Dynamic task-oriented heterogeneous unmanned aerial vehicle cooperative scheduling and path replanning method

Abstract

This invention belongs to the field of UAV trajectory planning technology, and particularly relates to a heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks. The method includes: S1, generating an initial global task allocation scheme; S2, using a rolling time-domain control module to control the UAVs to fly according to their corresponding trajectories, when a new dynamic task arrives: S3, calling a dynamic priority decay compensation model to evaluate the dynamic priority of the dynamic task to be allocated; calculating the value density of the dynamic task to be allocated by the UAV; using the value density as the bidding auction for the dynamic task to be allocated by the UAVs, obtaining the current global task allocation scheme; if the replanning trigger condition is met, proceeding to S4; S4, using the current global task allocation scheme as the elite individuals of a genetic algorithm to obtain the optimal global task allocation scheme, distributing it to the corresponding UAVs, and returning to S2. This eliminates local task deadlock and starvation, maximizing the overall execution benefit in a dynamic environment.

Description

Technical Field

[0001] This invention belongs to the field of UAV trajectory planning technology, and particularly relates to a method for collaborative scheduling and trajectory replanning of heterogeneous UAVs for dynamic tasks. Background Technology

[0002] In practical application scenarios such as disaster relief, wide-area detection and complex environment monitoring, the tasks faced by drone swarms are usually highly dynamic and time-series uncertain, and the potential benefits of the tasks show a significant decay characteristic as the waiting time increases.

[0003] Related technologies provide global optimization task allocation methods based on heuristic algorithms, but these methods mainly target statically known global targets and have the following key limitations in dynamic, sudden task scenarios: (1) Global replanning leads to insufficient real-time response capability: When a new task arrives, the complete offline genetic algorithm or particle swarm optimization algorithm needs to be re-executed. The global calculation time usually takes several seconds to tens of seconds. During this period, sudden tasks cannot be responded to in time, and high-value execution windows are easily missed.

[0004] (2) Lack of time-varying incentive mechanism leads to local optima and task starvation: In order to avoid the computational overhead of global replanning, some schemes adopt a simple greedy strategy to allocate new tasks. However, such methods do not consider the time-varying decay characteristics of task benefits and are prone to getting trapped in local optima. At the same time, due to the lack of compensation mechanism, low-value tasks with long waiting times cannot be assigned for a long time, resulting in task starvation.

[0005] (3) The allocation mechanism is lagging in adapting to the dynamic evolution of the system state: Under dynamic conditions, the constraints such as the remaining energy of the UAV and the environmental threat field change nonlinearly, and the offline solution cannot adapt in time, resulting in the failure of the execution layer constraints when the command is issued. Summary of the Invention

[0006] This application aims to address the technical problem that related UAV swarm mission planning algorithms struggle to balance real-time response to sudden dynamic tasks with optimal global system benefits, and provides a heterogeneous UAV swarm scheduling and replanning method for dynamic mission scenarios.

[0007] Firstly, this application provides a method for collaborative scheduling and trajectory replanning of heterogeneous UAVs for dynamic tasks. The method includes: Step S1, generating an initial global task allocation scheme based on the input initial static task set, environmental threat field, and the initial position, energy state, and safe return base point set of one or more heterogeneous UAVs. The global task allocation scheme includes the sequence of tasks to be executed and the corresponding trajectory for each UAV; Step S2, using a rolling time-domain control module to control the UAV to fly according to its corresponding trajectory, while monitoring whether new dynamic tasks arrive. When a new dynamic task arrives, the following steps are executed synchronously: Step S3, calling dynamic priority decay compensation. The compensation model evaluates the dynamic priority of the dynamic tasks to be assigned at the current moment; it calculates the value density of the dynamic tasks to be assigned by the UAV by combining the UAV state vector and the dynamic priority of the dynamic tasks to be assigned; it uses the value density as the bidding auction for the dynamic tasks to be assigned by the UAV, and obtains the current global task allocation scheme after all dynamic tasks to be assigned are completed; if the replanning trigger condition is met, proceed to step S4; in step S4, the current global task allocation scheme is used as the elite individuals of the genetic algorithm, the genetic algorithm is executed to obtain the optimal global task allocation scheme, the sequence of tasks to be executed in the optimal global task allocation scheme is sent to the corresponding UAV, and the process returns to step S2.

[0008] Secondly, this application provides a system based on a heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks. The system includes a processing center and multiple heterogeneous UAVs that communicate wirelessly with the processing center.

[0009] The beneficial technical effects of this application are as follows: The rolling time-domain control module controls the UAV to fly according to its corresponding trajectory while monitoring whether new dynamic tasks arrive. When a new dynamic task arrives, steps S3 and S4 are executed simultaneously to achieve local addition of dynamic tasks. When the replanning trigger condition is met, a genetic algorithm is executed to obtain the optimal global task allocation scheme, updating the UAV's task sequence to be executed, which greatly improves the cluster task completion efficiency and achieves background hot-start global optimization. Compared with the traditional static heuristic algorithm, the total benefit of this method increases by more than 65.3%, and the task completion rate is close to 100.0%, achieving a significant improvement in the overall system benefit with extremely low computational overhead. When a new dynamic task arrives, step S3 calls the dynamic priority decay compensation model to evaluate the dynamic task to be allocated. The system dynamically prioritizes tasks at the current moment, eliminating task starvation and deadlock in high-load scheduling and significantly reducing resource idle rate. It calculates the value density of dynamic tasks to be assigned by drones by combining drone state vectors and dynamic priorities of dynamic tasks to be assigned. Using value density as the bidding for drones enables heterogeneous perception auctions, significantly improving resource utilization efficiency and fully utilizing the differentiated characteristics of heterogeneous drones. By combining real-time auctions to quickly issue instructions with background hot-start global optimization, it meets the needs for response windows in rescue or time-sensitive scenarios, enabling rapid response to new and sudden tasks in milliseconds. At the same time, it achieves global overall planning and correction through cross-stage dual-trigger mechanism, effectively eliminating task local deadlock and starvation, and maximizing the total execution benefits of the cluster in dynamic environments. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a preferred embodiment of the heterogeneous UAV collaborative scheduling and trajectory replanning method for dynamic tasks according to the present invention. Figure 2 This is an example of the dynamic priority decay compensation model in this invention, showing the dynamic priority curves under different value decay rates. Figure 3 This is an example of the dynamic priority decay compensation model in this invention, showing the dynamic priority change rate under different value decay rates; Figure 4 This is a comparison chart of the task completion rates of the heterogeneous UAV cooperative scheduling and trajectory replanning method (MCHS-Dynamic) of the present invention and four single static programming and greedy rule benchmark algorithms. Figure 5 This is a comparison chart of the energy consumption of the heterogeneous UAV cooperative scheduling and trajectory replanning method (MCHS-Dynamic) for dynamic tasks of the present invention with four single static planning and greedy rule benchmark algorithms; Figure 6This is a comparison chart of the total movement distance between the heterogeneous UAV cooperative scheduling and trajectory replanning method (MCHS-Dynamic) of the present invention and four single static planning and greedy rule benchmark algorithms; Figure 7 This is a comparison chart of the total benefits of the heterogeneous UAV cooperative scheduling and trajectory replanning method (MCHS-Dynamic) for dynamic tasks of the present invention with four single static programming and greedy rule benchmark algorithms; Figure 8 A system block diagram of a preferred embodiment of the present invention. Detailed Implementation

[0011] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0012] The execution entity of the heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks disclosed in this invention includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in the embodiments of this application: a server, a terminal, or a computer. In other words, the heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0013] Consider a A cluster of heterogeneous drones It executes a mixed set of tasks, including initial static tasks and randomly emerging dynamic tasks. The set of dynamic tasks is denoted as:

[0014] This indicates the total number of dynamic tasks, and each dynamic task... ( It has the following time-sensitive attribute vectors:

[0015] The meanings of each symbol are as follows: Dynamic Tasks (Also known as dynamic task) The three-dimensional spatial coordinates of the mission objective point; Dynamic Tasks Task baseline reference value (normalized reference value, used to calculate dynamic priority; actual initial priority at task arrival time) , rather than itself), Indicates the first weight; Dynamic Tasks Time of appearance (arrival); Dynamic Tasks Required payload type identifier (mission type), such as reconnaissance mission (corresponding reconnaissance payload), strike mission (corresponding strike payload). Dynamic Tasks Value decay rate ( ); Dynamic Tasks Compensation recovery rate ( ).

[0016] The system global time is recorded as This can also be called the current time or the current moment, and the dynamic tasks to be assigned. The waiting time is defined as:

[0017] (when (At this time, the task is considered not yet completed, and the dynamic priority is set to zero).

[0018] UAV state vector definition: The drones ( At any moment The state vector is defined as:

[0019] The meanings of each symbol are as follows: Drones At any moment 3D position coordinate vector ; Drones At any moment Three-dimensional velocity vector ; Drones At any moment Current remaining energy (unit: joules J); Drones Maximum battery capacity (unit: joules J); Drones The energy consumption coefficient per unit distance (unit: J / m) characterizes the energy consumed by the UAV per unit distance it flies; Drones Professional type identifier (e.g., professional reconnaissance type, professional strike type); Drones The set of all load capacities (satisfying) ).

[0020] The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks provided by this invention, in a preferred embodiment, is described below. Figure 1 The method includes: Step S1: Based on the input initial static task set, environmental threat field, and the initial position, energy state, and safe return base point set of one or more heterogeneous UAVs, generate an initial global task allocation scheme and distribute it to the UAVs. The global task allocation scheme includes the sequence of tasks to be executed and the corresponding flight path for each UAV.

[0021] In this embodiment, the system first acquires and integrates the following input information: an initial static task set, an environmental threat field, the initial positions of one or more heterogeneous UAVs, the current energy state of the UAVs, and a set of safe return base points. The current energy state of the UAVs may include the current remaining energy in the UAV state vector, the energy consumption coefficient per unit distance, etc.

[0022] Among them, the environmental threat field is used to mark dangerous areas such as no-fly zones and radar detection zones; the safe return base point is the location where the UAV can safely return when it is low on energy or when the mission is terminated; the initial position and energy state of the heterogeneous UAV determine the physical range and endurance of the mission that the UAV can perform.

[0023] After obtaining all input, At all times, the system combines the kinematic limits and payload capacity of each UAV with the optimization objectives of mission feasibility, flight safety, and reasonable energy consumption, and uses constraint-aware adaptive optimization algorithms or genetic algorithms for offline global calculations. The algorithm comprehensively considers: whether the UAV can safely reach the mission point from its initial position; whether the flight path avoids dangerous areas in the environmental threat field; whether the energy required to execute the mission does not exceed the current remaining energy, and reserves energy for return; and whether the task is evenly distributed to avoid overloading any single UAV.

[0024] Through the aforementioned global optimization calculations, the system assigns a unique sequence of tasks to be executed to each UAV, and generates corresponding continuous flight paths according to the task execution order, ultimately forming an initial global task allocation scheme. It is used to record the sequence of tasks to be performed for each UAV, as well as the safe and feasible flight path (also called the baseline flight path) that corresponds one-to-one with the sequence of tasks to be performed. Step S2: Use the rolling time domain control module to control the UAV to fly according to its corresponding trajectory, while monitoring whether any new dynamic tasks arrive.

[0025] In this embodiment, the Receding Horizon Control (RHC) module can be deployed on the UAV's onboard computer. Each UAV is independently configured with one RHC module, deployed in a distributed manner, for independently controlling its own trajectory tracking and real-time obstacle avoidance. After obtaining the initial global task allocation scheme, the corresponding task sequence and trajectory for each UAV are sent to that UAV. The UAV's RHC module then begins to run independently, performing path tracking and real-time obstacle avoidance according to the allocated access sequence, such as controlling the UAV to fly along the trajectory, avoiding obstacles in real time, and maintaining flight stability. The processing center is only responsible for task allocation, dynamic priority calculation, auction, and global replanning, and does not directly control the flight actions of each UAV. When a new task is added to the end of the UAV's task sequence, the UAV's RHC continues to fly uninterruptedly without interruption or waiting for global recalculation, ensuring millisecond-level response.

[0026] In this embodiment, during RHC execution, the system continuously monitors for the arrival of new dynamic tasks. If no new dynamic task is detected, RHC rolling tracking is maintained; if a new task is detected, the system switches to dynamic response mode (i.e., steps S3 and S4) and adjusts the current baseline allocation scheme. The track is retained as the background reference for the system's global state (for subsequent steps of physical ranging and queue continuation), and dynamic priority evaluation is triggered simultaneously.

[0027] Therefore, when a new dynamic task arrives, the following steps are executed synchronously: Step S3: Call the dynamic priority decay compensation model to evaluate the dynamic priority of the dynamic task to be assigned at the current moment; combine the UAV state vector and the dynamic priority of the dynamic task to be assigned to calculate the value density of the dynamic task to be assigned by the UAV; use the value density as the bidding auction for the dynamic task to be assigned by the UAV; after all dynamic tasks to be assigned are assigned, obtain the current global task assignment scheme; if the replanning trigger condition is met, proceed to step S4. Step S4: Using the current global task allocation scheme as the elite individual of the genetic algorithm, execute the genetic algorithm to obtain the optimal global task allocation scheme, send the sequence of tasks to be executed in the optimal global task allocation scheme to the corresponding UAV, and return to execute step S2.

[0028] In a preferred embodiment, when the system detects the arrival of a new sudden task (i.e., when a dynamic task is detected), it immediately calculates the current instantaneous dynamic priority of that task and all pending tasks in the system that have not yet been executed. In step S3, the dynamic priority decay compensation model is invoked to evaluate the dynamic priority of the dynamic task to be assigned at the current moment, including: Step S3a1: Obtain the time-sensitive attribute vector of the dynamic task to be assigned at the current moment. The time-sensitive attribute vector includes the value decay rate, compensation recovery rate, waiting time, three-dimensional spatial coordinates of the task target point, and task benchmark reference value used to calculate dynamic priority. Step S3a2: Substitute the time-sensitive attribute vector of the dynamic task to be assigned at the current moment into the dynamic priority decay compensation model to obtain the dynamic priority of the dynamic task to be assigned at the current moment. The dynamic priority attenuation compensation model is expressed as follows: ; Indicates dynamic tasks to be assigned At the present moment Dynamic priority; Indicates dynamic tasks to be assigned The task benchmark reference value; Represents an exponential function; Indicates dynamic tasks to be assigned The rate of value decay; Indicates the current time Dynamic tasks to be assigned Waiting time; Indicates the first weight; Indicates the second weight; Dynamic tasks to be assigned The compensation and recovery rate.

[0029] Control the attenuation term separately With compensation items The weighting percentage. Dynamic task arrival time ( The initial priority is ;when When the priority asymptotically approaches .thus, As a reference for normalization Determines the proportion of initial effective perceived value. This serves as an upper limit reserve for famine relief. The exponential decay of time-dependent factors characterizes the value of a task; The item represents the asymptotic compensation and rebound of priority after a long waiting time.

[0030] According to the above model, when When the value is small, the decay term dominates, and its priority drops rapidly; when... When the value is large, the compensation term gradually becomes dominant, preventing tasks from waiting indefinitely without being scheduled. The parameters of this model... It can be preset or adjusted online according to the task type.

[0031] In this embodiment, preferably, the dynamic task to be assigned is determined. The switching wait time when the dynamic priority changes from decreasing to increasing for: .

[0032] Specifically, to determine the moment when the dominance of the attenuation term and the compensation term switches, for about Differentiate and set it equal to zero:

[0033] Switching time after sorting (That is, the moment when the priority changes from decreasing to increasing):

[0034] The meanings of each symbol are as follows: :Task The moment when the priority change rate is zero (in seconds), before which the decay term dominates, and after which the compensation term dominates; :Task Value decay coefficient ( Typical values ); :Task Value compensation coefficient ( Typical values ); Switching time calculations need to be performed independently for each task, and different tasks may vary. , , , The different switching times result from the differences in the two.

[0035] when At that time, it is necessary to meet the following requirements. The switching time is a finite positive value; when hour, and If the same sign (both negative) is used, the value will still be a limited positive value when switching, indicating that the anti-hunger compensation state is entered at an earlier stage. when At that time, if (Typical situation) Always negative, with monotonically decreasing priority, and no finite switching time; only when... In the case of degradation, the priority is constant, and the switching time is meaningless.

[0036] Under typical parameter configuration ( , , , The switching time is approximately:

[0037] By fusing the calculated attenuation and compensation terms, a composite dynamic priority for each task is obtained. At the same time, according to With switching time The relative relationship is used to determine whether the current stage is dominated by the decay term (priority drops rapidly) or the compensation term (priority rises asymptotically), and the final priority evaluation result is output.

[0038] In a preferred embodiment, the system maintains a short-term dynamic task buffer and a recent task arrival time window. (Typical value: 1.5–3 seconds). When a new dynamic task arrives and triggers S3: Count the number of dynamic tasks waiting to be assigned in the current buffer. .

[0039] like ( If the batch threshold is set (typically 3-5), then the batch enters the Batch-HAM-Auction (batch heterogeneous awareness multi-attribute auction) mode: the buffer... A batch set of tasks And according to the current dynamic priority of each task. Sort the data from highest to lowest to obtain an ordered batch sequence; otherwise, enter single-task HAM-Auction mode: batch set. It only includes the current single task. Then, it sorts the batch set. Each task in the process executes the subsequent single-task auction process sequentially (in batch mode, the status of the corresponding drone is updated in real time after each task allocation is completed, so that subsequent tasks in the same batch can use the latest queue and energy information). This switching mechanism ensures millisecond-level single-machine closed-loop response in low-density scenarios, and enables batch-based allocation of high-value tasks to prioritize the best resources in high-density task cluster scenarios, with negligible switching judgment overhead.

[0040] For the currently pending tasks (In batch mode, the current task is sorted) The system immediately initiates a distributed auction. Each drone ( Conduct independent feasibility self-assessment, including: Payload matching verification: Unmanned aerial vehicle (UAV) capability set Does it include the payload type required for this task? (Right now ).

[0041] Remaining energy safety constraint check (energy consumption included in the backlog of tasks): The meanings of each symbol are as follows: Drones At the present moment Instantaneous residual energy (unit: J); Drones The estimated total energy required to complete the currently backlogged task queue; Drones Fly from the end waypoint of its mission queue to the new mission point. The estimated additional energy required to eventually return to the nearest safe base point (in J); Energy safety margin factor (typically 0.15).

[0042] The time-sensitive attribute vector of the dynamic task to be assigned at the current moment also includes the payload type identifier required by the dynamic task, and the UAV state vector includes the UAV's specialty type identifier. and the set of load-bearing capabilities .

[0043] In step S3, the value density of the dynamic tasks to be assigned is calculated only for UAVs that meet the remaining energy safety constraint check. This value density is calculated by combining the UAV's state vector and the dynamic priority of the dynamic tasks to be assigned, including: Step S3b1: Based on the payload type identifier required for the dynamic task to be assigned and the UAV's specialty type identifier... and the set of load-bearing capabilities Obtain the matching degree between the dynamic tasks to be assigned and the capabilities of the drone. .

[0044] Specifically, dynamic tasks to be assigned With drones Ability matching degree This is used to distinguish between what can be performed and what is good at performing, and its quantitative definition is as follows: It should be noted that specialized drones receive a fixed 1.5x weighted advantage; drones with only general execution capabilities have their weighting coefficient varying depending on the dimension of their capability set. The increasing and decreasing values ​​prioritize task allocation to the most suitable specialized drones, suppressing inefficient bidding. The coefficient value of 1.5 and the decreasing function form mentioned above are empirical settings and can be adjusted according to specific application scenarios.

[0045] Step S3b2: Calculate the estimated additional energy consumption required for the UAV to fly from the end waypoint of its pending task sequence to the task target point of the dynamic task to be assigned and finally return to the nearest safe return base point. Used to define drones From the end of its pending task sequence to the new task The total incremental energy consumption for returning from the current location (in J).

[0046] Step S3b3: Calculate the load balancing factor of the UAV based on the length of the task sequence being executed by the UAV at the current moment. Indicates drone The load balancing factor , This represents the load penalty weighting coefficient. The typical value is This is used to control the degree to which task backlog inhibits the willingness to bid. Indicates drone The current length of the queue of tasks to be executed.

[0047] Step S3b4: Match the dynamic tasks to be assigned with the capabilities of the UAV. Estimated additional energy consumption of drones The length of the task sequence currently being executed by the UAV and the dynamic priority of the dynamic tasks to be assigned are substituted into the composite value density function to obtain the value density of the dynamic tasks to be assigned by the UAV.

[0048] The composite value density function is: ; in, Indicates dynamic tasks to be assigned Undertake drones Value density; Indicates dynamic tasks to be assigned At the present moment Dynamic priority; Indicates dynamic tasks to be assigned With drones Compatibility; Indicates drone From the end waypoint of its pending task sequence, fly to the task objective point of the dynamic task to be assigned. And the estimated additional energy required to eventually return to the nearest safe return base point; Indicates drone The load balancing factor , This represents the load penalty weighting coefficient. The typical value is This is used to control the degree to which task backlog inhibits bidding willingness. The composite value density function fully utilizes the differences in characteristics of heterogeneous UAVs, and its essence is the unit incremental energy value gain after capability matching adjustment. The higher the value density, the greater the bidding advantage.

[0049] In a preferred embodiment, step S3, which uses value density as the basis for allocating dynamic tasks to be assigned in a bidding process for drones, includes: among drones that meet the remaining energy safety constraint test, selecting the drone with the highest value density as the winning drone, assigning the dynamic tasks to the winning drone, and appending the dynamic tasks to the end of the winning drone's task sequence. The expression for the remaining energy safety constraint test is: ; in, Indicates drone At any moment The current remaining energy; Indicates drone The total estimated energy required to complete its sequence of tasks; Indicates drone Maximum battery capacity; This represents the energy safety margin coefficient.

[0050] In this implementation, the bidding results of various drones are aggregated across the entire network, and the feasible drone with the highest value density is selected. :

[0051] The system selects drones As the winning unit , will the task Add to At the end of the sequence of tasks to be executed, an execution command is immediately issued to maintain short-term collision avoidance flight of the RHC.

[0052] If in Batch-HAM-Auction mode, immediately update the queue length of the winning drone. Energy state and global task queue And return. Continue the auction for the next task in the batch sequence; until the batch set is complete. Once all tasks have been processed, clear the buffer and output the updated global task queue. (The current global task allocation scheme) and immediately trigger a global replanning check. If it is a single-task mode, directly complete this step and output the updated global task queue. (Current global task allocation scheme).

[0053] In a preferred embodiment, in step S4, the replanning trigger condition is that the number of times local tasks are appended (equivalent to the number of dynamically assigned tasks to be assigned) reaches a set threshold. (e.g., 3 times) trigger a global replanning, or, at a fixed time step. (e.g., 10 steps) triggers a global replanning. Local task additions are event-triggered, while fixed-time-step triggers are periodic. Dual trigger conditions (event and periodic) can be used for monitoring, balancing execution responsiveness and global optimization. Event triggers are threshold-based, while periodic triggers are a fallback, forcibly triggering a global optimization.

[0054] In a preferred embodiment, in step S4, the current global task allocation scheme is used. As elite individuals in a genetic algorithm (elite chromosomes) The genetic algorithm is used to obtain the optimal global task allocation scheme, including: Step S41: Construct an initial population. The initial population includes elite individuals and one or more chromosomes. Chromosomes are randomly initialized, and each chromosome represents a global task allocation scheme.

[0055] First, the current global task allocation scheme formed through auction in step S3 is encoded and transformed into an elite chromosome. Each chromosome corresponds to a complete global task allocation scheme, and each segment of the code on the chromosome represents the sequence of tasks to be executed and the flight path access order for a particular UAV.

[0056] Subsequently, the system constructs the initial population for the genetic algorithm: the elite chromosomes mentioned above are placed at the beginning of the initial population as a high-quality starting point for evolution; the other individuals in the population, excluding the elite chromosomes, are generated randomly to form a complete initial population.

[0057] Specifically, elite chromosomes Insert initial population matrix The first line: in, Elite chromosome row vectors represent the current task allocation scheme. The encoding is a one-dimensional decision sequence that can be processed by a genetic algorithm. : Encoding function that maps a multi-UAV mission queue to a chromosome sequence. : Initial population matrix for genetic algorithm, where For population size, This represents chromosome length. The first row of the initial population matrix will... This ensures that the currently feasible solution is preserved as an elite individual throughout the entire iteration process.

[0058] Step S42: After performing a preset number of iterations of optimization based on the initial population, the optimal global task allocation scheme is obtained.

[0059] After constructing the initial population containing elite individuals and random initial individuals, proceed to step S42: perform low-iteration (20-80 generations) genetic algorithm optimization based on the fitness function to obtain the optimal global task allocation scheme.

[0060] First, calculate the fitness value for each chromosome in the population (i.e., each global task allocation scheme).

[0061] In this embodiment, the fitness function is used to evaluate the merits of the task allocation scheme, with the core objective of maximizing the overall global benefit of the system, while simultaneously satisfying the requirements of UAV energy security constraints, environmental threat field constraints, payload matching constraints, and load balancing.

[0062] Specifically, the fitness function uses the sum of the dynamic priority gains of all tasks as the positive reward term, and the higher the reward, the greater the fitness. At the same time, penalty terms are set for schemes that violate the constraints, including: insufficient remaining energy of the UAV to complete the task, flight path entering the environmental threat zone, mismatch between task payload and UAV capability, excessive imbalance of single-unit task load, etc. If any of the above situations exist, the fitness value will be reduced accordingly.

[0063] Subsequently, using the fitness function as the optimization objective, standard genetic operations of selection, crossover, and mutation are performed on the initial population.

[0064] Since the initial population already contains elite feasible solutions derived from the auction results, the algorithm can converge quickly with only a few iterations (20 to 80 generations), without the need for a cold start global search from scratch. The computational cost is only 3% to 10% of that of traditional genetic algorithms, which greatly reduces computational latency while ensuring global optimum.

[0065] After the iteration terminates, the optimal chromosome with the highest fitness value is selected and decoded into an executable optimal global task allocation scheme. The optimal solution is then distributed to each UAV to update the task queue and flight path, and is used as a new system benchmark for rapid response and follow-up calculation of subsequent dynamic tasks.

[0066] Will After decoding, the data is sent to each drone, replacing its current local queue (a smooth transition mechanism can be used to avoid abrupt changes in flight paths). Simultaneously, Updated to the system's new background reference benchmark This is used by the S1–S3 stages when the next batch of dynamic tasks arrives, ensuring that rolling optimization can be continuously advanced.

[0067] Regarding the heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks provided by this invention, the parameters in the DPDM model ( , , The sensitivity analysis results of the impact on the time-varying curve of dynamic priority are as follows: Figure 2 and Figure 3 As shown, Figure 2 and Figure 3 The dynamic priority curves and dynamic priority change rates of the Dynamic Priority Decay Compensation Model (DPDM model) in this method under different value decay rates are shown.

[0068] The heterogeneous UAV cooperative scheduling and trajectory replanning method (MCHS-Dynamic) for dynamic tasks provided by this invention was compared and verified by experiments. The results of multi-metric comparisons between MCHS-Dynamic and four single static programming and greedy rule benchmark algorithms (Static-GA, Greedy-Auction, ER-GA, and MCHS-NoDecay) were presented, including task completion rate comparisons. Figure 4 Energy consumption comparison () Figure 5 Comparison of total travel distance () Figure 6 ), Total Revenue Comparison ( Figure 7 As can be seen, the MCHS-Dynamic of this application is optimal in every metric.

[0069] The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks provided by this invention has the following technical advantages: (1) Using the chimeric exponential decay term With reverse compensation item The Dynamic Priority Evaluation Function (DPDM) enables real-time value calculation and eliminates task starvation; (2) Based on the above dynamic priority, a heterogeneous sensing multi-attribute auction mechanism (HAM-Auction) is constructed to achieve distributed fast task allocation that satisfies millisecond-level single-node closed-loop response and avoids global communication blocking; (3) Use heterogeneous multi-attribute value density function to protect the multi-level capability specialization matching degree With queue load balancing factor This approach combines various methods to break away from the limitations of traditional auctions that rely solely on physical distance. It enables execution decisions to comprehensively reflect the energy efficiency differences, payload specialization, and task backlog of heterogeneous drones. The binary judgment of whether an operation can be executed is upgraded to a multi-level quantification of proficiency, with the highest weighting awarded when the payload required for the task perfectly matches the drone's specialization. The weighting coefficient decreases with the dimension of the capability set only when it is executable within the scope of general capabilities. This replaces the traditional auction practice of simply performing Boolean filtering without distinguishing execution validity; load balancing factor Introducing the current queue backlog length into the denominator of the evaluation metrics naturally increases the bidding cost for drones with more tasks, achieving an endogenous anti-congestion balance within the system without the need for an additional computational layer; and realizing optimal allocation of multiple objectives that comprehensively consider energy consumption, timeliness, and payload capacity.

[0070] (4) By using the event and periodic dual triggering mechanism, the local auction results are encoded as the initial individuals of the genetic algorithm elite and injected into the optimization engine to complete the global trajectory optimal replanning with extremely low iteration cost.

[0071] This achieves an organic unity between instantaneous response to sudden tasks and global optimization throughout the entire lifecycle of the cluster, and fully utilizes the differentiated characteristics of heterogeneous UAVs.

[0072] This invention also discloses a system for the above-mentioned heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks, such as... Figure 8 As shown, the system includes a processing center and multiple heterogeneous drones that communicate wirelessly with the processing center.

[0073] In this embodiment, the processing center is a computing device with data processing and scheduling decision-making capabilities. Specifically, it can be one or a combination of ground servers and edge servers. It is used to perform computationally intensive operations such as initial static task allocation, dynamic task priority calculation, heterogeneous perception auction allocation, and dual-trigger global replanning, and uniformly generate and issue global task allocation schemes and track instructions.

[0074] Each heterogeneous UAV is equipped with an onboard controller, a wireless communication module, a sensor module, and an independent Rolling Time Domain Control (RHC) module. The UAV receives task sequences and trajectory information from the processing center via the wireless communication module. The onboard controller controls the UAV to perform takeoff, landing, cruise, and mission operations, while the independent RHC module enables onboard trajectory tracking, real-time obstacle avoidance, and flight stability control. During flight, the UAV transmits real-time information such as its position, speed, remaining energy, and mission queue status back to the processing center, providing real-time status support for the center's dynamic scheduling and global replanning.

[0075] The above system can be implemented using a distributed computing architecture or a centralized architecture: in scenarios with good communication conditions, the core scheduling calculation is completed by the processing center; in complex environments, the onboard terminal and the processing center can work together to complete local rapid response and global optimization updates, ensuring the real-time performance and global optimality of scheduling in dynamic task scenarios.

[0076] The present invention also discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-mentioned heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks provided by the present invention. The computer program product should be understood as a software product that mainly implements its solution through a computer program, such as a program product integrated in the cloud or a software library.

[0077] In the description of this specification, the references to terms such as "an embodiment," "some embodiments," "example," "specific example," "a implementation," "a preferred implementation," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0078] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks, characterized in that, The method includes: Step S1: Based on the input initial static task set, environmental threat field, and the initial position, energy state, and safe return base point set of one or more heterogeneous UAVs, generate an initial global task allocation scheme. The global task allocation scheme includes the sequence of tasks to be executed and the corresponding flight path for each UAV. Step S2: Use the rolling time domain control module to control the UAV to fly according to its corresponding trajectory, while monitoring whether new dynamic tasks arrive. When a new dynamic task arrives, execute the following steps simultaneously: Step S3: Call the dynamic priority decay compensation model to evaluate the dynamic priority of the dynamic task to be assigned at the current moment; combine the UAV state vector and the dynamic priority of the dynamic task to be assigned to calculate the value density of the dynamic task to be assigned by the UAV; use the value density as the bidding auction for the dynamic task to be assigned by the UAV; after all dynamic tasks to be assigned are assigned, obtain the current global task assignment scheme; if the replanning trigger condition is met, proceed to step S4. Step S4: Using the current global task allocation scheme as the elite individual of the genetic algorithm, execute the genetic algorithm to obtain the optimal global task allocation scheme, send the sequence of tasks to be executed in the optimal global task allocation scheme to the corresponding UAV, and return to execute step S2.

2. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 1, characterized in that, In step S3, the dynamic priority decay compensation model is invoked to evaluate the dynamic priority of the dynamic task to be assigned at the current moment, including: Obtain the time-sensitive attribute vector of the dynamic task to be assigned at the current moment. The time-sensitive attribute vector includes the value decay rate, compensation recovery rate, waiting time, three-dimensional spatial coordinates of the task target point, and task benchmark reference value used to calculate dynamic priority. Substitute the time-sensitive attribute vector of the dynamic task to be assigned at the current moment into the dynamic priority decay compensation model to obtain the dynamic priority of the dynamic task to be assigned at the current moment. The dynamic priority attenuation compensation model is expressed as follows: ； Indicates dynamic tasks to be assigned At the present moment Dynamic priority; Indicates dynamic tasks to be assigned The task benchmark reference value; Represents an exponential function; Indicates dynamic tasks to be assigned The rate of value decay; Indicates the current time Dynamic tasks to be assigned Waiting time; Indicates the first weight; Indicates the second weight; Dynamic tasks to be assigned The compensation recovery rate; This indicates the second weight.

3. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 2, characterized in that, Determine the dynamic tasks to be assigned The switching wait time when the dynamic priority changes from decreasing to increasing for: 。 4. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 2, characterized in that, The time-sensitive attribute vector of the dynamic task to be assigned at the current moment also includes the payload type identifier required by the dynamic task, and the UAV state vector includes the UAV's professional type identifier and the set of payload capabilities it possesses. In step S3, the value density of the UAV undertaking the dynamic task to be assigned is calculated by combining the UAV state vector and the dynamic priority of the dynamic task to be assigned, including: The matching degree between the dynamic task to be assigned and the UAV's capabilities is obtained based on the payload type identifier required by the dynamic task to be assigned, the UAV's professional type identifier, and the set of payload capabilities it possesses. Calculate the estimated additional energy consumption required for an unmanned aerial vehicle (UAV) to fly from the end waypoint of its pending mission sequence to the mission objective point of the dynamic mission to be assigned, and finally return to the nearest safe return base point. ; The load balancing factor of the drone is calculated based on the length of the task sequence being executed by the drone at the current moment; By substituting the matching degree between the dynamic task to be assigned and the UAV's capabilities, the estimated additional energy consumption of the UAV, the length of the task sequence currently being executed by the UAV, and the dynamic priority of the dynamic task to be assigned into the composite value density function, the value density of the UAV undertaking the dynamic task to be assigned is obtained.

5. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 4, characterized in that, The composite value density function is: ； in, Indicates dynamic tasks to be assigned Undertake drones Value density; Indicates dynamic tasks to be assigned At the present moment Dynamic priority; Indicates dynamic tasks to be assigned With drones Compatibility; Indicates drone From the end waypoint of its pending task sequence, fly to the task objective point of the dynamic task to be assigned. And the estimated additional energy required to eventually return to the nearest safe return base point; Indicates drone The load balancing factor , This represents the load penalty weighting coefficient.

6. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 5, characterized in that, In step S3, the value density of the dynamic tasks to be assigned is calculated only for UAVs that meet the remaining energy security constraint test.

7. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 6, characterized in that, In step S3, the dynamic tasks to be assigned to the drone are allocated through a bidding process using value density as the basis for the bidding. This includes: Among the drones that meet the remaining energy safety constraint test, the drone with the highest value density is selected as the winning drone. The dynamic tasks to be assigned are then assigned to the winning drone, and the dynamic tasks are appended to the end of the winning drone's task sequence.

8. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to any one of claims 1-7, characterized in that, The replanning is triggered when the number of times a local task is added reaches a set threshold, or when it is triggered at a fixed time step.

9. The heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks according to claim 8, characterized in that, In step S4, the current global task allocation scheme is used as the elite individual in the genetic algorithm, and the genetic algorithm is executed to obtain the optimal global task allocation scheme, including: Construct an initial population, which includes elite individuals and one or more chromosomes. Chromosomes are randomly initialized, and each chromosome represents a global task allocation scheme. After performing a preset number of iterations of optimization based on the initial population, the optimal global task allocation scheme is obtained.

10. A system based on any one of claims 1-9 for a heterogeneous UAV cooperative scheduling and trajectory replanning method for dynamic tasks, characterized in that, The system includes a processing center and multiple heterogeneous drones that communicate wirelessly with the processing center.

Images