily leading to local optima or conflicts. In the classic market auction-based approach, drones, acting as selfish, rational individuals, bid for tasks based solely on local information. This can easily lead to excessive competition, resulting in frequent task switching and replanning, causing system oscillations and severely impairing overall execution efficiency and stability. Furthermore, designing collaborative rules that incentivize individual participation, constrain speculative behavior, and adapt to environmental changes is a core challenge for distributed methods.

[0005] In existing technologies, there are few solutions that can effectively balance the relationship between individual autonomy, global efficiency, and system stability. Therefore, there is an urgent need for an innovative collaborative task planning method that can leverage the resilience and survivability of distributed systems while, through ingenious mechanism design, guide self-interested drones to spontaneously achieve overall system efficiency, stability, and collaboration in the pursuit of local benefits. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A multi-UAV collaborative mission planning method based on dynamic contracts and consensus auctions includes the following steps: S1: When the emergency command vehicle system mounted on the vehicle issues task T1, the dynamic contract generation system in the emergency command vehicle system generates a dynamic commitment contract C(T1).

[0008] Furthermore, the vehicles mentioned are central command vehicles and dispatch vehicles; The emergency command vehicle system is a special vehicle system integrating communication and command functions, including a task management and injection system, a dynamic contract generation system, a global situational awareness and data fusion system, communication and network management, and other auxiliary systems. It is a software algorithm suite that integrates task management, contract calculation, global optimization, situational fusion, communication management, and stability monitoring. The task management and injection system is used to issue task T1, and the dynamic contract generation system is used to generate dynamic commitment contract C(T1).

[0009] Furthermore, the dynamic contract generation system, based on the attributes of task T1, including target location distance, type BaseValue (type), resource consumption cost, and deadline t, further... deadline ; The target location distance is the geometric distance from a fixed "reference point" preset by the emergency command vehicle system to mission T; the type includes strategic reconnaissance, tactical reconnaissance, precision strike, and material transportation; the resource consumption cost is the amount of resources required to execute mission T; the deadline t deadline To meet the deadline for completing the task T, the resources include electricity or fuel; The dynamic commitment contract C(T1) includes the basic task value V. b The promised bonus B(t) and the penalty for breach of contract P.

[0010] Furthermore, the dynamic commitment contract C(T1) is expressed as: C(T) = (V b ,B(t),P), where: The basic value V of the task b Calculated from the inherent attributes of the task; Calculation method:

[0011] in: Weight 1+2+ 3+ 4 = 1; λ is the geographical proximity gain coefficient; η is the time urgency gain coefficient; γ is the resource consumption penalty coefficient; ϵ is a constant; w i These are the weighting coefficients; λ, η, γ, ϵ and w i All settings are manually configured during system initialization. distance is the target location, cost is the resource consumption, and t is the target location. deadline The deadline; BaseValue(type) is a baseline value function for task types, returning a corresponding preset baseline value based on the task type. The type includes strategic reconnaissance, tactical reconnaissance, precision strike, and material transport. Specifically, the baseline value for strategic reconnaissance ranges from 100 to 200, for tactical reconnaissance from 100 to 200, for precision strike from 200 to 300, and for material transport from 100 to 200. The specific values ​​of the baseline values ​​for each task type in BaseValue(type) are manually set during the initialization of the emergency command vehicle system, reflecting the strategic / tactical importance of different task types. in This indicates that the value gain generated by distance decays rapidly with increasing distance, which will guide the drone swarm to spontaneously form a "regional responsibility system," naturally reducing the total travel loss and total mission delay of the system and improving overall efficiency; This means that value increases with time sensitivity, ensuring that urgent tasks receive an overwhelmingly high value score in consensus auctions, thus almost inevitably being allocated and executed quickly, meeting the most critical time constraints in tactical operations; all drones and the central controller use the exact same formula and parameters to calculate the V for the same task. b This ensures the objectivity and consistency of value assessment, providing the mathematical basis for voters to make fair comparisons and decisions during subsequent consensus auctions. The combination of positive and negative terms—positive terms incentivizing task acceptance and negative terms providing cost constraints—makes the assessment more rational and comprehensive, avoiding rash decisions that simply pursue high value while ignoring execution costs.

[0012] The promised bonus B(t) is expressed as: B(t) = B0 × (1 - e^(-t / τ)) is the commitment bonus that accumulates over time, accumulating with the time t for which the task is committed to be executed. It is used to incentivize drones to faithfully execute assigned tasks in the long term. The reward for drones that continuously execute tasks is increased exponentially. Here, B0 is the bonus cap, t is the commitment time, and τ is the time constant. P = P0 × (1 + α·R) is the penalty for breach of contract based on the system replanning frequency R, and its value increases with the system's historical replanning frequency R. Where P0 is the base penalty; α is the adjustment gain coefficient; and R is the reprogramming frequency.

[0013] The emergency command vehicle system utilizes a carefully designed dynamic commitment contract C(T1)=(V b B(t), P), guides and converges the individual's pursuit of local interests into the long-term, stable and efficient global goal of the entire system.

[0014] S2: The emergency command vehicle system is based on the capability vectors A of each UAV. i Calculate the matching degree M(A,T) between the UAV and the dynamic commitment contract C(T1), perform initial task allocation, form an initial allocation set {(A,C(T1))}, and achieve preliminary optimization of resource allocation; Furthermore, the capability vector A i Represented as: A i =[Cap type Cap position Cap resource Cap speed Cap specialty Cap status ]; Among them, the task type capability Cap type The set of task types that drones can perform; location capability (Cap) position The current coordinates of the drone (x) i ,y i ,z i Resource Capacity resource : Remaining battery or fuel, payload; speed capability (Cap) of the drone. speed The maximum cruising speed of the drone, v max Exclusive Professional Capabilities specialty This includes the types of sensors mounted on the UAV, weapon accuracy levels, communication relay capabilities, and anti-jamming levels; state capabilities (Cap). status Current task status (idle / busy), health status (normal / faulty); the task status is set to idle when the emergency command vehicle system is initialized.

[0015] Furthermore, the matching degree M(A,T) is calculated using the following formula: M(A,T)=a·M type +b·M distance +c·M resource +d·M time ; Where a+b+c+d=1, and a, b, c, and d are weighting coefficients; M type For task type matching degree, if the task type capability Cap type If the task type is included, then M type =1; otherwise, M type =0; M distance M represents the distance matching degree. distance =1 / (1+d(A,T) / D_scale), where d(A,T) is the actual distance from the current position of the UAV to the position of mission T; D_scale = distance scaling constant; M resource For resource matching degree, M resource =min(1,E remain / E required ), where E required To estimate the total energy consumption required to perform the task, E remain This refers to the current remaining fuel / electricity of the drone. M time For time matching degree, if (d(A,T) / v max +t exec )<(t deadline -t current ), then M time =1; otherwise, M time =0; where d(A,T) is the distance from the current position of the UAV to the position of the target T; v max t is the maximum cruising speed of the drone; exec The time required to complete the task itself; t deadline -t current This represents the remaining available time.

[0016] The emergency command vehicle system calculates the matching degree M(A,T) based on the capabilities of each UAV. It then assigns task T to UAV A, which has the highest matching degree M(A,T). S3: Drone A, which has the highest M(A,T) in S2, executes task T1 assigned in S2. When drone A locally senses a new high-value task T2, drone A calculates the expected net benefit G1=V of abandoning task T1 and bidding for the high-value task T2. b 2-P1, and the expected net return G2 for persisting with task T1: G2=V b 1+B(t1); Drone A initiates a consensus auction request for high-value task T2 to local neighboring drones if and only if G1>θ·G2; Where θ is the preset switching threshold, V b1 represents the base value of task T1, P1 represents the penalty for breach of contract in task T1, and V represents the base value of task T1. b 2 represents the base value of task T2, and B(t1) represents the accumulated bonus of task T1. The local neighborhood drones refer to other drone groups that are within the current communication range of drone A, which initiates the auction request, and can directly communicate with and participate in the decision-making process. The consensus auction request is a collaborative invitation initiated by the drone that discovers the new task to the drones in the local neighborhood.

[0017] S4: The drones that receive the consensus auction request collaboratively determine the best allocation of the high-value task T2 through a weighted voting mechanism, forming a local consensus and completing the reallocation; Furthermore, the execution process of the weighted voting mechanism includes the following steps: S41: Drone V that received a consensus auction request within a local neighborhood i As a potential bidder, based on the contract C(T2) of the new task T2 and its current state, it conducts a local bidding value assessment, generates a quantitative assessment value G3, and decides whether to participate in the bidding based on a preset threshold E.

[0018] Furthermore, the local bidding value assessment is for the drone V. i Calculate the expected net gain G of abandoning the existing task (if any) and instead executing task T2. 3i The calculation formula is as follows: G 3i =(V b 2-P i )-θ(V b 3+B(t3)); Among them, V b 2 represents the basic task value of T2; P i For drones V i The penalty for abandoning the existing mission, if the drone V i If P is in a no-task state, then i =0; θ is the preset switching threshold, which is the same as the threshold used when the drone initiates the auction request in S3; V b 3 is the drone V i The basic mission value of performing existing tasks, if the drone V i If V is in a no-task state, then b 3 is 0; B(t3) is the value of the drone V. i The accumulated bonuses from performing existing tasks, if drone V i If the task is not active, then B(t3) is 0.

[0019] Furthermore, if G 3i>E, then the drone V i Marked as qualified bidder B j And generate an evaluation report; If G 3i If V ≤ E, then the drone V i Bidder C marked as unqualified j They will not participate in the bidding.

[0020] The evaluation report includes the drone's unique ID and drone capability vector A. i Estimated completion time t j G 3i ; Estimated completion time t j Based on its current location, speed, and task flow, the estimated time required to complete task T2; S42: Unmanned Aerial Vehicle (UAV) i For drone B j Calculate the overall voting weight W ij The comprehensive voting weight W ij =Trust ij ×Sim ij ×Cap ij ,in, Dynamic Trust ij Based on drones V i With drone B j The historical cooperation results are dynamically updated, and the update rules are as follows: Trust ij(t+1) =Trust ij(t) +Δ1×I success -Δ2×I breach Where Δ1 represents the incremental value of successful cooperation, and Δ2 represents the reduction in value due to default; I success with I breach The indicator function takes a value of 1 when there is successful cooperation or a breach of contract, and 0 otherwise; the dynamic trust level is... ij Constrained within the interval [0,1]; Context similarity Sim ij : Measuring the V of drones i The degree of similarity between the current situation and bidding task T2 is used to determine the accuracy of the judgment. The formula is as follows: Sim ij =β×Sim cosine +(1-β)×Sim context ; Among them, Sim cosine Sim is a cosine similarity algorithm based on task feature vectors, considering type, environment, and urgency. contextSimilarity is based on contextual differences such as time and location; β is the weighting coefficient. Capability Confidence ij Drone V i For drone B j The ability to successfully complete task T2 is assessed using the following formula: Cap ij =ω1×Match ij +ω2×Hist ij +ω3×Status ij ; Among them, Match ij For the task requirements and B j The matching degree of the capability vector; Hist ij For B j Historical success rate and quality score for performing similar tasks; Status ij For B j Current real-time status (resource availability, health); ω1, ω2, ω3 are weighting coefficients and their sum is 1; S43: V per drone i It allocates its voting rights to the bidders it supports; typically, drones V i Cast your vote for weight W ij The highest bidder; the emergency command vehicle system aggregates all weighted votes and calculates the bid proposal D. j Vote (D) j ): Let V = {V1, V2, ..., V} m} is the voter (the drone V that receives the consensus auction request) i )gather; B = {B1, B2, ..., B} n} is the bidder (Drone B participating in the bidding) j )gather;

[0021] Where m is the total number of elements in set V, and n is the total number of elements in set B; W ij For drones V i For drone B j The overall weight is calculated as follows: ， W ij =Trust ij ×Sim ij ×Cap ij ; I ij Let V be the indicator function, representing the unmanned aerial vehicle (UAV). iShould I vote for drone B? j :

[0022] S44: The emergency command vehicle system has a preset consensus threshold T c If a certain solution D exists... j Vote (D) j )>T c If this is achieved, it is considered a partial consensus, and the high-value task T2 and its dynamic commitment contract C(T2) are assigned to the corresponding winning drone A. j If consensus cannot be reached in the first round, a multi-round voting process will be initiated, by lowering the consensus threshold T. c To promote the formation of consensus on auctions; S45: After the auction consensus described in S44 is reached, the winning drone A... j The system immediately acquires the right to execute the high-value task T2 and commits to accumulating a bonus B(t) starting from time 0 in the dynamic commitment contract C(T2); the contract for task T1 is marked as defaulted, and the corresponding penalty P1 is recorded. The result of the entire consensus auction is reported to the emergency command vehicle system.

[0023] S5: The emergency command vehicle system continuously monitors all task reassignment events generated by S3 and S4, calculates the smoothed system replanning frequency through an exponential smoothing filtering algorithm, and feeds this frequency back to the dynamic contract generation system of S1, thereby dynamically adjusting the penalty P for breach of contract in all future contracts, thus forming a closed-loop stability control mechanism.

[0024] Furthermore, the execution process and principle of the closed-loop stability control mechanism include the following steps: S51: The system monitoring and stability controller of the emergency command vehicle system listens for and records replanning events consisting of the following two types of events in real time: S3 Triggering Event: Consensus auction request initiated by any drone based on G1>θ·G2; S4 Completion Event: Any consensus is reached, resulting in the task being reassigned; Within a fixed time window Δt (ideally 1 minute), the number of replanning events that occur is counted and denoted as the instantaneous replanning count N. t ; S52: Calculate the instantaneous replanning frequency r based on the instantaneous count in S51. t :r t =N t / Δt The instantaneous replanning frequency r t It directly reflects the degree of system volatility within the most recent time window.

[0025] S53: To avoid excessive impact from instantaneous fluctuations on the system, an exponential smoothing filter algorithm is used to smooth the instantaneous frequency, and the system reprogramming frequency R is calculated. t The calculation formula is: R t =λ⋅R t−1 +(1−λ)⋅r t ; Among them, R t The system replanning frequency at time t (times / unit of time) is a core smoothing indicator reflecting the overall stability of the system. R t−1 The system replanning frequency of the previous moment carries historical state information; r t The instantaneous reprogramming frequency at the current moment is the latest observation value; λ is the forgetting factor, and its value ranges from 0 to 1.

[0026] S54: Calculate the latest system replanning frequency R t The dynamic contract generation system provides real-time feedback to step S1, and subsequently generates a dynamic commitment contract C(T)=(V) for any newly released or discovered task. b When B(t), the penalty P for breach of contract will be based on the smoothed R. t (instead of the original r) t Perform dynamic calculations: P = P0 × (1 + α⋅R) t ); Where P0 is the base penalty and α is the adjustment gain coefficient.

[0027] The above mechanism constitutes a complete negative feedback closed loop, realizing the adaptive stability control of the system: When the emergency command vehicle system detects frequent and persistent replanning: the smoothing frequency R t The penalty P for breach of the new dynamic commitment contract increases steadily, and the penalty P for the drone abandoning the current mission during the S3 decision-making process increases significantly. i As the value increases, the expected net benefit of switching to a new task G1=V j -P i The reduction effectively curbed the incentive for drones to arbitrarily default on contracts and initiate auctions, and the system's replanning behavior gradually decreased, returning to stability.

[0028] When the system is running stably: smoothing frequency R t By maintaining a low level, the penalty P for breach of the new dynamic commitment contract is reduced accordingly, the cost of switching drone missions is reduced, and the system maintains its ability to respond quickly and flexibly to sudden high-value missions and local environmental changes.

[0029] Through closed-loop stability control in step S5, the method innovatively employs an exponential smoothing filter algorithm to intelligently distinguish between the "trend" and "noise" of system fluctuations, thereby dynamically adjusting the game rules (penalties in the contract). This enables multi-UAV systems to autonomously and robustly maintain overall global stability without sacrificing local response agility, significantly improving the long-term robustness and overall mission performance of large-scale distributed autonomous systems in complex dynamic environments.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. A hybrid architecture of "centralized initialization (S2) + distributed replanning (S3-S4)" is adopted. The central controller (emergency command vehicle) only performs an optimized allocation once based on global information at the initial stage of mission deployment, providing an efficient starting point for the system. Thereafter, the real-time replanning decision-making power is completely delegated to the UAV swarm, which autonomously completes the process through local communication and consensus auction. This frees the central controller from heavy real-time micro-management, significantly reduces communication load and computational pressure, and allows the system to maintain basic operation through distributed negotiation even after the central node fails, significantly improving the system's scalability and resilience. 2. Each task is accompanied by quantified value, reward, and cost tags (V). b ,B(t),P). All local decisions made by the drone (such as whether to initiate an auction S3, whether to bid S41) are translated into the calculation of "expected net revenue" based on this contract. V b The design elements in the formula (such as distance decay and urgency incentives) subtly guide the behavior of drones towards overall efficiency; B(t) incentivizes long-term commitments; and P punishes arbitrary breaches of contract. In this way, in the distributed decision-making process where each drone is purely pursuing its own self-interest maximization, its behavior is automatically guided by the economic rules of the contract, spontaneously giving rise to the coordinated and efficient behaviors required by the system, thus achieving global coordination. 3. The system continuously monitors the replanning frequency R and obtains the smoothing index R through exponential smoothing filtering. t This information is fed back to the contract generation stage to dynamically adjust the penalty P for breach of contract in new tasks. When system replanning is frequent (R... t When the system is elevated, P automatically increases, raising the cost of switching missions for the UAV and thus suppressing speculative behavior, allowing the system to "cool down." When the system is stable, P decreases, maintaining responsiveness. This forms a powerful negative feedback loop, enabling the system to have self-diagnosis and self-adjustment capabilities, fundamentally suppressing oscillations, and achieving a dynamic balance between stability and agility in dynamic environments. 4. A structured "consensus auction" and "weighted voting" mechanism (S4) is proposed. This mechanism combines the competitive efficiency of market auctions with the collaborative fairness of consensus mechanisms. The weighted voting weight W... ijBy integrating trust levels, scenario similarity, and capability confidence, the decision-making process is based on objective capabilities while also incorporating historical cooperation reputation and current situational understanding. This ensures that tasks are assigned to the most reliable and suitable drones, rather than simply "highest bidder wins." The voting process is efficient, quantifiable, and can reach consensus through multiple iterations, significantly improving the quality, fairness, and credibility of local group decision-making. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the implementation of Embodiment 1 of the present invention. Detailed Implementation

[0032] To provide a further understanding of the purpose, structure, features, and functions of the present invention, detailed descriptions are provided below with reference to specific embodiments.

[0033] A multi-UAV collaborative mission planning method based on dynamic contracts and consensus auctions includes the following steps: S1: When the emergency command vehicle system mounted on the vehicle issues task T1, the dynamic contract generation system in the emergency command vehicle system generates a dynamic commitment contract C(T1).

[0034] Furthermore, the vehicles mentioned are central command vehicles and dispatch vehicles; The emergency command vehicle system is a special vehicle system integrating communication and command functions, including a task management and injection system, a dynamic contract generation system, a global situational awareness and data fusion system, communication and network management, and other auxiliary systems. It is a software algorithm suite that integrates task management, contract calculation, global optimization, situational fusion, communication management, and stability monitoring. The task management and injection system is used to issue task T1, and the dynamic contract generation system is used to generate dynamic commitment contract C(T1).

[0035] Furthermore, the dynamic contract generation system, based on the attributes of task T1, including target location distance, type BaseValue (type), resource consumption cost, and deadline t, further... deadline ; The target location distance is the geometric distance from a fixed "reference point" preset by the emergency command vehicle system to mission T; the type includes strategic reconnaissance, tactical reconnaissance, precision strike, and material transportation; the resource consumption cost is the amount of resources required to execute mission T; the deadline t deadline To meet the deadline for completing the task T, the resources include electricity or fuel; The dynamic commitment contract C(T1) includes the basic task value V. b The promised bonus B(t) and the penalty for breach of contract P.

[0036] Furthermore, the dynamic commitment contract C(T1) is expressed as: C(T) = (V b ,B(t),P), where: The basic value V of the task b Calculated from the inherent attributes of the task; Calculation method:

[0037] in:

[0038] Weight 1+ 2+ 3+ 4 = 1; λ is the geographical proximity gain coefficient; η is the time urgency gain coefficient; γ is the resource consumption penalty coefficient; ϵ is a constant; w i These are the weighting coefficients; λ, η, γ, ϵ and w i All settings are manually configured during system initialization. distance is the target location, cost is the resource consumption, and t is the target location. deadline The deadline; BaseValue(type) is a baseline value function for task types, returning a corresponding preset baseline value based on the task type. The type includes strategic reconnaissance, tactical reconnaissance, precision strike, and material transport. Specifically, the baseline value for strategic reconnaissance ranges from 100 to 200, the baseline value for tactical reconnaissance ranges from 100 to 200 (preferred values ​​are 130 and 120), the baseline value for precision strike ranges from 200 to 300 (preferred value is 250), and the baseline value for material transport ranges from 100 to 200 (preferred value is 150). The specific values ​​of the baseline values ​​for each task type in BaseValue(type) are manually set during the initialization of the emergency command vehicle system, reflecting the strategic / tactical importance of different task types. in This indicates that the value gain generated by distance decays rapidly with increasing distance, which will guide the drone swarm to spontaneously form a "regional responsibility system," naturally reducing the total travel loss and total mission delay of the system and improving overall efficiency; This means that value increases with time sensitivity, ensuring that urgent tasks receive an overwhelmingly high value score in consensus auctions, thus almost inevitably being allocated and executed quickly, meeting the most critical time constraints in tactical operations; all drones and the central controller use the exact same formula and parameters to calculate the V for the same task. bThis ensures the objectivity and consistency of value assessment, providing the mathematical basis for voters to make fair comparisons and decisions during subsequent consensus auctions. The combination of positive and negative terms—positive terms incentivizing task acceptance and negative terms providing cost constraints—makes the assessment more rational and comprehensive, avoiding rash decisions that simply pursue high value while ignoring execution costs.

[0039] The promised bonus B(t) is expressed as: B(t) = B0 × (1 - e^(-t / τ)) is the commitment bonus that accumulates over time, accumulating with the time t for which the task is committed to be executed. It is used to incentivize drones to faithfully execute assigned tasks in the long term. The reward for drones that continuously execute tasks is increased exponentially. Here, B0 is the bonus cap, t is the commitment time, and τ is the time constant. P = P0 × (1 + α·R) is the penalty for breach of contract based on the system replanning frequency R, and its value increases with the system's historical replanning frequency R. Where P0 is the base penalty; α is the adjustment gain coefficient; and R is the reprogramming frequency.

[0040] The emergency command vehicle system utilizes a carefully designed dynamic commitment contract C(T1)=(V b B(t), P), guides and converges the individual's pursuit of local interests into the long-term, stable and efficient global goal of the entire system.

[0041] The dynamic commitment contract C(T1) is designed to include a fundamental value V. b The "dynamic commitment contract" of time-accumulated bonus B(t) and system state feedback penalty P encodes the system's global goals (efficiency, stability, and cooperation) into quantitative rules that drive each UAV to perform selfish economic calculations. This enables UAVs to spontaneously exhibit the collaborative, efficient, and stable group behavior required by the system in the distributed decision-making process of purely pursuing their own profit maximization. This solves the fundamental contradiction between individual autonomous decision-making and overall system stability in large-scale multi-UAV systems under dynamic and uncertain environments when pursuing task execution efficiency.

[0042] S2: The emergency command vehicle system is based on the capability vectors A of each UAV. i Calculate the matching degree M(A,T) between the UAV and the dynamic commitment contract C(T1), perform initial task allocation, form an initial allocation set {(A,C(T1))}, and achieve preliminary optimization of resource allocation; Furthermore, the capability vector A i Represented as: A i =[Cap type Cap position Cap resource Cap speed Capspecialty Cap status ]; Among them, the task type capability Cap type The set of task types that drones can perform; location capability (Cap) position The current coordinates of the drone (x) i ,y i ,z i Resource Capacity resource : Remaining battery or fuel, payload; speed capability (Cap) of the drone. speed The maximum cruising speed of the drone, v max Exclusive Professional Capabilities specialty This includes the types of sensors mounted on the UAV, weapon accuracy levels, communication relay capabilities, and anti-jamming levels; state capabilities (Cap). status Current task status (idle / busy), health status (normal / faulty); the task status is set to idle when the emergency command vehicle system is initialized.

[0043] This capability vector definition unifies heterogeneous physical UAVs into a structured, quantifiable digital model. As a standardized "data interface" for the entire system to conduct resource assessment and task adaptation, it provides an objective, consistent, and computable basis for the global matching degree calculation (S2) of the central command vehicle and the distributed consensus voting (S4) among UAVs. Thus, it fundamentally solves the core problem of resource scheduling and coordination in large-scale, multi-type UAV swarms: how to efficiently and automatically determine "who can perform what task at what level".

[0044] Furthermore, the matching degree M(A,T) is calculated using the following formula: M(A,T)=a·M type +b·M distance +c·M resource +d·M time ; Where a+b+c+d=1, and a, b, c, and d are weighting coefficients; M type For task type matching degree, if the task type capability Cap type If the task type is included, then M type =1; otherwise, M type =0; M distance M represents the distance matching degree. distance =1 / (1+d(A,T) / D_scale), where d(A,T) is the actual distance from the current position of the UAV to the position of mission T; D_scale = distance scaling constant; M resourceFor resource matching degree, M resource =min(1,E remain / E required ), where E required To estimate the total energy consumption required to perform the task, E remain This refers to the current remaining fuel / electricity of the drone. M time For time matching degree, if (d(A,T) / v max +t exec )<(t deadline -t current ), then M time =1; otherwise, M time =0; where d(A,T) is the distance from the current position of the UAV to the position of the target T; v max t is the maximum cruising speed of the drone; exec The time required to complete the task itself; t deadline -t current This represents the remaining available time.

[0045] This matching degree calculation formula integrates four key constraints—type compatibility, geographical proximity, resource sufficiency, and time feasibility—with optimization objectives into a quantifiable comprehensive score. It provides an efficient and automated decision-making tool for emergency vehicle command systems, solving the core allocation problem of how to quickly and objectively select the best executor capable of reliably and efficiently completing specific tasks from among numerous heterogeneous UAVs. Its advantage lies in transforming complex multi-objective optimization into a single-objective score with adjustable weights, ensuring both the physical feasibility and timeliness of allocation, and enabling flexible adaptation to different tactical priorities (such as rapid response and resource conservation) through weight coefficients.

[0046] The emergency command vehicle system calculates the matching degree M(A,T) based on the capabilities of each UAV. It then assigns task T to UAV A, which has the highest matching degree M(A,T). S3: Drone A, which has the highest M(A,T) in S2, executes task T1 assigned in S2. When drone A locally senses a new high-value task T2, drone A calculates the expected net benefit G1=V of abandoning task T1 and bidding for the high-value task T2. b 2-P1, and the expected net return G2 for persisting with task T1: G2=V b 1+B(t1); Drone A initiates a consensus auction request for high-value task T2 to local neighboring drones if and only if G1>θ·G2; Where θ is the preset switching threshold, V b 1 represents the base value of task T1, P1 represents the penalty for breach of contract in task T1, and V represents the base value of task T1. b2 represents the base value of task T2, and B(t1) represents the accumulated bonus of task T1. The local neighborhood drones refer to other drone groups that are within the current communication range of drone A, which initiates the auction request, and can directly communicate with and participate in the decision-making process. The consensus auction request is a collaborative invitation initiated by the drone that discovers the new task to the drones in the local neighborhood.

[0047] S4: The drones that receive the consensus auction request collaboratively determine the best allocation of the high-value task T2 through a weighted voting mechanism, forming a local consensus and completing the reallocation; Furthermore, the execution process of the weighted voting mechanism includes the following steps: S41: Drone V that received a consensus auction request within a local neighborhood i As a potential bidder, based on the contract C(T2) of the new task T2 and its current state, it conducts a local bidding value assessment, generates a quantitative assessment value G3, and decides whether to participate in the bidding based on a preset threshold E.

[0048] Furthermore, the local bidding value assessment is for the drone V. i Calculate the expected net gain G3 of abandoning the existing task (if any) and instead executing task T2. i The calculation formula is as follows: G 3i =(V b 2-P i )-θ(V b 3+B(t3)); Among them, V b 2 represents the basic task value of T2; P i For drones V i The penalty for abandoning the existing mission, if the drone V i If P is in a no-task state, then i =0; θ is the preset switching threshold, which is the same as the threshold used when the drone initiates the auction request in S3; V b 3 is the drone V i The basic mission value of performing existing tasks, if the drone V i If V is in a no-task state, then b 3 is 0; B(t3) is the value of the drone V. i The accumulated bonuses from performing existing tasks, if drone V i If the task is not active, then B(t3) is 0.

[0049] The steps involve defining a local bidding decision rule based on economic rationality, transforming the complex coordination problem in a distributed system into a comparison of expected net benefits that each drone can independently calculate. This allows for the selection of truly motivated and capable qualified bidders to perform new tasks. Its advantage lies in reducing invalid bidding and communication overhead at the source. Furthermore, through cost-benefit calculations linked to a global contract mechanism (considering penalties for breach of contract and accumulated bonuses), it ensures that the spontaneity of local decisions does not compromise the overall stability and efficiency of the system. This effectively solves the core challenge of achieving efficient, orderly, and stable task competition and redistribution in large-scale distributed clusters.

[0050] Furthermore, if G 3i >E, then the drone V i Marked as qualified bidder B j And generate an evaluation report; If G 3i If V ≤ E, then the drone V i Bidder C marked as unqualified j They will not participate in the bidding.

[0051] The evaluation report includes the drone's unique ID and drone capability vector A. i Estimated completion time t j G 3i ; Estimated completion time t j Based on its current location, speed, and task flow, the estimated time required to complete task T2; S42: Unmanned Aerial Vehicle (UAV) i For drone B j Calculate the overall voting weight W ij The comprehensive voting weight W ij =Trust ij ×Sim ij ×Cap ij ,in, Dynamic Trust ij Based on drones V i With drone B j The historical cooperation results are dynamically updated, and the update rules are as follows: Trust ij(t+1) =Trust ij(t) +Δ1×I success -Δ2×I breach Where Δ1 represents the incremental value of successful cooperation, and Δ2 represents the reduction in value due to default; I success with I breach The indicator function takes a value of 1 when there is successful cooperation or a breach of contract, and 0 otherwise; the dynamic trust level is... ijConstrained within the interval [0,1]; Context similarity Sim ij : Measuring the V of drones i The degree of similarity between the current situation and bidding task T2 is used to determine the accuracy of the judgment. The formula is as follows: Sim ij =β×Sim cosine +(1-β)×Sim context ; Among them, Sim cosine Sim is a cosine similarity algorithm based on task feature vectors, considering type, environment, and urgency. context Similarity is based on contextual differences such as time and location; β is the weighting coefficient. Capability Confidence ij Drone V i For drone B j The ability to successfully complete task T2 is assessed using the following formula: Cap ij =ω1×Match ij +ω2×Hist ij +ω3×Status ij ; Among them, Match ij For the task requirements and B j The matching degree of the capability vector; Hist ij For B j Historical success rate and quality score for performing similar tasks; Status ij For B j Current real-time status (resource availability, health); ω1, ω2, ω3 are weighting coefficients and their sum is 1; This weighted voting formula integrates three dimensions—historical cooperation trust, current scenario similarity, and comprehensive capability assessment—to construct a multi-dimensional and quantifiable evaluation framework for each voter. Its advantage lies in transforming scattered local knowledge and individual experience into objective collective decision-making basis, thereby solving the key problem of how drone swarms can collaboratively make reliable, efficient, and fair task allocation decisions based on limited and heterogeneous local information in a distributed environment lacking central authority.

[0052] S43: V per drone i It allocates its voting rights to the bidders it supports; typically, drones V i Cast your vote for weight W ij The highest bidder; the emergency command vehicle system aggregates all weighted votes and calculates the bid proposal D. j Vote (D) j ): Let V = {V1, V2, ..., V} m} is the voter (the drone V that receives the consensus auction request) i )gather; B = {B1, B2, ..., B} n} is the bidder (Drone B participating in the bidding) j )gather;

[0053] Where m is the total number of elements in set V, and n is the total number of elements in set B; W ij For drones V i For drone B j The overall weight is calculated as follows: ， W ij =Trust ij ×Sim ij ×Cap ij ; I ij Let V be the indicator function, representing the unmanned aerial vehicle (UAV). i Should I vote for drone B? j :

[0054] By weighted voting aggregation and vote rate calculation, each voter's individual preferences (reflected in the overall weight of the bidders) are transformed into the group's quantitative support for each candidate solution. Its advantage lies in both respecting individual differences based on professional judgment (reflected through weight differentiation) and ensuring the objectivity and comparability of collective decision-making through normalized mathematical aggregation. This solves the key problem of how to efficiently and fairly integrate scattered and heterogeneous individual opinions into a collective choice that can reach a clear consensus in distributed collaboration.

[0055] S44: The emergency command vehicle system has a preset consensus threshold T c If a certain solution D exists... j Vote (D) j )>T c If this is achieved, it is considered a partial consensus, and the high-value task T2 and its dynamic commitment contract C(T2) are assigned to the corresponding winning drone A. j If consensus cannot be reached in the first round, a multi-round voting process will be initiated, by lowering the consensus threshold T. c To promote the formation of consensus on auctions; S45: After the auction consensus described in S44 is reached, the winning drone A... jThe system immediately acquires the right to execute the high-value task T2 and commits to accumulating a bonus B(t) starting from time 0 in the dynamic commitment contract C(T2); the contract for task T1 is marked as defaulted, and the corresponding penalty P1 is recorded. The result of the entire consensus auction is reported to the emergency command vehicle system.

[0056] S5: The emergency command vehicle system continuously monitors all task reassignment events generated by S3 and S4, calculates the smoothed system replanning frequency through an exponential smoothing filtering algorithm, and feeds this frequency back to the dynamic contract generation system of S1, thereby dynamically adjusting the penalty P for breach of contract in all future contracts, thus forming a closed-loop stability control mechanism.

[0057] Furthermore, the execution process and principle of the closed-loop stability control mechanism include the following steps: S51: The system monitoring and stability controller of the emergency command vehicle system listens for and records replanning events consisting of the following two types of events in real time: S3 Triggering Event: Consensus auction request initiated by any drone based on G1>θ·G2; S4 Completion Event: Any consensus is reached, resulting in the task being reassigned; Within a fixed time window Δt (preferably Δt is 1 minute), the number of replanning events that occur is counted and denoted as the instantaneous replanning count N. t ; S52: Calculate the instantaneous replanning frequency r based on the instantaneous count in S51. t :r t =N t / Δt The instantaneous replanning frequency r t It directly reflects the degree of system volatility within the most recent time window.

[0058] S53: To avoid excessive impact from instantaneous fluctuations on the system, an exponential smoothing filter algorithm is used to smooth the instantaneous frequency, and the system reprogramming frequency R is calculated. t The calculation formula is: R t =λ⋅R t−1 +(1−λ)⋅r t ; Among them, R t The system replanning frequency at time t (times / unit of time) is a core smoothing indicator reflecting the overall stability of the system. R t−1 The system replanning frequency of the previous moment carries historical state information; r t The instantaneous reprogramming frequency at the current moment is the latest observation value; λ is the forgetting factor, and its value ranges from 0 to 1.

[0059] S54: Calculate the latest system replanning frequency R t The dynamic contract generation system provides real-time feedback to step S1, and subsequently generates a dynamic commitment contract C(T)=(V) for any newly released or discovered task. b When B(t), the penalty P for breach of contract will be based on the smoothed R. t (instead of the original r) t Perform dynamic calculations: P = P0 × (1 + α⋅R) t ); Where P0 is the base penalty and α is the adjustment gain coefficient.

[0060] The above mechanism constitutes a complete negative feedback closed loop, realizing the system's adaptive stability control. By monitoring system replanning events in real time and using an exponential smoothing filter algorithm to obtain frequency indicators reflecting the long-term turbulence trend of the system, and using these as key parameters to dynamically adjust the penalties for breach of contract in all future task contracts, an adaptive negative feedback adjustment system is constructed. Its advantage lies in its ability to intelligently distinguish between short-term fluctuations and long-term trends in system turbulence, and to achieve self-correction of system behavior through the self-adjustment of economic rules. This effectively solves the fundamental contradiction between the individual flexibility requirements and the overall stability requirements in a distributed autonomous decision-making system.

[0061] When the emergency command vehicle system detects frequent and persistent replanning: the smoothing frequency R t The penalty P for breach of the new dynamic commitment contract increases steadily, and the penalty P for the drone abandoning the current mission during the S3 decision-making process increases significantly. i As the value increases, the expected net benefit of switching to a new task G1=V j -P i The reduction effectively curbed the incentive for drones to arbitrarily default on contracts and initiate auctions, and the system's replanning behavior gradually decreased, returning to stability.

[0062] When the system is running stably: smoothing frequency R t By maintaining a low level, the penalty P for breach of the new dynamic commitment contract is reduced accordingly, the cost of switching drone missions is reduced, and the system maintains its ability to respond quickly and flexibly to sudden high-value missions and local environmental changes.

[0063] Through closed-loop stability control in step S5, the method innovatively employs an exponential smoothing filter algorithm to intelligently distinguish between the "trend" and "noise" of system fluctuations, thereby dynamically adjusting the game rules (penalties in the contract). This enables multi-UAV systems to autonomously and robustly maintain overall global stability without sacrificing local response agility, significantly improving the long-term robustness and overall mission performance of large-scale distributed autonomous systems in complex dynamic environments.

[0064] The present invention has been described in the above-described embodiments; however, these embodiments are merely examples for implementing the present invention. It must be noted that the disclosed embodiments do not limit the scope of the present invention. Conversely, any modifications and refinements made without departing from the spirit and scope of the present invention are within the scope of patent protection of the present invention.

Claims

1. A multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions, characterized in that: S1: When the emergency command vehicle system mounted on the vehicle issues task T1, the dynamic contract generation system in the emergency command vehicle system generates a dynamic commitment contract C(T1); S2: The emergency command vehicle system is based on the capability vectors A of each UAV. i Calculate the matching degree M(A,T) between the UAV and the dynamic commitment contract C(T1), perform initial task allocation, form an initial allocation set {(A,C(T1))}, and achieve preliminary optimization of resource allocation; S3: When drone A executes task T1 assigned by S2, when drone A locally senses a new high-value task T2, drone A calculates the expected net profit G1=V1-P1 of abandoning the current task T1 and bidding for the high-value task T2, and the expected net profit of sticking to the current task T1: G2=V2+B2; drone A initiates a consensus auction request for the high-value task T2 to the drones in the local neighborhood if and only if G1>θ·G2. Where θ is the preset switching threshold, V b 1 represents the base value of task T1, P1 represents the penalty for breach of contract in task T1, and V represents the base value of task T1. b 2 represents the base value of task T2, and B(t1) represents the accumulated bonus of task T1. S4: The drones that receive the consensus auction request collaboratively determine the best allocation of the high-value task T2 through a weighted voting mechanism, forming a local consensus and completing the reallocation; S5: The emergency command vehicle system continuously monitors all task reassignment events generated by S3 and S4, calculates the smoothed system replanning frequency through an exponential smoothing filtering algorithm, and feeds this frequency back to the dynamic contract generation system of S1, thereby dynamically adjusting the penalty P for breach of contract in all future contracts, thus forming a closed-loop stability control mechanism.

2. As described in claim 1, characterized in that: The dynamic contract generation system generates contracts based on the attributes of task T1, including target location distance, type BaseValue (type), resource consumption cost, and deadline t. deadline ; The target location distance is the geometric distance from a fixed "reference point" preset by the emergency command vehicle system to mission T; the type includes strategic reconnaissance, tactical reconnaissance, precision strike, and material transportation; the resource consumption cost is the amount of resources required to execute mission T; the deadline t deadline To meet the deadline for completing the task T, the resources include electricity or fuel; The dynamic commitment contract C(T1) includes the basic task value V. b The promised bonus B(t) and the penalty for breach of contract P.

3. The multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions as described in claim 2, characterized in that: The dynamic commitment contract C(T1) is expressed as: C(T) = (V b ,B(t),P), where: The basic value V of the task b Calculated from the inherent attributes of the task; Calculation method: ； in: Weight 1+2+3+ 4 = 1; λ is the geographical proximity gain coefficient; η is the time urgency gain coefficient; γ is the resource consumption penalty coefficient; ϵ is a constant; w i These are the weighting coefficients; λ, η, γ, ϵ and w i All settings are manually configured during system initialization. distance is the target location, cost is the resource consumption, and t is the target location. deadline The deadline; The promised bonus B(t) is expressed as: B(t) = B0 × (1 - e^(-t / τ)) is the commitment bonus that accumulates over time, accumulating with the time t for which the task is committed to be executed. It is used to incentivize drones to faithfully execute assigned tasks in the long term. The reward for drones that continuously execute tasks is increased exponentially. Here, B0 is the bonus cap, t is the commitment time, and τ is the time constant. P = P0 × (1 + α·R) is the penalty for breach of contract based on the system replanning frequency R, and its value increases with the system's historical replanning frequency R. Where P0 is the base penalty; α is the adjustment gain coefficient; and R is the reprogramming frequency.

4. The multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions as described in claim 1, characterized in that: The matching degree M(A,T) is calculated using the formula: M(A,T) = a·M type +b·M distance +c·M resource +d·M time ; Where a+b+c+d=1, and a, b, c, and d are weighting coefficients; M type For task type matching degree, if the task type capability Cap type If the task type is included, then M type =1; otherwise, M type =0; M distance M represents the distance matching degree. distance =1 / (1+d(A,T) / D_scale), where d(A,T) is the actual distance from the current position of the UAV to the position of mission T; D_scale = distance scaling constant; M resource For resource matching degree, M resource =min(1,E remain / E required ), where E required To estimate the total energy consumption required to perform the task, E remain This refers to the current remaining fuel / electricity of the drone. M time For time matching degree, if (d(A,T) / v max +t exec )<(t deadline -t current ), then M time =1; otherwise, M time =0; where d(A,T) is the distance from the current position of the UAV to the position of the target T; v max t is the maximum cruising speed of the drone; exec The time required to complete the task itself; t deadline -t current This represents the remaining available time.

5. The multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions as described in claim 1, characterized in that: The execution process of the weighted voting mechanism includes the following steps: S41: Drone V that received a consensus auction request within a local neighborhood i As a potential bidder, based on the contract C(T2) of the new task T2 and its current state, it conducts a local bidding value assessment, generates a quantitative assessment value G3, and decides whether to participate in the bidding based on a preset threshold E. Furthermore, the local bidding value assessment is for the drone V. i Calculate the expected net gain G of abandoning the existing task (if any) and instead executing task T2. 3i The calculation formula is as follows: G 3i =(V b 2-P i )-θ(V b 3+B(t3); Among them, V b 2 represents the basic task value of T2; P i For drones V i The penalty for abandoning the existing mission, if the drone V i If P is in a no-task state, then i =0; θ is the preset switching threshold, which is the same as the threshold used when the drone initiates the auction request in S3; V b 3 is the drone V i The basic mission value of performing existing tasks, if the drone V i If V is in a no-task state, then b 3 is 0; B(t3) is the value of the drone V. i The accumulated bonuses from performing existing tasks, if drone V i If the task is not active, then B(t3) is 0; Furthermore, if G 3i >E, then the drone V i Marked as qualified bidder B j And generate an evaluation report; If G 3i If V ≤ E, then the drone V i Bidder C marked as unqualified j They will not participate in the bidding. The evaluation report includes the drone's unique ID and drone capability vector A. i Estimated completion time t j G 3i ; Estimated completion time t j Based on its current location, speed, and task flow, the estimated time required to complete task T2; S42: Unmanned Aerial Vehicle (UAV) i For drone B j Calculate the overall voting weight W ij The comprehensive voting weight W ij =Trust ij ×Sim ij ×Cap ij ,in, Dynamic Trust ij Based on drones V i With drone B j The historical cooperation results are dynamically updated, and the update rules are as follows: Trust ij(t+1) =Trust ij(t) +Δ1×I success -Δ2×I breach Where Δ1 represents the incremental value of successful cooperation, and Δ2 represents the incremental value of default; I success with I breach The indicator function takes a value of 1 when there is successful cooperation or a breach of contract, and 0 otherwise; the dynamic trust level is... ij Constrained within the interval [0,1]; Context similarity Sim ij : Measuring the V of drones i The degree of similarity between the current situation and bidding task T2 is used to determine the accuracy of the judgment. The formula is as follows: Yes ij =β×Sim cosine +(1-β)×Sim context ; Among them, Sim cosine Sim is a cosine similarity algorithm based on task feature vectors, considering type, environment, and urgency. context Similarity is based on contextual differences such as time and location; β is the weighting coefficient. Capability Confidence ij Drone V i For drone B j The ability to successfully complete task T2 is assessed using the following formula: Cap ij =ω1×Match ij +ω2×Hist ij +ω3×Status ij ; Among them, Match ij For the task requirements and B j The matching degree of the capability vector; Hist ij For B j Historical success rate and quality score for performing similar tasks; Status ij For B j Current real-time status (resource availability, health); ω1, ω2, ω3 are weighting coefficients and their sum is 1; S43: V per drone i It allocates its voting rights to the bidders it supports; typically, drones V i Cast your vote for weight W ij The highest bidder; the emergency command vehicle system aggregates all weighted votes and calculates the bid proposal D. j Vote (D) j ): Let V = {V1, V2, ..., V} m } for voter V i gather; B = {B1, B2, ..., B} n For bidder B j gather; ； Where m is the total number of elements in set V, and n is the total number of elements in set B; W ij For drones V i For drone B j The overall weight is calculated as follows: ， W ij =Trust ij ×Sim ij ×Cap ij ; I ij Let V be the indicator function, representing the unmanned aerial vehicle (UAV). i Should I vote for drone B? j : 6.S44: The emergency command vehicle system has a preset consensus threshold T. c If a certain solution D exists... j Vote (D) j )>T c If this is achieved, it is considered a partial consensus, and the high-value task T2 and its dynamic commitment contract C(T2) are assigned to the corresponding winning drone A. j If consensus cannot be reached in the first round, a multi-round voting process will be initiated, by lowering the consensus threshold T. c To promote the formation of consensus on auctions; S45: After the auction consensus described in S44 is reached, the winning drone A... j The system immediately acquires the right to execute the high-value task T2 and commits to accumulating a bonus B(t) starting from time 0 in the dynamic commitment contract C(T2); the contract for task T1 is marked as defaulted, and the corresponding penalty P1 is recorded. The result of the entire consensus auction is reported to the emergency command vehicle system.

7. The multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions as described in claim 1, characterized in that: Furthermore, the execution process and principle of the closed-loop stability control mechanism include the following steps: S51: The system monitoring and stability controller of the emergency command vehicle system listens for and records replanning events consisting of the following two types of events in real time: S3 Triggering Event: Consensus auction request initiated by any drone based on G1>θ·G2; S4 Completion Event: Any consensus is reached, resulting in the task being reassigned; Within a fixed time window Δt (ideally 1 minute), the number of replanning events that occur is counted and denoted as the instantaneous replanning count N. t ; S52: Calculate the instantaneous replanning frequency r based on the instantaneous count in S51. t :r t =N t / Δt The instantaneous replanning frequency r t It directly reflects the degree of system volatility within the most recent time window; S53: To avoid excessive impact from instantaneous fluctuations on the system, an exponential smoothing filter algorithm is used to smooth the instantaneous frequency, and the system reprogramming frequency R is calculated. t； S54: Calculate the latest system replanning frequency R t The dynamic contract generation system provides real-time feedback to step S1, and subsequently generates a dynamic commitment contract C(T)=(V) for any newly released or discovered task. b When B(t), the penalty P will be based on the smoothed R. t Perform dynamic calculations: P=P0×(1+α⋅R t ); Where P0 is the base penalty and α is the adjustment gain coefficient; The above mechanism constitutes a complete negative feedback closed loop, realizing the adaptive stability control of the system.

8. The multi-UAV collaborative task planning method based on dynamic contracts and consensus auctions as described in claim 6, characterized in that: The system replanning frequency R t The calculation formula is: R t =λ⋅R t−1 +(1−λ)⋅r t ; Among them, R t The system replanning frequency at time t is expressed in times per unit time and is a core smoothing indicator reflecting the overall stability of the system. R t−1 The system replanning frequency of the previous moment carries historical state information; r t The instantaneous reprogramming frequency at the current moment is the latest observation value; λ is the forgetting factor, and its value ranges from 0 to 1.

Images