Multi-unmanned aerial vehicle task allocation and cooperative tracking method for multi-target

Abstract

This invention discloses a multi-target, multi-UAV task allocation and cooperative tracking method, comprising: establishing a state model of a multi-UAV system and dynamic targets, and constructing a communication topology; constructing a benefit matrix for each UAV with respect to each dynamic target; executing a kWTA task selection mechanism to obtain selected UAVs and obtain real-time task allocation results; generating expected distances and expected azimuth angles relative to the formation center for the selected UAVs, calculating the target positions of each selected UAV, and forming a tracking configuration; designing a fixed-time cooperative control law for the selected UAVs to converge the states of each UAV to the desired tracking configuration; establishing corresponding control barrier function constraints; constructing a constrained optimization problem to generate control inputs for each UAV and drive UAV movement. This invention can simultaneously meet the requirements of real-time task reallocation, fixed-time convergence, and dynamic obstacle avoidance in dynamic environments.

Description

Technical Field

[0001] This invention relates to the field of multi-UAV cooperative control technology, specifically to a multi-UAV task allocation and cooperative tracking method for multiple targets. Background Technology

[0002] Existing multi-UAV cooperative dynamic target tracking technologies typically employ a layered approach of "task allocation—cooperative control—safety constraints." This means the upper layer determines the matching relationship between the UAV and the target or sub-task; the middle layer handles formation keeping and target tracking control; and the lower layer superimposes collision avoidance and obstacle avoidance constraints. While this layered architecture offers certain modular design advantages, facilitating the design and implementation of allocation algorithms, control laws, and safety mechanisms separately, the relatively independent nature and limited coupling of each layer make it prone to issues such as delayed task reconfiguration, uncoordinated control responses, and delayed intervention of safety constraints in complex scenarios involving rapid target maneuvering, dynamic obstacle changes, and time-varying communication topologies. These problems ultimately limit the overall real-time performance, robustness, and security of the system.

[0003] In task allocation, existing research typically employs methods such as the Hungarian algorithm, integer programming, centralized optimization, auction algorithms, and distributed consensus negotiation to achieve dynamic matching between multiple UAVs and targets. Hungarian algorithms and integer programming methods often achieve good or even globally optimal allocation results. However, these methods largely rely on a central node for unified modeling and centralized solution. Therefore, when the target state changes rapidly, the number of UAVs increases, or communication links suffer from latency or packet loss, they are prone to heavy computational burdens, long redistribution cycles, and sensitivity to single points of failure. In contrast, auction algorithms and distributed negotiation methods have advantages in real-time performance and scalability. However, because their decision-making process relies more on local information, they are prone to slow convergence, allocation results trapped in local optima, frequent task switching, and significant allocation jitter when the target is continuously maneuvering and the payoff function changes rapidly. Therefore, they struggle to balance real-time performance and allocation quality.

[0004] In terms of cooperative control, existing methods for multi-UAV dynamic target tracking mainly include consensus control, leader-follower control, model predictive control, finite-time control, and fixed-time control. Asymptotic consensus control is theoretically mature and has a relatively simple control structure, making it suitable for general formation cooperative problems. However, since the system state typically converges asymptotically only as time approaches infinity, it struggles to meet the requirements of convergence speed and response timeliness for rapidly maneuvering target tracking. While finite-time control can enable the system state to converge within a finite time, thereby improving tracking and reconfiguration speed to some extent, its convergence time usually depends on the initial system state. Therefore, when the initial distribution of UAVs is relatively discrete, missions undergo sudden switching, or formations require rapid reconfiguration, system performance is prone to significant fluctuations due to differences in initial conditions, making it difficult to guarantee a uniform and stable dynamic response.

[0005] Regarding safety constraints, existing multi-UAV cooperative tracking methods mainly achieve inter-UAV collision avoidance and obstacle avoidance through artificial potential fields, virtual repulsion, collision penalty terms, and constraint optimization. Artificial potential field methods are relatively simple to implement and easy to deploy in engineering, but they are prone to getting stuck in local minima and causing oscillations. Penalty-based methods are sensitive to weight parameters and typically cannot provide strict theoretical guarantees of safety. While adding collision avoidance constraints to the controller output can correct the trajectory to some extent, it may also disrupt the original formation tracking performance and even cause conflicts between the controlled target and the safety target.

[0006] Therefore, the existing technology generally has the following shortcomings:

[0007] First, task allocation, collaborative control, and security constraints are usually designed in a hierarchical and independent manner, lacking a unified "allocation-control-security" closed-loop collaborative mechanism. This results in insufficient information transmission between modules and incomplete feedback chains, making it difficult to achieve optimal collaboration at the system level.

[0008] Secondly, when target maneuvering, obstacle changes, and communication topology switching occur simultaneously, existing methods often struggle to simultaneously achieve rapid online task redistribution, stable formation reconfiguration, and continuous satisfaction of safety constraints, resulting in insufficient system adaptability in complex dynamic environments.

[0009] Third, existing security constraint mechanisms typically only operate during the control execution phase and fail to participate in the task reconstruction and trajectory update process. Therefore, they are prone to problems such as being "feasible from an allocation perspective but not feasible from a security perspective" or "trackable in the short term but unsustainable in the long term."

[0010] Fourth, with the increase in the number of drones and the increase in environmental complexity, existing methods often show a significant decline in real-time performance, system robustness and overall collaborative efficiency, making it difficult to meet the collaborative tracking needs in large-scale, multi-constraint and highly dynamic environments.

[0011] In summary, current technologies lack a multi-UAV cooperative dynamic target tracking method that can organically integrate kWTA competitive allocation, fixed-time distributed cooperative control, and control barrier function safety constraints in dynamic environments, thereby simultaneously addressing the three key issues of real-time task reassignment, fast convergence within a fixed timeframe, and dynamic safe obstacle avoidance. In other words, how to construct a unified cooperative optimization framework that combines rapid task reconfiguration, stable cooperative control, and stringent safety guarantees remains a pressing technical problem to be solved in this field. Summary of the Invention

[0012] The purpose of this invention is to provide a multi-UAV task allocation and cooperative tracking method for multiple targets, which addresses the following technical problems in existing multi-UAV cooperative dynamic target tracking: task allocation is highly dependent on the central node, online real-time redistribution is insufficient, cooperative control convergence speed is slow, convergence time is greatly affected by the initial state of the system, and it is difficult to uniformly guarantee inter-UAV collision avoidance and static / dynamic obstacle avoidance with the target tracking process.

[0013] To achieve the above functions, this invention designs a multi-target multi-UAV task allocation and cooperative tracking method. For an environment containing multiple UAVs, multiple dynamic targets, and several obstacles, the following steps S1-S7 are executed to complete the task allocation and cooperative tracking of multiple UAVs:

[0014] Step S1: For a multi-UAV system consisting of multiple UAVs, establish a state model of the multi-UAV system; based on the multiple dynamic targets tracked by each UAV, construct a multi-dynamic target system and its state model; construct a communication topology diagram according to the communication relationships between UAVs.

[0015] Step S2: Based on the relative relationship between the UAV and the dynamic target, the UAV's own resource status, and environmental safety constraints, construct a comprehensive benefit function to represent the task allocation problem as an optimization problem with selection constraints; construct a benefit matrix based on the comprehensive benefit function value of each UAV for each dynamic target;

[0016] Step S3: Execute the kWTA task selection mechanism on the payout matrix to select several winning individuals from the candidate drones for each dynamic target, obtain the selected drones, and get the real-time task allocation results.

[0017] Step S4: Based on the real-time task allocation results, generate the expected distance and expected azimuth angle relative to the formation center for the selected UAVs, calculate the target position of each selected UAV, and form a tracking configuration;

[0018] Step S5: For the selected UAV, design a fixed-time cooperative control law that includes position error term, velocity error term, and cooperative consistency term, so that the state of each UAV converges to the desired tracking configuration within a fixed time.

[0019] Step S6: Establish corresponding control barrier function constraints for the minimum safe distance between drones, the safe distance between drones and static obstacles, and the collision risk between drones and dynamic obstacles;

[0020] Step S7: Combine the task allocation results obtained in Step S3, the fixed-time cooperative control law obtained in Step S5, and the control barrier function constraints constructed in Step S6 into a constrained optimization problem, and solve it online in each sampling period to generate control inputs for each UAV in real time. Send the control inputs to the UAV actuators to drive the UAVs to move and complete the task allocation and cooperative tracking of multiple UAVs.

[0021] The present invention also designs a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the multi-target multi-UAV task allocation and cooperative tracking method.

[0022] The present invention also designs a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the multi-target multi-UAV task allocation and cooperative tracking method.

[0023] Beneficial effects: Compared with the prior art, the advantages of the present invention include:

[0024] 1. This invention uses the kWTA local competition mechanism to replace the traditional centralized allocation, transforming the task allocation problem into a continuous-time competition network for online solution. It relies solely on neighborhood information to complete the decision, effectively overcoming the shortcomings of existing technologies such as strong dependence on the central node, long redistribution cycle, and high computational complexity. It significantly improves the real-time performance and system scalability of online task reconstruction in dynamic environments.

[0025] 2. This invention designs a distributed fixed-time cooperative control law that includes target tracking error, neighbor consistency error, and configuration preservation error. By introducing a nonlinear power error term, the upper bound of the system convergence time is made independent of the UAV's initial state. This invention solves the problems of slow convergence speed and convergence time dependence on initial conditions in existing control methods, and is particularly suitable for complex dynamic scenarios with frequent task switching and dispersed initial formation.

[0026] 3. This invention embeds requirements such as inter-machine safety distance and static / dynamic obstacle avoidance into the online optimization process in the form of control barrier function (CBF) inequality constraints, which mathematically guarantees the forward invariance of the system's safety set. This invention overcomes the shortcomings of existing safety mechanisms, such as delayed intervention and inability to provide strict guarantees, and realizes an active and unified closed-loop design of safety constraints and tracking control, fundamentally eliminating the risk of "allocation is feasible but safety is not feasible".

[0027] 4. This invention, through the rolling execution of the integrated optimization framework of "allocation-control-security", exhibits strong robustness when facing complex disturbances such as enhanced target maneuverability, communication delay, and partial UAV failure. This invention achieves deep coupling and mutual promotion of task allocation, cooperative control and security constraints, significantly improving the overall cooperative efficiency, adaptability and engineering practicality of multi-UAV systems in highly dynamic and complex environments with multiple constraints. Attached Figure Description

[0028] Figure 1 This is a flowchart of a multi-target, multi-UAV task allocation and cooperative tracking method provided by an embodiment of the present invention;

[0029] Figure 2 These are task allocation time consumption curves for three methods provided in the embodiments of the present invention at different scales;

[0030] Figure 3 These are tracking error convergence curves of four methods provided in the embodiments of the present invention under the same initial error conditions;

[0031] Figure 4 This is a graph showing the change of minimum distance between UAVs over time according to an embodiment of the present invention;

[0032] Figure 5 This is a comparison chart of the formation reconstruction time of the method of the present invention and the comparative method provided in the embodiments of the present invention;

[0033] Figure 6 This is a comparison chart of safety constraint satisfaction rate and number of collision events provided by an embodiment of the present invention;

[0034] Figure 7 This is a comparison chart of computation time scalability provided according to an embodiment of the present invention. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0036] This invention provides a multi-target, multi-UAV task allocation and cooperative tracking method, applicable to environments containing multiple UAVs, multiple dynamic targets, and several obstacles, referring to... Figure 1 Perform the following steps S1-S7 to complete the task allocation and collaborative tracking of multiple drones:

[0037] Step S1: For a multi-UAV system consisting of multiple UAVs, establish a state model of the multi-UAV system; based on the multiple dynamic targets tracked by each UAV, construct a multi-dynamic target system and its state model; construct a communication topology diagram according to the communication relationships between UAVs.

[0038] The specific steps of step S1 are as follows:

[0039] Step S1.1: Establish the state model of the multi-UAV system. Suppose the multi-UAV system contains N UAVs, and the state of the i-th UAV is defined as follows:

[0040] ;

[0041] ;

[0042] ;

[0043] in, Indicates the position of the i-th drone; This represents the speed of the i-th drone; This represents the control input (acceleration control quantity) for the i-th UAV. It represents spatial dimensions, consisting of three spatial dimensions: x, y, and z.

[0044] A second-order integral model is used to determine the state of each UAV, as shown in the following equation:

[0045] ;

[0046] ;

[0047] in, Indicates the position of the i-th drone The first derivative, Represents the speed of the i-th drone. The first derivative;

[0048] Step S1.2: Establish the state model of the multi-dynamic target system as follows:

[0049] ;

[0050] Where M is the number of dynamic targets, and j represents the number of one of the dynamic targets; a dynamic target is a dynamically moving tracking target, or multiple tracking slots around a single dynamic target;

[0051] Let the position and velocity of the j-th dynamic target be respectively and The motion of a dynamic target can be based on a known model, an estimated model, or observation data updated in real time by online sensors.

[0052] Step S1.3: Construct the communication topology diagram based on the communication relationships between the UAVs as follows:

[0053] ;

[0054] ;

[0055] in, Represents the communication topology. Let N represent the set of drones, and N represent the number of drones. Represents the set of communication edges;

[0056] If the i-th drone and the m-th drone can communicate, then the adjacency matrix between the i-th drone and the m-th drone is... As shown in the following formula:

[0057] ;

[0058] The communication topology between drones is a bidirectional graph. If the i-th drone and the m-th drone can communicate, then it is assumed that the j-th drone and the m-th drone can also communicate. The adjacency matrix... Satisfy the following formula:

[0059] ;

[0060] in, Let represent the set of communicable neighbors of the i-th drone, that is, the set of all other drones that can directly communicate, sense, or interact with the i-th drone. This indicates that the m-th drone is a communicable neighbor of the i-th drone; This indicates that the m-th drone is not a communicable neighbor of the i-th drone.

[0061] The communication topology does not require direct communication between any two drones, but it does require that a swarm of drones participating in the same collaborative task remain connected as a whole.

[0062] In other words, not every drone needs to communicate with all other drones. However, if several drones are tracking the same target, they must be indirectly connected through at least one communication link.

[0063] This communication topology defines adjacency relationships and local information exchange rules, which are used for subsequent task allocation, formation maintenance, and distributed control.

[0064] The environment model includes static obstacles (fixed positions). ) and dynamic obstacles (with time-varying position and velocity information: Static obstacles can be represented as fixed regions or sets of boundaries; dynamic obstacles are preferably acquired in real time through online sensing modules.

[0065] Step S2: Based on the relative relationship between the UAV and the dynamic target, the UAV's own resource status, and environmental safety constraints, construct a comprehensive benefit function to represent the task allocation problem as an optimization problem with selection constraints; construct a benefit matrix based on the comprehensive benefit function value of each UAV for each dynamic target;

[0066] The specific method for step S2 is as follows:

[0067] For the i-th UAV and the j-th dynamic target, construct a comprehensive benefit function. As shown in the following formula:

[0068] ;

[0069] in, Let be the distance function. It is a azimuth function. For velocity matching function, Let the energy gain function be... For communication revenue function, For risk functions; , , , , , In order, they are respectively , , , , , Weighting coefficients;

[0070] Distance function The specific formula is as follows:

[0071] ;

[0072] in, Let be the Euclidean distance from the i-th UAV to the j-th dynamic target. Indicates the position of the i-th drone. Let be the position of the j-th dynamic target; the smaller the distance, the greater the benefit of the distance function.

[0073] Orientation function The specific formula is as follows:

[0074] ;

[0075] in, The velocity of the i-th drone is represented by the azimuth function, which represents the cosine of the angle between the drone's flight direction and the target direction.

[0076] Velocity matching function The specific formula is as follows:

[0077] ;

[0078] in, Let be the velocity of the j-th dynamic target; when the velocity of the i-th UAV... and the velocity of the j-th dynamic target When there is a perfect match: =0, at this time The speed matching function value reaches its maximum value;

[0079] Energy gain function The specific formula is as follows:

[0080] ;

[0081] in, Let be the remaining energy of the i-th drone. This represents the energy of the drone when fully charged; the remaining energy of the drone is normalized to the value when it is fully charged: , The energy gain function value is maximized when the drone runs out of power. ,so The energy gain function value is minimized;

[0082] Communication revenue function The specific formula is as follows:

[0083] ;

[0084] When the i-th drone can communicate with all other drones: The communication benefit function value is maximized when the i-th drone has no communicable neighbors: , The communication benefit function value is 0;

[0085] Risk function The specific formula is as follows:

[0086] ;

[0087] in, This represents the Euclidean distance from the i-th drone to the k-th obstacle. This indicates the position of the k-th obstacle; the closer the drone is to the obstacle, the more... The smaller the value, the higher the risk function value;

[0088] Based on the comprehensive benefit function, the comprehensive benefit function value of each UAV and each dynamic target is calculated, and a benefit matrix is ​​constructed to uniformly characterize factors such as "distance, direction, speed, energy, communication, and security" to achieve the comprehensive optimal allocation of multiple objectives.

[0089] Step S3: Execute the kWTA (k-Winners-Take-All) task selection mechanism on the payout matrix to select several winning individuals from the candidate drones for each dynamic target, obtain the selected drones, and get the real-time task allocation results;

[0090] Step S3 transforms the kWTA competition process into an equivalent quadratic programming problem or a continuous-time competition network solution problem, and uses a finite-time neural dynamics solver or a fixed-time neural dynamics solver for online solution to improve the task redistribution speed in dynamic environments and reduce the computational dependence on the central node. The specific steps are as follows:

[0091] Step S3.1: For the payoff matrix, execute the kWTA task selection mechanism and output binary selection variables. The following equation is satisfied:

[0092] ;

[0093] in, This indicates that the i-th drone is assigned to the j-th dynamic target. This indicates that the variable has not been assigned; the selected variable must meet the quantity constraints on both the target side and the UAV side to ensure that the assignment result meets the task requirements and resource constraints.

[0094] Step S3.2: Based on the binary selection variables and the comprehensive benefit function, construct the optimization problem as follows:

[0095] ;

[0096] The constraints for the optimization problem are as follows:

[0097] ;

[0098] ;

[0099] in, This represents the number of drones needed to be allocated for the j-th dynamic target; the first constraint indicates that for any j-th dynamic target, from... Selected from drones Each drone is responsible for this dynamic objective; the second constraint states that each drone can be assigned at most one task, and for the i-th drone, all its corresponding tasks... Within the system, at most one value can be equal to 1, while the rest are 0. This avoids the conflict of "a drone being assigned multiple tasks at the same time".

[0100] Traditional task allocation methods, such as manual sorting or greedy algorithms, first sort by reward and then allocate tasks sequentially. Which of these methods... The value of 1 is manually set and fixed; however, in this optimization problem, this invention only provides the objective and constraints, and the solver will automatically calculate all... The combination that yields high returns without violating constraints, corresponding to It will be set to 1; combinations with low returns or that may cause conflicts will be set accordingly. It will be set to 0.

[0101] The objective function aims to maximize the total reward across all assigned tasks. Only when... At that time, the corresponding comprehensive return function value Only then will it be added to the sum. When that happens, the corresponding items disappear. This is determined by selecting which... Choose some values ​​to be 1, and others to be 0, so that the combined return function value of all selected values ​​is... The sum is the largest.

[0102] Step S3.3: The continuous competition dynamics are implemented as follows:

[0103] Constructing continuous allocation variables: , is defined as the competition intensity between the i-th UAV and the j-th dynamic target. The closer the value is to 1, the greater the competitive advantage of the i-th UAV over the j-th dynamic target, and the more likely the j-th dynamic target should be assigned to it. The closer it is to 0, the greater the competitive disadvantage of the i-th drone against the j-th dynamic target, and the less likely it is to be assigned the j-th dynamic target;

[0104] Assign variables The evolution equation is defined as: ;

[0105] : It will grow larger, and the intensity of competition will increase;

[0106] : It will become smaller, and the intensity of competition will decrease.

[0107] The formula can be summarized as follows:

[0108] The rate of change of competition intensity = (incentives for self-reward) - (suppression of global task competition) - (suppression of multi-task competition within the UAV itself).

[0109] in, As a column constraint, when too many drones compete for the same task / too many tasks are selected, It will automatically enlarge and suppress all. growth, to prevent exceeding One task was selected; For execution constraints, when a drone competes for multiple tasks simultaneously, It will automatically increase in size to suppress its competitiveness against other tasks and prevent "one machine from doing multiple tasks"; To control the convergence rate of the allocation variable, the competitive dynamic gain is used.

[0110] For binary selection variables Introducing allocation variables The specific formula is as follows:

[0111] ;

[0112] in, For the threshold of the allocation variable, It is not an independent threshold, but a cutoff value derived from the sorting results. The determination process is as follows:

[0113] right All elements in column j Sort in descending order to obtain the sequence:

[0114] ;

[0115] in, This represents the k-th largest element in the j-th column;

[0116] Define the threshold for the allocation variable : Obtain k selected drones.

[0117] Thus, all greater than or equal to The first k elements ( All of them will be discretized to 1, while the remaining Nk elements will be discretized to 0, thus satisfying the condition. Constraints.

[0118] Step S3 allows for rapid updates to task allocation results in case of target maneuvering, drone straggling, changes in communication links, or sudden obstacle appearances. When the benefits... When changes occur, the network will automatically adjust. The allocation plan is updated in real time.

[0119] In layman's terms, each drone is "competing for tasks," with those offering higher rewards being more likely to be selected. Simultaneously, the system automatically avoids having too many drones occupying a single task, and conflicts arising from a single drone performing multiple tasks simultaneously. This transforms the non-convex, computationally intensive 0-1 integer programming problem into a continuous competition within the system, where each drone has a specific task... The final benefit is continuous. The allocation variable values ​​and their corresponding meanings for the drone in the embodiment are shown in Table 1:

[0120] Table 1. Assignment Variable Values ​​and Their Meanings for Unmanned Aerial Vehicles (UAVs)

[0121]

[0122] Step S4: Based on the real-time task allocation results, generate the expected distance and expected azimuth angle relative to the formation center for the selected UAVs, calculate the target position of each selected UAV, and form a tracking configuration;

[0123] The specific method for step S4 is as follows:

[0124] The target position of the drone is calculated using the following formula:

[0125] ;

[0126] in, Indicates the target location of the i-th drone. Indicates the position of the j-th dynamic target; This indicates that the i-th drone has arrived at the formation center. The expected distance is a scalar; This indicates that the i-th drone is relative to the formation center. The expected azimuth is a scalar; Indicates the expected distance Convert to a length of , direction is A two-dimensional vector;

[0127] Calculate the target positions of all selected UAVs to form a tracking configuration.

[0128] For single-target cooperative tracking scenarios, multiple desired tracking slots can be set around the target, such as left front slot, right front slot, lateral slot, or surrounding slot, and the selected drone can be mapped to the corresponding slot.

[0129] For multi-target scenarios, the number and configuration parameters of tracking drones can be dynamically allocated to different targets based on target priority, target movement direction, threat level, and observability requirements.

[0130] Unselected drones can execute standby, fill in, edge patrol, or connectivity maintenance strategies to participate in mission reconfiguration in subsequent sampling cycles.

[0131] Step S5: For the selected UAV, design a fixed-time cooperative control law that includes position error term, velocity error term, and cooperative consistency term, so that the state of each UAV converges to the desired tracking configuration within a fixed time.

[0132] The specific method for step S5 is as follows:

[0133] Step S5.1: Let the tracking configuration error of the i-th UAV be as follows:

[0134] ;

[0135] ;

[0136] in, Indicates the target location of the drone; Let i be the current position of the i-th drone and its target position. The difference, the target is →0, meaning fly to the desired location; Let i be the current velocity of the i-th UAV and the velocity of the dynamic target. The difference, the target is →0, meaning maintaining the same speed as the dynamic target; Let the velocity be the velocity of the j-th dynamic target;

[0137] Step S5.2: The control law adopts a position-velocity dual closed-loop structure. The upper layer generates a reference trajectory based on the allocation result and the slot mapping relationship. The lower layer designs position and velocity closed-loop controllers based on the fixed-time stability principle. A fixed-time cooperative control law containing position error terms, velocity error terms, and cooperative consistency terms is constructed, as shown in the following formula:

[0138] ;

[0139] in, For the control input of the i-th UAV, This is the position error term, enabling the UAV to quickly converge to the target position. , For the speed error term, ensure the drone's speed keeps up with the dynamic target's speed. , For a coherent consistency item, Let m be the location of the m-th drone; let i be the location of the i-th drone and its neighboring drones. The weighted sum of the location differences. When the i-th drone is too far from its neighbor... It will grow larger, and the control input will "pull" it back into formation; when the i-th drone is too close to its neighbor, It will shrink or even reverse, and the control input will "push" it away to maintain the formation spacing; This represents the error term with exponentiation. This is a typical form of fixed-time / finite-time control. The larger the error, the stronger the control input, and the faster the convergence speed is compared to ordinary linear control. , The exponential parameter satisfies the fixed-time convergence requirement; It represents the acceleration of a dynamic target. As feedforward compensation, it allows the control input of the UAV to directly offset the changes in the target's acceleration, thus preventing the UAV from being unable to keep up and experiencing tracking lag when the target accelerates / decelerates. , , To control the gain, used to adjust the control strength, and .

[0140] By designing parameters appropriately, the upper bound of the system convergence time can be made independent of the initial state, thereby ensuring that configuration establishment and target cooperative tracking can still be completed within a fixed time when switching tasks, rapidly reconfiguring formations, or initially dispersing deployments. This results in a fixed-time cooperative controller that integrates "position stabilization + velocity tracking + target feedforward + formation coordination".

[0141] Furthermore, the control law can also simultaneously consider velocity constraints, acceleration constraints, and actuator saturation constraints to improve the engineering feasibility of the method.

[0142] This control method enables drones to not only fly quickly to a designated location, but also maintain synchronized movement with the target while automatically maintaining formation.

[0143] Step S6: Establish corresponding control barrier function (CBF) constraints for the minimum safe distance between drones, the safe distance between drones and static obstacles, and the collision risk between drones and dynamic obstacles.

[0144] The specific method for step S6 is as follows:

[0145] For any i-th drone and m-th drone, the inter-drone safety function is constructed as follows:

[0146] ;

[0147] in, This represents the inter-drone safety function between the i-th and m-th drones. To maintain the minimum safe distance between machines, Let i be the position of the i-th drone; Let m be the position of the m-th drone;

[0148] When the safety constraints are met When this occurs, it indicates that the system is within the safe set;

[0149] Introduce control barrier function constraints:

[0150] ;

[0151] in, To control the convergence coefficient of the barrier function, the convergence speed of the safe set is controlled.

[0152] Expanding the above equation:

[0153] ;

[0154] in, Let be the speed of the i-th drone; Let m be the speed of the m-th drone;

[0155] The control input of the i-th drone and Connecting them:

[0156] ;

[0157] ;

[0158] Taking the derivative of the expansion, we get:

[0159] ;

[0160] in, This is the control input for the m-th UAV;

[0161] The constraint form of the second-order control barrier function is:

[0162] ;

[0163] in, , These are the safety convergence coefficients of the second-order control barrier function, all of which are positive positive constants used to adjust the response speed of the safety constraints.

[0164] Will , Substituting and rearranging all terms, we get:

[0165] ;

[0166] Among them, let ;

[0167] ;

[0168] The inter-machine control barrier function constraint is obtained as follows:

[0169] ;

[0170] , For machine room safety constraints;

[0171] For both static and dynamic obstacles, corresponding safety functions can be constructed and control barrier function constraints can be formed.

[0172] For the i-th drone and the k-th static obstacle, construct a static obstacle safety function:

[0173] ;

[0174] in, Let the fixed position of the k-th static obstacle be... Let i be the position of the i-th drone. Minimum safe distance from static obstacles;

[0175] Because static obstacles have fixed positions The derivatives are:

[0176] ;

[0177] ;

[0178] Substitute the second-order control barrier function constraints:

[0179] ;

[0180] The static obstacle control barrier function constraints are as follows:

[0181] ;

[0182] Right now:

[0183] ;

[0184] in, , Safety constraints for static obstacles;

[0185] For the i-th drone and the l-th dynamic obstacle, construct a dynamic obstacle safety function:

[0186] ;

[0187] in, The time-varying position of dynamic obstacles. , These are its velocity and acceleration, respectively;

[0188] remember , The derivatives are:

[0189]

[0190] ;

[0191] in, Let be the real-time acceleration vector of the l-th dynamic obstacle;

[0192] Substituting the second-order control barrier function constraints, we obtain the dynamic obstacle control barrier function constraints:

[0193] ;

[0194] Right now:

[0195] ;

[0196] in, , Safety constraints for dynamic obstacles;

[0197] The three types of safety constraints—machine room safety, static obstacles, and dynamic obstacles—are merged:

[0198] ;

[0199] The control input is optimized as follows:

[0200] ;

[0201] ;

[0202] in, , This represents the constraint matrix consisting of inter-machine safety constraints, static obstacle safety constraints, and dynamic obstacle safety constraints. This is the optimized control input for the i-th UAV; This is a safety constraint inequality constructed from the control barrier function.

[0203] Objective function: to adjust the control input As close as possible to the nominal control This ensures that the mission is not affected.

[0204] Constraints: All control barrier function constraints must be satisfied to ensure that the UAV does not enter the danger zone; this inequality explicitly limits the control input. The feasible domain is determined to ensure that the system is always within the safe set.

[0205] If the system dynamics is a high-order model, then a high-order control barrier function (HOCBF) can be used to handle high-order constraints.

[0206] Preferably, the control barrier function constraint and the fixed-time collaborative control objective of step S5 are uniformly incorporated into the distributed optimization problem, and the control input is corrected in real time so that the system can always meet the safety constraints while maintaining the target tracking and configuration convergence performance.

[0207] The control barrier function constraint is like adding a "safety shell" to the drone. Once it approaches a dangerous area, the control system will automatically adjust its movement direction to avoid collision.

[0208] Step S7: Combine the task allocation results obtained in Step S3, the fixed-time cooperative control law obtained in Step S5, and the control barrier function constraints constructed in Step S6 into a constrained optimization problem, and solve it online in each sampling period to generate control inputs for each UAV in real time. Send the control inputs to the UAV actuators to drive the UAVs to move and complete the task allocation and cooperative tracking of multiple UAVs.

[0209] The specific method for step S7 is as follows:

[0210] The constrained optimization problem is constructed as follows:

[0211] ;

[0212] ;

[0213] in, The control input for the UAV is obtained from the fixed-time cooperative control law; The control input for the i-th UAV after satisfying the safety constraints is the control input variable to be solved. This is a safety constraint inequality constructed from the control barrier function.

[0214] While minimizing changes to the original control objectives (target tracking and formation keeping), the control inputs are adjusted to the smallest possible extent to ensure that the system always meets safety constraints.

[0215] Rolling closed-loop execution process:

[0216] During system operation, the following steps are repeatedly executed at a fixed sampling period:

[0217] (1) Obtain information on the target status, the UAV's own status, the status of neighboring UAVs, and environmental obstacles;

[0218] (2) Update the revenue function based on the current state information and construct a new task allocation revenue matrix;

[0219] (3) Execute the kWTA competition mechanism to obtain the task allocation result at the current moment;

[0220] (4) Generate the corresponding expected tracking configuration based on the task allocation results;

[0221] (5) Calculate the nominal control input based on the fixed-time cooperative control law, and construct safety constraints in combination with the control barrier function;

[0222] (6) Solve the above optimization problem to obtain the actual control input that satisfies the safety constraints;

[0223] (7) Send control inputs to the UAV actuators to drive the UAV to move;

[0224] (8) Enter the next sampling period and repeat the above process.

[0225] In layman's terms, step S7 is equivalent to a "real-time decision loop system." At every moment, the UAV will re-perceive the environment, reallocate tasks, recalculate control actions, and execute the optimal motion while ensuring that no collision occurs, thereby achieving continuous and stable tracking of dynamic targets.

[0226] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the multi-target multi-UAV task allocation and cooperative tracking method.

[0227] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the multi-target multi-UAV task allocation and cooperative tracking method.

[0228] In summary, the present invention aims to solve the following problems:

[0229] Firstly, under conditions of continuous maneuvering of dynamic targets, changes in the number of UAVs, and possible switching of communication topologies, how to achieve online rapid allocation and real-time reconstruction of tracking tasks or tracking slots based on target status, UAV status, and environmental constraint information;

[0230] Secondly, after the task allocation results change, how can the UAVs undertaking the tracking task complete formation reconstruction within a fixed time limit independent of the initial state and form a stable cooperative tracking of the target?

[0231] Third, given the presence of static obstacles, dynamic obstacles, and the risk of inter-machine collisions in the environment, how can safety constraints be explicitly embedded into the collaborative optimization and control process to improve system operational safety, robustness, and engineering feasibility while ensuring tracking accuracy and system response speed?

[0232] Therefore, the present invention essentially aims to solve the problem of how to construct a multi-UAV closed-loop cooperative tracking control method that uniformly couples task allocation, cooperative control, and safety constraints in a dynamic environment, so as to simultaneously meet the requirements of real-time task redistribution, fixed-time convergence, and dynamic safety obstacle avoidance.

[0233] The following section provides a detailed explanation of the effects of each technology, using simulation data and graphs:

[0234] 1. Significantly improved efficiency and low computational overhead in real-time task reallocation: This invention employs a kWTA local competition mechanism, transforming the task allocation problem into an online solution using a continuous-time competition network. Decision-making relies solely on neighborhood information, fundamentally eliminating the dependence of centralized allocation on a central node. Simulation experiments, within a UAV scale range of N=5 to N=20, systematically compared the task allocation time of this invention's method, the Auction algorithm, and the Hungarian algorithm (number of dynamic targets M=3, number of obstacles 5).

[0235] Figure 2 The figure shows the task allocation time curves for the three methods at different scales. Experimental results show that when N=20, the single allocation time of the method of this invention is about 50 ms, the auction algorithm is about 100 ms, and the Hungarian algorithm is about 350 ms. The method of this invention is about 50% faster than the auction algorithm and about 85% faster than the Hungarian algorithm. More importantly, the computation time of the method of this invention increases approximately linearly with the scale, with an increase of only about 18% (N=5→N=20), while the centralized method (Hungarian algorithm) increases by more than 900%. Under the constraint of a real-time control period of 50 ms, the centralized method can no longer meet the real-time requirements when N>14, while the method of this invention still remains within the control period when N=20.

[0236] 2. Excellent fixed-time convergence performance, with convergence time independent of the initial state: This invention designs a distributed fixed-time cooperative control law based on target tracking error, neighbor consistency error, and configuration preservation error. By introducing a nonlinear power-law error term, the upper bound T of the system convergence time is made more efficient. max It is independent of the initial state. Figure 3 The tracking error convergence curves of four methods are shown under the same initial error conditions (initial tracking error of approximately 3.2 m, N=10 UAVs).

[0237] Simulation results show that the proposed method converges the tracking error to below 0.05 m within approximately 2.9 s and remains stable without oscillations. The auction algorithm + finite-time method takes approximately 6-8 s to approach convergence and exhibits significant lag. The methods using only kWTA allocation and only superimposed CBF post-processing have the slowest convergence speed and exhibit large oscillations and residual errors. The proposed method has a convergence speed approximately 2.5 times that of the suboptimal method, and in tests with a wide range of random initial distributions (initial errors uniformly distributed between 1 and 8 m), the maximum convergence time does not exceed 3.5 s in 100 independent simulations, with a standard deviation of only 0.18 s, verifying the consistency of fixed-time convergence.

[0238] 3. Strict and effective CBF safety constraints with zero safety violation rate: This invention unifies inter-machine safety constraints, static obstacle constraints, and dynamic obstacle constraints into the online optimization process in the form of a control barrier function (CBF) inequality, mathematically guaranteeing the forward invariance of the safety set, so that safety constraints are always actively satisfied during the control execution phase, rather than being modified afterward. Figure 4 The curves showing the change of the minimum distance between UAVs under four methods during a 60-second simulation (minimum safety threshold of 1.5 m, with 3 dynamic obstacles) are presented.

[0239] The results show that the method of this invention maintains a minimum distance of over 2.0 m between all UAVs throughout the 60 s process, never falling below the safety threshold of 1.5 m, and no collision events occur (collision event count = 0). In contrast, the kWTA Only method experienced 12 safety violations where the distance fell below the safety threshold. Although the CBF Post-processing method reduced this to 5 violations, it still could not completely avoid them because the CBF constraint is only superimposed on the control output layer and cannot adjust the tracking trajectory in advance. The Auction+Finite-time method experienced 3 safe approach events.

[0240] 4. Fast and stable formation reconfiguration with strong robustness to disturbances: When the target suddenly maneuvers, the UAV experiences a partial communication failure, or some UAVs temporarily fail, this invention enables the system to complete formation reconfiguration within a bounded time by rolling execution of kWTA competitive allocation and fixed-time control closed-loop update, and the reconfiguration time is not affected by the magnitude of the initial deviation. Figure 5 The formation reconstruction time of the proposed method and the comparative method are compared under five scenarios: no disturbance, communication delay (0.1 s), enhanced target maneuver, UAV failure (one UAV temporarily withdraws), and combined disturbance (bar chart, N=10, M=3).

[0241] Data shows that under the most severe combined perturbation scenario, the average formation reconstruction time of the method in this invention is 3.1 s, while the auction algorithm + finite-time method is 7.2 s, and the kWTA Only method is as high as 10.3 s. The reconstruction time of the method in this invention is reduced by about 57% compared with the suboptimal method and by about 70% compared with the kWTA Only method. In 100 Monte Carlo stochastic simulation tests, the reconstruction time variance of the method in this invention is less than 0.12 s² under various perturbation scenarios, demonstrating excellent uniform convergence and robustness. In addition, compared with the prior art, the trajectory oscillation amplitude of this invention is reduced by about 40% in multi-target switching scenarios, and the root mean square value (RMSE) of formation error is reduced by about 35%, resulting in more stable system operation.

[0242] 5. Leading safety constraint compliance rate, effectively eliminating collision risks: Figure 6 This paper presents the safety constraint satisfaction rate and the number of collision / approach events for five methods in a 60-second simulation (including 5 dynamic obstacles). The method of this invention consistently satisfies the CBF inequality constraint throughout the entire simulation period, achieving a 100.0% safety constraint satisfaction rate and zero collision events. In contrast, the traditional Artificial Potential Field (APF) method, due to its inherent local minima problem, achieves a safety satisfaction rate of only 83.3% and as many as 9 collision events. While CBF post-processing shows improvement (91.5%, 5 events), it still cannot completely eliminate risks due to the delayed intervention of the safety mechanism. This invention eliminates the "lag" problem of safety constraints by embedding CBF constraints pre-emptively into the control layer and unifying the closed loop with task allocation.

[0243] 6. Highly scalable and practical for engineering applications: This invention adopts a distributed architecture, where each drone only needs to interact with its local neighbors, eliminating the need for a central node for global planning. Figure 7The computation time per step of the proposed method (distributed) and the centralized optimization method were compared under different scales, and a control cycle limit (50 ms) was marked. The results show that the proposed method takes approximately 15 ms per step when N=20, far below the 50 ms control cycle limit; the centralized method exceeds the control cycle (approximately 105 ms) when N>14, failing to meet real-time requirements, and takes approximately 240 ms when N=20, 16 times longer than the proposed method. In the full range of tests from N=5 to N=20, the computation time per step of the proposed method increases by only 18%, exhibiting near-linear scalability and demonstrating excellent scale scalability. Furthermore, the proposed method has a clear control law structure and well-defined interfaces between modules, making it suitable for engineering deployment on embedded flight control platforms.

[0244] 7. Overall Technical Performance Comparison Overview: Table 2 summarizes the comparison results between the method of this invention and existing technologies in terms of key technical indicators. All data are derived from the above simulation experiments and are reproducible and statistically significant.

[0245] Table 2. Comparison of the overall technical performance of the method of the present invention and existing technologies

[0246]

[0247] In summary, this invention achieves a technological breakthrough by unifying kWTA competitive allocation, fixed-time cooperative control, and CBF safety constraints into a closed-loop design. This results in significantly better performance than existing hierarchical independent design methods in terms of task allocation real-time performance, control convergence speed, safety assurance, robustness, and scalability. It realizes a three-in-one technical breakthrough of "allocation-control-safety" and has important theoretical significance and engineering application value for large-scale multi-UAV cooperative tracking tasks in complex dynamic environments.

[0248] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for multi-target multi-UAV task allocation and cooperative tracking, characterized in that, For an environment containing multiple drones, multiple dynamic targets, and several obstacles, perform the following steps S1-S7 to complete the task allocation and collaborative tracking of multiple drones: Step S1: For a multi-UAV system consisting of multiple UAVs, establish a state model of the multi-UAV system; based on the multiple dynamic targets tracked by each UAV, construct a multi-dynamic target system and its state model; construct a communication topology diagram according to the communication relationships between UAVs. Step S2: Based on the relative relationship between the UAV and the dynamic target, the UAV's own resource status, and environmental safety constraints, construct a comprehensive benefit function to represent the task allocation problem as an optimization problem with selection constraints; A payoff matrix is ​​constructed based on the comprehensive payoff function value of each drone for each dynamic target; Step S3: Execute the kWTA task selection mechanism on the payout matrix to select several winning individuals from the candidate drones for each dynamic target, obtain the selected drones, and get the real-time task allocation results. Step S4: Based on the real-time task allocation results, generate the expected distance and expected azimuth angle relative to the formation center for the selected UAVs, calculate the target position of each selected UAV, and form a tracking configuration; Step S5: For the selected UAV, design a fixed-time cooperative control law that includes position error term, velocity error term, and cooperative consistency term, so that the state of each UAV converges to the desired tracking configuration within a fixed time. Step S6: Establish corresponding control barrier function constraints for the minimum safe distance between drones, the safe distance between drones and static obstacles, and the collision risk between drones and dynamic obstacles; Step S7: Combine the task allocation results obtained in Step S3, the fixed-time cooperative control law obtained in Step S5, and the control barrier function constraints constructed in Step S6 into a constrained optimization problem, and solve it online in each sampling period to generate control inputs for each UAV in real time. Send the control inputs to the UAV actuators to drive the UAVs to move and complete the task allocation and cooperative tracking of multiple UAVs.

2. The method for multi-target multi-UAV task allocation and cooperative tracking according to claim 1, characterized in that, The specific steps of step S1 are as follows: Step S1.1: Establish the state model of the multi-UAV system. Suppose the multi-UAV system contains N UAVs, and the state of the i-th UAV is defined as follows: ； ； ； in, Indicates the position of the i-th drone; This represents the speed of the i-th drone; This represents the control input for the i-th drone; It represents spatial dimensions, consisting of three spatial dimensions: x, y, and z. A second-order integral model is used to determine the state of each UAV, as shown in the following equation: ； ； in, Indicates the position of the i-th drone The first derivative, Represents the speed of the i-th drone. The first derivative; Step S1.2: Establish the state model of the multi-dynamic target system as follows: ； Where M is the number of dynamic targets, and j represents the number of one of the dynamic targets; Let the position and velocity of the j-th dynamic target be respectively and ; Step S1.3: Construct the communication topology diagram based on the communication relationships between the UAVs as follows: ； ； in, Represents the communication topology. Let N represent the set of drones, and N represent the number of drones. Represents the set of communication edges; If the i-th drone and the i-th If drones can communicate, then the i-th drone and the i-th drone can communicate. Adjacency matrix between drones As shown in the following formula: ； The communication topology between drones is a bidirectional graph. If the i-th drone and the i-th drone... If one drone can communicate, then the default is the first one. The i-th drone and the i-th drone can also communicate, and the adjacency matrix... Satisfy the following formula: ； in, Let i represent the set of communicable neighbors of the i-th drone. Indicates the first Each drone is a communicable neighbor of the i-th drone; Indicates the first The i-th drone is not a communicable neighbor of the i-th drone.

3. The method for multi-target multi-UAV task allocation and cooperative tracking according to claim 1, characterized in that, The specific method for step S2 is as follows: For the i-th UAV and the j-th dynamic target, construct a comprehensive benefit function. As shown in the following formula: ； in, Let be the distance function. It is a azimuth function. For velocity matching function, Let the energy gain function be... For communication revenue function, For risk functions; , , , , , In order, they are respectively , , , , , Weighting coefficients; Distance function The specific formula is as follows: ； in, Let be the Euclidean distance from the i-th UAV to the j-th dynamic target. Indicates the position of the i-th drone. Let j be the position of the j-th dynamic target; Orientation function The specific formula is as follows: ； in, This represents the speed of the i-th drone; Velocity matching function The specific formula is as follows: ； in, Let be the velocity of the j-th dynamic target; when the velocity of the i-th UAV... and the velocity of the j-th dynamic target When there is a perfect match: =0, at this time The speed matching function value reaches its maximum value; Energy gain function The specific formula is as follows: ； in, Let be the remaining energy of the i-th drone. This represents the energy of the drone when fully charged; the remaining energy of the drone is normalized to the value when it is fully charged: , The energy gain function value is maximized when the drone runs out of power. ,so The energy gain function value is minimized; Communication revenue function The specific formula is as follows: ； When the i-th drone can communicate with all other drones: The communication benefit function value is maximized when the i-th drone has no communicable neighbors: , The communication benefit function value is 0; Risk function The specific formula is as follows: ； in, This represents the Euclidean distance from the i-th drone to the k-th obstacle. Indicates the position of the k-th obstacle; Based on the comprehensive benefit function, the comprehensive benefit function value of each UAV and each dynamic target is calculated, and a benefit matrix is ​​constructed.

4. The multi-target multi-UAV task allocation and cooperative tracking method according to claim 1, characterized in that, The specific steps of step S3 are as follows: Step S3.1: For the payoff matrix, execute the kWTA task selection mechanism and output binary selection variables. The following equation is satisfied: ； in, This indicates that the i-th drone is assigned to the j-th dynamic target. This indicates that it has not been allocated; Step S3.2: Based on the binary selection variables and the comprehensive benefit function, construct the optimization problem as follows: ； The constraints for the optimization problem are as follows: ； ； in, This represents the number of drones that need to be allocated to the j-th dynamic target; Step S3.3: The continuous competition dynamics are implemented as follows: Constructing continuous allocation variables: , is defined as the competition intensity between the i-th UAV and the j-th dynamic target. The closer the value is to 1, the greater the competitive advantage of the i-th UAV over the j-th dynamic target, and the more likely the j-th dynamic target should be assigned to it. The closer it is to 0, the greater the competitive disadvantage of the i-th drone against the j-th dynamic target, and the less likely it is to be assigned the j-th dynamic target; Assign variables The evolution equation is defined as: ; in, For column constraints, For row constraints, To control the convergence rate of the allocation variable, the competitive dynamic gain is used. For binary selection variables Introducing allocation variables The specific formula is as follows: ； in, For the threshold of the allocation variable, It is not an independent threshold, but a cutoff value derived from the sorting results. The determination process is as follows: right All elements in column j Sort in descending order to obtain the sequence: ； in, This represents the k-th largest element in the j-th column; Define the threshold for the allocation variable : Obtain k selected drones.

5. A multi-target, multi-UAV task allocation and cooperative tracking method according to claim 1, characterized in that, The specific method for step S4 is as follows: The target position of the drone is calculated using the following formula: ； in, Indicates the target location of the i-th drone. Indicates the position of the j-th dynamic target; This indicates that the i-th drone has arrived at the formation center. The expected distance; This indicates that the i-th drone is relative to the formation center. The expected azimuth angle; Indicates the expected distance Convert to a length of , direction is A two-dimensional vector; Calculate the target positions of all selected UAVs to form a tracking configuration.

6. The multi-target multi-UAV task allocation and cooperative tracking method according to claim 1, characterized in that, The specific method for step S5 is as follows: Step S5.1: Let the tracking configuration error of the i-th UAV be as follows: ； ； in, Indicates the target location of the drone; Let i be the current position of the i-th drone and its target position. The difference; Let i be the current velocity of the i-th UAV and the velocity of the dynamic target. The difference; Let the velocity be the velocity of the j-th dynamic target; Step S5.2: Construct a fixed-time cooperative control law that includes a position error term, a velocity error term, and a cooperative consistency term, as shown in the following formula: ； in, For the control input of the i-th UAV, For the position error term, For the speed error term, For a coherent consistency item, Let m be the position of the m-th drone; This represents the error term with exponentiation. , For exponential parameters; Indicates the acceleration of a dynamic target; , , To control the gain, and .

7. A multi-target, multi-UAV task allocation and cooperative tracking method according to claim 1, characterized in that, The specific method for step S6 is as follows: For any i-th drone and m-th drone, the inter-drone safety function is constructed as follows: ； in, This represents the inter-drone safety function between the i-th and m-th drones. To maintain the minimum safe distance between machines, Let i be the position of the i-th drone; Let m be the position of the m-th drone; When the safety constraints are met When this occurs, it indicates that the system is within the safe set; Introduce control barrier function constraints: ； in, To control the convergence coefficient of the barrier function; The inter-machine control barrier function constraint is: ； , For machine room safety constraints; among which, , ; , represents the safe convergence coefficient of the second-order control barrier function; Let m be the speed of the m-th drone; The control input for the m-th UAV; For the i-th drone and the k-th static obstacle, construct a static obstacle safety function. : ； in, Let the fixed position of the k-th static obstacle be... Let i be the position of the i-th drone. Minimum safe distance from static obstacles; The static obstacle control barrier function constraint is: ； in, , Safety constraints for static obstacles; , ; For the i-th drone and the l-th dynamic obstacle, construct a dynamic obstacle safety function. : ； in, The time-varying position of dynamic obstacles. , These are its velocity and acceleration, respectively; The dynamic obstacle control barrier function constraints are: ； in, , Safety constraints for dynamic obstacles; , ;remember , ; The three types of safety constraints—machine room safety, static obstacles, and dynamic obstacles—are merged: ； The control input is optimized as follows: ； ； in, , This represents the constraint matrix consisting of inter-machine safety constraints, static obstacle safety constraints, and dynamic obstacle safety constraints. This is the optimized control input for the i-th UAV; These are the safety constraint inequalities constructed from the control barrier function; The control input is optimized as follows: ； ； in, This represents the constraint matrix composed of the gradients of the safety function. This represents a constraint vector containing state terms and safety margins. This is the optimized control input for the i-th UAV; This is a safety constraint inequality constructed from the control barrier function.

8. A multi-target, multi-UAV task allocation and cooperative tracking method according to claim 1, characterized in that, The specific method for step S7 is as follows: The constrained optimization problem is constructed as follows: ； ； in, The control input for the UAV is obtained from the fixed-time cooperative control law; The control input for the i-th UAV after satisfying the safety constraints is the control input variable to be solved. This is a safety constraint inequality constructed from the control barrier function.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements each step of the multi-target multi-UAV task allocation and cooperative tracking method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the multi-target multi-UAV task allocation and cooperative tracking method according to any one of claims 1 to 8.

Images