Unmanned aerial vehicle swarm perception resource joint scheduling method based on alternating direction penalty method
Patent Information
- Application Number
- CN202310884614.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-18
- Publication Date
- 2026-09-22
- Estimated Expiration
- 2043-07-18
AI Technical Summary
[0096]本发明有益效果：本发明所述的多无人机编队感知资源联合调度方法，首先构造基于信号能量检测的多无人机联合感知模型、提出多无人机感知资源联合调度问题，然后将原问题转化为带约束的新问题、写为增广拉格朗日函数形式，最后设计基于交替方向惩罚法的求解方式，实现对多无人机编队感知资源有效调度。该方法在各无人机频域、空域、波束域等资源受限条件下实现了目标感知性能最优化。
Smart Images
Figure CN116963294B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of communication sensing technology, specifically to the design of a joint scheduling method for sensing resources of unmanned aerial vehicle (UAV) swarms based on the alternating direction penalty method. Background Technology
[0002] As a novel intelligent platform designed for complex environments, unmanned swarm combat platforms possess advantages that single UAV combat units lack. Environmentally-based collaborative perception and understanding technology is fundamental for autonomous decision-making by unmanned combat platforms. However, with the rapid development of electronic attack technology, countries worldwide have successively developed sophisticated camouflage technologies such as electromagnetic silence, electronic feints, and electronic deception. Simultaneously, due to the increasing complexity of the electromagnetic environment and the influence of terrain, features, and electromagnetic waves, swarms encounter regular signal fading when sensing target signals, placing increasingly stringent demands on signal detection theory and estimation methods for UAV receiving equipment. Energy detection, as a widely applicable signal detection method, can accurately detect the presence or absence of target signals by making a binary judgment on their existence, thus providing reliable prior information for swarm target detection. Therefore, researching weak signal energy detection technology in complex electromagnetic environments has become an important issue for unmanned combat platforms in the new era. However, research on improving the energy detection performance of systems is relatively limited. Among them, LIU X, LI F, NA Z's 2017 IEEE Access paper, "Optimal resource allocation insimultaneous cooperative spectrum sensing and energy harvesting for multichannel cognitive radio," proposed a cooperative spectrum sensing and energy harvesting model to enhance the energy detection performance of multichannel cognitive radio communication systems. FAN R, JIANG H's 2010 IEEE Transactions on Wireless Communications paper, "Optimal multi-channel cooperative sensing in cognitive radio networks," investigated an optimal multi-channel spectrum sensing mechanism to maximize the throughput of network users on all sub-channels while maintaining the energy detection probability of each sub-channel above a preset threshold. LIU X, BI G, JIA M et al.'s 2013 Radio Science paper, "Joint optimization of sensing threshold and transmission power in wideband cognitive radio with energy detection," jointly considered power allocation and sensing threshold optimization problems, aiming to improve the throughput of the sensing network while ensuring the system's false alarm probability and energy detection performance. However, there is currently no research specifically addressing the energy detection problem of UAV swarms.Therefore, this invention proposes for the first time a multi-UAV perception resource scheduling strategy based on energy detection. Several conditions need to be considered for this problem: First, in real-world scenarios, UAV hardware limitations prevent omnidirectional perception, and directional antennas suffer from main lobe and sidelobe limitations, resulting in limited antenna coverage. Second, when a UAV performs signal detection on a target, the received signal strength of the directional antenna is related to the target distance. Thus, the joint scheduling problem of perception network resources based on multiple UAVs involves multiple perception nodes, multiple observation targets, multiple types of resource scheduling, and the selection of various management methods. This problem contains multiple optimization variables and is a highly coupled, large-scale, complex optimization problem. Summary of the Invention
[0003] To address the above problems, the present invention aims to overcome the shortcomings of existing technologies by providing a joint scheduling method for UAV swarm perception resources based on the alternating direction penalty method. For multiple targets to be perceived, various perception indicators, optimization variables, and environmental constraints are analyzed to construct a deployment and beam direction optimization problem for multi-sensor UAVs. Then, based on the augmented Lagrangian function and the alternating direction penalty algorithm, a joint scheduling optimization algorithm for perception resources is designed to optimize target perception performance under resource constraints in the frequency domain, airspace, and beam domain of each UAV.
[0004] A joint scheduling method for UAV swarm perception resources based on alternating direction penalty includes the following steps: First, the number of UAVs in the swarm, the UAV perception frequency band range, the beamwidth of the UAV directional antenna, the deployable area of the UAVs, the distance constraints between UAVs, the number of targets to be sensed, the operating frequency of the targets to be sensed, the estimated position of the targets to be sensed, and environmental parameters are set. Then, a multi-UAV joint perception model based on signal energy detection is constructed, and a multi-UAV perception resource joint scheduling problem is proposed. Finally, the original problem is transformed into an equivalent problem, and an iterative algorithm based on alternating direction penalty is designed to efficiently solve the problem.
[0005] A joint scheduling scheme for sensing resources of UAV swarms based on alternating direction penalty method includes the following steps:
[0006] A method for joint scheduling of sensing resources for unmanned aerial vehicle (UAV) swarms based on alternating direction penalty, characterized by the following steps:
[0007] S1: Derive the energy detection model for multiple UAVs targeting;
[0008] S2: Construct the joint scheduling problem of UAV swarm sensing resources based on the energy detection model;
[0009] S3: By performing an equivalent transformation of the problem, problem P1 is transformed into an equivalent problem P2;
[0010] S4: The problem is solved using an alternating direction penalty algorithm to obtain the optimal deployment location and antenna orientation of the drone swarm.
[0011] A method for joint scheduling of UAV swarm perception resources based on alternating direction penalty, characterized in that S1 specifically comprises:
[0012] In a three-dimensional plane, suppose the number of drones is M and the position of the i-th drone is q. i =[x i y i H] T Assume the number of targets to be sensed is K, and the position of the k-th target is q. t，k =[x t，k y t，k , z t，k ] T The distance, horizontal angle, and pitch angle from UAV i to target k can be expressed as follows:
[0013] d i，k =||q t，k -q i ||2
[0014] ψ i，k =arctan(x t，k -x i y t，k -y i )
[0015]
[0016] The antenna gain of UAV i receiving the signal from target k is:
[0017]
[0018] The signal-to-noise ratio of the signal received by UAV i from target k is:
[0019]
[0020] Where θ represents the antenna main lobe width, β0 is the channel gain when the distance is 1, and P... k For the target signal transmission power, N t The number of directional antenna elements, σ 2 This represents noise power.
[0021] The probability of UAV i detecting the energy of target k is:
[0022]
[0023] Where Q represents the right-tail function of the standard normal distribution, N is the number of received signal samples, and P...fa The false alarm probability of the sensing system.
[0024] Assume the spectrum sensing range of UAV i is (f i min f i max If the probability of UAV i detecting the signal of target k at frequency point f is:
[0025]
[0026] For target k at frequency f, the joint detection probability of signals from M drones is:
[0027]
[0028] A method for joint scheduling of UAV swarm perception resources based on alternating direction penalty, characterized in that S2 constructs a joint scheduling problem for UAV swarm perception resources, specifically as follows:
[0029] The detection probability and function of the system are defined as follows:
[0030]
[0031] The joint scheduling problem of perception resources aims to maximize the sum of the detection probabilities of all targets by a swarm of drones, as follows:
[0032] P1:
[0033]
[0034]
[0035]
[0036]
[0037] in, Let S be the set of drone locations, D be the deployable area, and S be the location of the drones. l To constrain the minimum distance between the UAV and the target, R l Minimum distance constraint between drones.
[0038] A method for joint scheduling of UAV swarm perception resources based on alternating direction penalty is characterized by S3 decoupling the non-convex constraints that are difficult to handle in problem P1 by equivalent transformation, transforming it into an equivalent problem P2, specifically:
[0039] The minimum distance constraint between the UAV and the perceived target in the model is defined here:
[0040]
[0041] Define column vector b as all b i，k The long vector formed, the feasible region D of this vector. b ={b i，k |||b i，k ||2>s l The minimum distance constraint between the UAV and the perceived target can be equivalently written as the following constraint:
[0042] A1q-q t =b
[0043] Where A1 is the coefficient matrix in the transformation process, q t For all q t，k The vector formed.
[0044] For the collision constraints between drones in the model, define the following variables:
[0045]
[0046] Next, define column vector c as all c i，j The long vector formed, the feasible region D of this vector. c ={c ij |||c ij ||2>R l The collision constraints between drones can be equivalently written as follows:
[0047] A2q=c
[0048] Where A2 is the coefficient matrix corresponding to the transformation process.
[0049] Therefore, problem P1 is equivalent to the following problem:
[0050] P2:
[0051] stA1q-q t =b
[0052] A2q=c
[0053]
[0054] b∈D b ,
[0055] c∈D c ,
[0056]
[0057] A joint scheduling method for UAV swarm perception resources based on the alternating direction penalty algorithm is proposed. S4 uses the alternating direction penalty algorithm to solve the problem, obtaining the optimal deployment positions and antenna directions for multiple UAVs. Specifically:
[0058] For problem P2, its augmented Lagrangian function is expressed as follows:
[0059]
[0060] In the formula, χ and μ are Lagrange multiplier vectors, and ρ1 and ρ2 are penalty factors.
[0061] S4.1. Update variable b, i.e., solve the following problem:
[0062]
[0063] stb∈D b .
[0064] Based on the special composition of variable b, the problem is decomposed into M×K subproblems that are solved in parallel. The v-th subproblem is:
[0065]
[0066] st||b v ||2≥S l .
[0067] The closed-form solution to the above problem can be expressed as:
[0068]
[0069] S4.2. Update variable c, i.e., solve the following problem:
[0070]
[0071] stc∈D c .
[0072] Based on the special structure of variable c, the problem is decomposed into The v-th subproblem is solved in parallel.
[0073]
[0074] st||c v ||2≥R l .
[0075] The closed-form solution to the above problem can be expressed as:
[0076]
[0077] S4.3. Update variable q, i.e., solve the following problem:
[0078]
[0079] stq i ∈D q .
[0080] The gradient projection algorithm is used to optimize the solution of the above problem.
[0081] S4.4. Update the variable ψ, that is, solve the following problem:
[0082]
[0083]
[0084] The gradient descent algorithm is used to optimize the solution of this problem.
[0085] S4.5. Update the penalty factors ρ1 and ρ2 as follows:
[0086]
[0087]
[0088] In the formula, 0 < ∈ 1，b ,∈ 1，c <1, ∈ 2，b ,∈ 2，c Greater than but close to 1. This represents the residuals corresponding to the constraint terms in the model.
[0089] S4.6: Update the Lagrange multiplier vectors χ and μ as follows:
[0090]
[0091]
[0092] In the formula, also, and They represent vectors respectively and The absolute value of the largest element in the set. and All are sufficiently large positive numbers.
[0093] S4.7: Stopping Condition: The variables can be continuously updated according to the above steps until the following condition is met:
[0094] ε l+1 =||A1q l+1 -qt -b l+1 ||2+||A2q l+1 -c l+1 ||2≤η
[0095] Where η is the iteration threshold.
[0096] The beneficial effects of this invention are as follows: The multi-UAV formation sensing resource joint scheduling method described in this invention first constructs a multi-UAV joint sensing model based on signal energy detection and proposes a multi-UAV sensing resource joint scheduling problem. Then, the original problem is transformed into a new problem with constraints, expressed in augmented Lagrangian function form. Finally, a solution method based on alternating direction penalty is designed to achieve effective scheduling of multi-UAV formation sensing resources. This method achieves optimal target sensing performance under resource constraints in the frequency domain, airspace domain, and beam domain of each UAV. Attached Figure Description
[0097] Figure 1 This is a system model diagram of the method used in the embodiments of the present invention.
[0098] Figure 2 This is a diagram of the perception network deployment in the embodiments of the present invention.
[0099] Figure 3 This is a performance diagram of the perceived resource scheduling in the embodiments of the present invention. Detailed Implementation
[0100] The present invention will become more apparent from the following detailed description with reference to the accompanying drawings.
[0101] A method for joint scheduling of sensing resources for unmanned aerial vehicle (UAV) swarms based on alternating direction penalty, characterized by the following steps:
[0102] S1: Derive the energy detection model for multiple UAVs targeting;
[0103] S2: Construct the joint scheduling problem of UAV swarm sensing resources based on the energy detection model;
[0104] S3: By performing an equivalent transformation of the problem, problem P1 is transformed into an equivalent problem P2;
[0105] S4: The problem is solved using an alternating direction penalty algorithm to obtain the optimal deployment location and antenna orientation of the drone swarm.
[0106] S1 specifically refers to:
[0107] Figure 1 The scenario and system model shown are in a three-dimensional plane. Assume the number of drones is M and the position of the i-th drone is q. i =[xi y i H] T Assume the number of targets to be sensed is K, and the position of the k-th target is q. t，k =[x t，k y t，k , z t，k ] T The distance, horizontal angle, and pitch angle from UAV i to target k can be expressed as follows:
[0108] d i，k =||q t，k -q i ||2
[0109] ψ i，k =arctan(x t，k -x i y t，k -y i )
[0110]
[0111] The antenna gain of UAV i receiving the signal from target k is:
[0112]
[0113] The signal-to-noise ratio of the signal received by UAV i from target k is:
[0114]
[0115] Where θ represents the antenna main lobe width, β0 is the channel gain when the distance is 1, and P... k For the target signal transmission power, N t The number of directional antenna elements, σ 2 This represents noise power.
[0116] The probability of UAV i detecting the energy of target k is:
[0117]
[0118] Where Q represents the right-tail function of the standard normal distribution, N is the number of received signal samples, and P... fa The false alarm probability of the sensing system.
[0119] Assume the spectrum sensing range of UAV i is (f i min f i max If the probability of UAV i detecting the signal of target k at frequency point f is:
[0120]
[0121] For target k at frequency f, the joint detection probability of signals from M drones is:
[0122]
[0123] S2 addresses the joint scheduling problem of UAV swarm perception resources, specifically:
[0124] The detection probability and function of the system are defined as follows:
[0125]
[0126] The joint scheduling problem of perception resources aims to maximize the sum of the detection probabilities of all targets by a swarm of drones, as follows:
[0127] P1:
[0128]
[0129]
[0130]
[0131]
[0132] in, Let S be the set of drone locations, D be the deployable area, and S be the location of the drones. l To constrain the minimum distance between the UAV and the target, R l Minimum distance constraint between drones.
[0133] S3 decouples the non-convex constraints that are difficult to handle in problem P1 by performing an equivalent transformation, transforming it into an equivalent problem P2, specifically:
[0134] The minimum distance constraint between the UAV and the perceived target in the model is defined here:
[0135]
[0136] Define column vector b as all b i，k The long vector formed, the feasible region D of this vector. b ={b i，k |||b i，k ||2>s l The minimum distance constraint between the UAV and the perceived target can be equivalently written as the following constraint:
[0137] A1q-q t =b
[0138] Where A1 is the coefficient matrix in the transformation process, q t For all q t，k The vector formed.
[0139] For the collision constraints between drones in the model, define the following variables:
[0140]
[0141] Next, define column vector c as all c i，j The long vector formed, the feasible region D of this vector. c ={c ij |||c ij ||2>R l The collision constraints between drones can be equivalently written as follows:
[0142] A2q=c
[0143] Where A2 is the coefficient matrix corresponding to the transformation process.
[0144] Therefore, problem P1 is equivalent to the following problem:
[0145] P2:
[0146] stA1q-q t =b
[0147] A2q=c
[0148]
[0149] b∈D b ,
[0150] c∈D c ,
[0151]
[0152] S4 employs an alternating direction penalty algorithm to solve the problem, obtaining the optimal deployment locations and antenna pointing for multiple UAVs. Specifically:
[0153] For problem P2, its augmented Lagrangian function is expressed as follows:
[0154]
[0155] In the formula, χ and μ are Lagrange multiplier vectors, and ρ1 and ρ2 are penalty factors.
[0156] S4.1. Update variable b, i.e., solve the following problem:
[0157]
[0158] stb∈D b .
[0159] Based on the special composition of variable b, the problem is decomposed into M×K subproblems that are solved in parallel. The v-th subproblem is:
[0160]
[0161] st||b v ||2≥S l .
[0162] The closed-form solution to the above problem can be expressed as:
[0163]
[0164] S4.2. Update variable c, i.e., solve the following problem:
[0165]
[0166] stc∈D c .
[0167] Based on the special structure of variable c, the problem is decomposed into The v-th subproblem is solved in parallel.
[0168]
[0169] st||c v ||2≥R l .
[0170] The closed-form solution to the above problem can be expressed as:
[0171]
[0172] S4.3. Update variable q, i.e., solve the following problem:
[0173]
[0174] stq i ∈D q .
[0175] The gradient projection algorithm is used to optimize the solution of the above problem.
[0176] S4.4. Update the variable ψ, that is, solve the following problem:
[0177]
[0178]
[0179] The gradient descent algorithm is used to optimize the solution of this problem.
[0180] S4.5. Update the penalty factors ρ1 and ρ2 as follows:
[0181]
[0182]
[0183] In the formula, 0 < ∈ 1，b ,∈ 1，c <1, ∈ 2，b ,∈ 2，c Greater than but close to 1. This represents the residuals corresponding to the constraint terms in the model.
[0184] S4.6: Update the Lagrange multiplier vectors χ and μ as follows:
[0185]
[0186]
[0187] In the formula, also, and They represent vectors respectively and The absolute value of the largest element in the set. and All are sufficiently large positive numbers.
[0188] S4.7: Stopping Condition: The variables can be continuously updated according to the above steps until the following condition is met:
[0189] ε l+1 =||A1q l+1 -q t -b l+1 ||2+||A2q l+1 -c l+1 ||2≤η
[0190] Where η is the iteration threshold.
[0191] Figure 2This is a perception network deployment diagram for an example with 5 drones. Specific parameter settings: the minimum distance between a drone and the target is 500 meters, and the minimum distance between drones is 50 meters; the target coordinates are (2000, 1580, 500), (3400, 1300, 500), and (2600, 2600, 500); the frequency band sets are {4000, 5000, ..., 12000, 13000}, {1000, 1500, ..., 4000, 4500}, and {200, 2200, 4200, ..., 16200, 18200}; and the transmit power is 4 × 10⁻⁶. -6 dBm, 4×10 -6 dBm, 4×10 -6 dBm; the spectrum sensing ranges of the five drones are [150, 400], [500, 4000], [2000, 4500], [1500, 5000], and [4000, 10000], respectively.
[0192] Figure 2 For the performance comparison of the deployed solution in the example with the baseline deployment solution, see [link to example]. Figure 3 . Figure 3 A performance comparison chart of sensing node deployment is provided, where the vertical axis represents the detection performance and speed of the sensing network, and the horizontal axis represents the number of sensing nodes. Figure 3 It can be seen that as the number of sensing nodes increases, the sensing performance gradually improves. Furthermore, it can be observed that, for the same network scenario, the system performance of the method described in this invention is superior to that of other solutions.
[0193] Performance simulations demonstrate that the method of this invention achieves effective scheduling of sensing resources for multi-UAV formations. This method optimizes target sensing performance under resource constraints in the frequency, airspace, and beam domains of each UAV. It is foreseeable that the method of this invention will be well-suited to future wireless sensing technologies, thereby improving the overall performance of sensing networks.
[0194] This invention is not limited to the specific embodiments described above. Those skilled in the art can implement this invention using various other specific embodiments based on the content disclosed herein. Therefore, any design that adopts the design structure and concept of this invention with some simple changes or modifications falls within the protection scope of this invention.
Claims
1. A method for joint scheduling of UAV swarm perception resources based on alternating direction penalty, characterized in that, Includes the following steps: First, the following parameters are defined: the number of UAVs in the swarm, the UAV sensing frequency band range, the beamwidth of the UAV directional antenna, the deployable area of the UAVs, the distance constraints between UAVs, the number of targets to be sensed, the operating frequency of the targets to be sensed, the estimated position of the targets to be sensed, and environmental parameters. Then, a multi-UAV joint sensing model based on signal energy detection is constructed, and a multi-UAV sensing resource joint scheduling problem is proposed. Finally, the original problem is transformed into an equivalent problem, and an iterative algorithm based on alternating direction penalties is designed to efficiently solve the problem. The specific algorithm includes the following steps: S1: Derive the energy detection model for multiple UAVs targeting; S2: Construct the joint scheduling problem of UAV swarm sensing resources based on the energy detection model; S3: By performing an equivalent transformation of the problem, problem P1 is transformed into an equivalent problem P2; S4: The problem is solved using an alternating direction penalty algorithm to obtain the optimal deployment location and antenna orientation of the drone swarm; S1 specifically refers to: In a three-dimensional plane, assume the number of drones is... , No. The location of the drone is Assume the number of targets to be perceived is , No. The location of the target is Then the drone To the target The distance, horizontal angle, and pitch angle are expressed as follows: drones Received target The antenna gain of the signal is: drones Receive target The signal-to-noise ratio of the signal is: in Indicates the main lobe width of the antenna, Channel gain at a distance of 1 For the target signal transmission power, For the number of directional antenna elements, Noise power; Get drone For the target The energy detection probability is: in Represented as the right-tail function of the standard normal distribution, The number of received signal samples, The false alarm probability of the sensing system; Assuming a drone The spectrum sensing range is Then drone For the target At frequency The signal detection probability on is: For the goal At frequency superior, The probability of joint detection of signals from multiple drones is: ； S2 addresses the joint scheduling problem of UAV swarm perception resources, specifically: The detection probability and function of the system are defined as follows: The joint scheduling problem of perception resources aims to maximize the sum of the detection probabilities of all targets by a swarm of drones, as follows: in, For drone location collection, For the deployable area, Minimum distance constraint between the UAV and the target. Minimum distance constraint between drones; S3 decouples the non-convex constraints that are difficult to handle in problem P1 by performing an equivalent transformation, transforming it into an equivalent problem P2, specifically: The minimum distance constraint between the UAV and the perceived target in the model is defined here: Define column vectors For all The long vector formed, and the feasible region of this vector. The minimum distance constraint between the UAV and the perceived target is equivalently written as the following constraint: in, This is the coefficient matrix during the transformation process. For all The vector formed; For the collision constraints between drones in the model, define the following variables: Next, define column vectors. For all The long vector formed, and the feasible region of this vector. The collision constraints between drones are equivalently written as follows: in, This is the coefficient matrix corresponding to the transformation process; Problem P1 is equivalently transformed into the following problem: s.t. ； S4 uses an alternating direction penalty algorithm to solve the problem and obtain the optimal deployment location and antenna pointing of multiple UAVs; Step S4 specifically includes: For problem P2, its augmented Lagrangian function is expressed as follows: In the formula, χ and µ Let Lagrange multiplier vectors be used. , As a penalty factor; S4.
1. Update variables That is, to solve the following problem: According to variables Break down the problem into The subproblems are solved in parallel; among them, the first subproblem is solved in parallel. The sub-problem is: The closed-form solution to the above problem is expressed as: S4.
2. Update variables That is, to solve the following problem: According to variables Break down the problem into The subproblems are solved in parallel; among them, the first subproblem is solved in parallel. The sub-problem is: The closed-form solution to the above problem is expressed as: S43. Update variables That is, to solve the following problem: . The gradient projection algorithm is used to optimize and solve the above problem; S4.
4. Update variables That is, to solve the following problem: The gradient descent algorithm is used to optimize the solution of this problem; S4.
5. Update the penalty factor as follows: , : In the formula, 0 < , <1, , Greater than but close to 1; These are the residuals corresponding to the constraint terms in the model; S4.6: Update the Lagrange multiplier vector and as follows: In the formula, ;also, and They represent vectors respectively and The absolute value of the largest element in the set; and All are sufficiently large positive numbers; S47. Stopping condition: Update the variables continuously according to the above steps until the following condition is met: Where η is the iteration threshold.