Unmanned cluster collaborative navigation method and system based on distributed cluster self-organizing model

Abstract

The invention relates to the technical field of unmanned clusters, and designs an unmanned cluster collaborative navigation method and system based on a distributed cluster self-organizing model. The method comprises the following steps: 1, acquiring a task environment, and preprocessing the task environment; 2, expressing the unmanned cluster platform by adopting characteristics; step 3, carrying out cluster inside and outside interaction framework criterion design based on a Boedes ring on the unmanned cluster platform expressed by the characteristics in the step 2, so that each unmanned platform adjusts the motion state of the unmanned platform according to the motion state of the surrounding platform and the external environment information sensed by a carried sensor; 4, on the basis of the interaction framework criterion designed in the step 3, each unmanned aerial vehicle uses a model prediction controller carried by the unmanned aerial vehicle to calculate the control input of the next moment on line, and the flight state is updated by using a fourth-order Runge-Kutta method; and each unmanned aerial vehicle performs online updating on the flight state of the surrounding unmanned platform by using a sensor carried by the unmanned aerial vehicle and returns to execute the step 3 until the unmanned aerial vehicle arrives at the target point. The method and the device are used for solving the problem of realizing a collaborative navigation behavior of an overall level of a cluster under the condition of reducing communication dependency.

Description

Technical Field

[0001] The invention relates to the technical field of unmanned clusters, and designs an unmanned cluster collaborative navigation method and a system thereof based on a distributed cluster self-organizing model. Background Art

[0002] In recent years, with the rapid development of drone technology, unmanned swarms have become an important technical field. Especially under the conditions of information warfare, drone swarms play an increasingly important role in combat. Unmanned swarms can be used to perform various tasks, such as surveillance, search, search and rescue, etc. In order to meet the diverse mission requirements, unmanned swarms must be able to effectively adapt to complex and dynamic external environments. Existing solutions usually use large and medium-sized unmanned aerial platforms to perform tasks, but with increasingly stringent air defense measures, it is difficult to penetrate deep behind enemy lines. In contrast, small unmanned platforms are becoming a sharp sword behind enemy lines on the modern battlefield in a large-scale form due to their low cost and low detectability. However, how to achieve safe and autonomous collaborative flight from the mission starting point to the predetermined target point in a complex environment is still a problem that needs to be solved.

[0003] For the above problems, the existing methods can be divided into three types: centralized, distributed, and distributed, according to the form of information interaction and processing with the ground station during flight. Considering that it is difficult to ensure communication with the rear after the cluster is released deep into the enemy area, it is difficult to ensure the real-time transmission of instructions even if a communication relay is established through an aerial platform. Therefore, local coordination methods based on distributed structures are becoming mainstream. It can be seen that the operating environment of unmanned clusters can be regarded as a communication-restricted environment. The existing group navigation methods mainly rely on communication networks for frequent information exchange to achieve stable navigation behavior. However, this dependence brings challenges to the realization of coordinated and cooperative behaviors in environments with limited communication and rich obstacles. How to achieve collaborative navigation behavior at the overall level of the cluster under the condition of reducing communication dependence is a technical problem that urgently needs to be solved in this field. Summary of the invention

[0004] The present invention provides an unmanned cluster collaborative navigation method based on a distributed cluster self-organizing model, which is used to solve the problem of realizing collaborative navigation behavior at the overall level of the cluster under the condition of reducing communication dependence.

[0005] The invention provides an unmanned cluster cooperative navigation system based on a distributed cluster self-organizing model, which is used to realize an unmanned cluster cooperative navigation method based on a distributed cluster self-organizing model.

[0006] The present invention is achieved through the following technical solutions:

[0007] An unmanned cluster collaborative navigation method based on a distributed cluster self-organizing model, the method comprising the following steps:

[0008] Step 1: Obtain the task environment and preprocess it;

[0009] Step 2: Describe the unmanned cluster platform using characteristics;

[0010] Step 3: For the unmanned swarm platforms characterized in step 2, design the interaction framework criteria inside and outside the swarm based on the Boyd ring, so that each unmanned platform can adjust its own motion state according to the motion state of the surrounding platforms and external environment information perceived by the sensors on board;

[0011] Step 4: Based on the interactive framework criteria designed in step 3, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and uses the fourth-order Runge-Kutta method to update the flight status; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute step 3 until it reaches the target point.

[0012] Furthermore, the step 1 specifically includes the following steps:

[0013] Step 1.1: Based on the existing elevation DEM data or obtained through digital computer simulation, the obstacle shape is approximated by its circumscribed circle, and its radius is expressed as T r ; A cylindrical model with randomly distributed positions is used to simulate the obstacles that unmanned swarms may encounter during low-altitude flight;

[0014] Step 1.2: For non-convex obstacles that may appear in the environment, the unmanned swarm converts them into convex obstacles by adjusting the flight altitude and further approximates them using the circumscribed cylinder in step 1.1.

[0015] Furthermore, the step 2 specifically includes the following steps:

[0016] Step 2.1: Simplify the platform kinematic model;

[0017] Specifically, step 2.1 is as follows: assuming that the unmanned clusters participating in the task are labeled as N={1,...,N}, the following simplified motion model is adopted for each unmanned platform i, expressed as formula (1):

[0018]

[0019] Among them, x i ,y i ,h i is the three-dimensional coordinate of the unmanned platform i, and are the control inputs of the position holding, heading holding and altitude holding autopilots of platform i respectively; τ v τ ψ, τ h and τ λ are the time constants of the three autopilots;

[0020] After determining the time constant, the model is simplified to a second-order integrator model, as shown in formula (2):

[0021]

[0022] Among them, p i =(x i ,y i ,h i ) T represents the position vector, represents the velocity vector,

[0023] represents the control input of each platform; at each time step k, set v min ≤v i (k)≤v max ,u min ≤u i (k)≤u max ; Based on the control input, the fourth-order Runge-Kutta method is used to update the motion state of each unmanned platform at each moment;

[0024] Step 2.2: Based on the simplified platform kinematic model in step 2.1, divide the platform task roles;

[0025] Specifically, the step 2.2 includes dividing the unmanned platform into an informed member or an uninformed member according to whether the unmanned platform has global target information;

[0026] Among them, the informed unmanned platform has global navigation information and can guide the uninformed unmanned platform to fly to the target;

[0027] An uninformed unmanned platform can only perceive the local state changes of its neighbors and use this information to determine its state in the next time step;

[0028] During the initialization process, informed individuals are randomly assigned to the population according to the specified proportion Δ.

[0029] Furthermore, the step 3 specifically includes the following steps:

[0030] Step 3.1: Each unmanned platform observes the motion state of the unmanned platforms in the adjacent airspace based on the sensors it carries and obtains its predicted motion state at the current moment and in the future with the help of the corresponding filtering prediction algorithm;

[0031] Step 3.2: Each unmanned platform makes local decisions based on the state observations of adjacent platforms and environmental observation information to achieve dynamic adaptive adjustment of key parameters;

[0032] Step 3.3: After obtaining the optimized parameter set Then, the motion model in step 2.1 is used to predict T p The future state of the time step, serving as a reference trajectory A decoupled MPC controller is used for the horizontal and vertical directions, where is the reference trajectory before optimization based on the current state; it provides the final control input and is then used to update the state at the next time step.

[0033] Furthermore, the step 3.1 is specifically as follows: step 3.1.1: at the current speed v i (k) and Establish a reference coordinate system for the reference direction, define counterclockwise rotation as positive; define the field of view as where α max for the maximum viewing angle; then Represents the local perception range R of unmanned platform i sen The detectable individuals within the qth angular interval; divide the field of view into Q equal parts; then the individuals within the qth angular interval It can be expressed as formula (3), where ∠p ij (k) represents the relative angle between unmanned platform i and unmanned platform j, ||p ij (k)||2 represents the relative distance between unmanned platform i and unmanned platform j;

[0034]

[0035] Formula (4) is used to calculate the nearest perceptible individual in each subinterval and form a neighbor set Specifically,

[0036]

[0037] Since the relative positions between unmanned platforms will not mutate, when platform i detects platform j, it will create and store the state information structure of platform j locally;

[0038] Step 3.1.2: If the current unmanned platform calculates that it is at the front of the cluster flight based on the relative position and relative speed information, if the unmanned platform is currently in the role of non-informed, it will apply to the informed unmanned platform in the group to obtain target information through the communication network to lead the cluster to fly towards the target. Once it is no longer in a dominant position, it will automatically switch to the follower role; the information flow only includes the target and the corresponding sign information, and does not involve the exchange of real-time status information; if the current unmanned platform is in the role of informed, skip this step;

[0039] Step 3.1.3: If the current unmanned platform detects an external obstacle and needs to take obstacle avoidance action, define the critical neighbor set under the obstacle detection condition in formula (5): Where <·> represents the angle between two vectors, and λ represents the critical neighbor angle range;

[0040]

[0041] Key Neighbor Set is the set of sensing neighbors When the critical neighbor set of an individual is an empty set, it uses the communication network to obtain global target information from the locally grouped informant platform and acts as an "informant" to continue flying toward the target and restores to its original identity after returning to the cluster.

[0042] If the current unmanned platform does not detect any obstacles, it continues to use the perceived neighbor set. Perform subsequent state parameter calculation updates.

[0043] Further, the step 3.2 includes the following steps:

[0044] Step 3.2.1: If the current flight status triggers the optimization flag and reaches the optimization interval T0, go to step 3.2.2;

[0045] If the current flight status is safe, the current parameter set is maintained and the flight continues. When the fixed optimization interval T1 is reached, the process also goes to step 3.2.2.

[0046] At other times, the flight status is updated according to step 3.3;

[0047] Step 3.2.2: Design of interaction criteria between members within the cluster;

[0048] Step 3.2.3: Based on the interaction criteria in step 3.2.2, further define the information interaction items between the unmanned platform and the environment;

[0049] Step 3.2.4: Summarize various control items to obtain the overall control input, as shown in formula (20);

[0050] Step 3.2.5: After obtaining the virtual control input, combine the motion equation to predict its next T p Step status in Represents the predicted state at each time step; a local performance evaluation function is defined for each unmanned platform based on its own and its neighbors’ predicted states;

[0051] Step 3.2.6: Based on the above local optimization indicators, the local optimization target of each unmanned platform is defined as formula (26):

[0052]

[0053] Among them, γ∈(0,1) is the state decay factor that characterizes the importance between different times.

[0054] Furthermore, the step 3.2.2 is specifically as follows:

[0055] Step 3.2.2.1: Calculate the relative velocity alignment term to prevent the cluster from dispersing due to local velocity inconsistency during flight, which can be expressed as formula (6);

[0056]

[0057] First, the relative distance between the neighboring member and the unmanned platform i is calculated as Get the normalized vector It is used as a local weight to calculate the total alignment term, as shown in formula (7):

[0058]

[0059] in, is a constant that needs to be determined. It reflects the influence of the neighbors of unmanned platform i on its speed at the current time k;

[0060] Step 3.2.2.2: Calculate member local cohesion to ensure perception range The members within are kept close to each other as shown in formula (8),

[0061]

[0062] Among them, d coh It is the minimum distance between unmanned platforms to produce cohesion. is the unit vector pointing from unmanned platform i to unmanned platform j; the local overall cohesion term is calculated using the relative distance weighted method as shown in formula (9):

[0063]

[0064] in, is a cohesion term coefficient that needs to be determined dynamically;

[0065] Step 3.2.2.3: Calculate the local repulsion term to ensure that members always maintain the minimum safe distance; the repulsion effect is defined as formula (10)

[0066]

[0067] Among them, d rep is the maximum interaction range where unmanned platforms begin to exclude each other; is the avoidance direction from unmanned platform j to unmanned platform i; the total repulsion term between unmanned platform i and its adjacent individuals is expressed as formula (11):

[0068]

[0069] in, is the weight parameter of the exclusion term;

[0070] Step 3.2.2.4: In order to maintain stability during flight, a speed maintenance term is added, as shown in formula (12)

[0071]

[0072] in, is the reference speed; is the coefficient, is the average flight direction of the aircraft, as shown in formula (13):

[0073]

[0074] in, is the unit vector of unmanned platform j;

[0075] definition To characterize all interactions within the swarm, as shown in formula (14):

[0076]

[0077] Further, the step 3.2.3 includes the following steps:

[0078] Step 3.2.3.1: Design the goal-oriented interaction term for the informed individuals of global navigation information, as shown in formula (15);

[0079]

[0080] Among them, p g is the target coordinate, C taris the attraction term gain, Ω is the set of informed unmanned platforms, and this step is skipped for uninformed individuals;

[0081] Step 3.2.3.2: Assume that the field of view of unmanned platform i is α i (k); After the obstacle enters the detection range, its boundary intersection point P can be calculated l =(x l ,y l ), P r =(x r ,y r ) ; use the law of cosines to calculate the angle θ it occupies i,o (k) Assume that unmanned platform i detects an obstacle at time k Then the feasible angle of unmanned platform i is expressed as formula (16):

[0082]

[0083] The feasible angle A i (k) is divided into W sub-intervals A i,w (k); Evaluate the angular difference between the center line of each subinterval and the heading of the current local neighbor, and select the subinterval with the smallest relative angular difference as the local obstacle avoidance direction; calculated as formula (17);

[0084]

[0085] When calculating the obstacle avoidance direction, the key neighbor set is used The external obstacle avoidance behavior term is defined as shown in formula (18):

[0086]

[0087] in, and Represents the boundary angle of each sub-interval; By analyzing the navigation direction, the expected navigation item of the unmanned platform i is defined, where represents the gain factor; this factor allows to adapt its heading by adaptively modifying its navigation gain, thus adapting to the different requirements of the task in different situations; similarly, define To represent the sum of the interactions between the unmanned platform and the external environment, as shown in formula (19):

[0088]

[0089] Further, the step 3.2.6 includes the following steps,

[0090] Step 3.2.6.1: Initialize the meta-heuristic optimization algorithm, including the feasible range of each parameter, the number of populations, the maximum number of iterations, and the set of factors to be optimized is expressed as Represents the weights of each item in the virtual control input;

[0091] Step 3.2.6.2: Substitute the set of variables to be optimized into formula (26) to calculate the function value of the current individual motion state in the future period of time;

[0092] Step 3.2.6.3: Perform multiple iterations of optimization to obtain the control parameter set at the future moment

[0093] An unmanned cluster collaborative navigation system based on a distributed cluster self-organizing model, the system uses the unmanned cluster collaborative navigation method based on the distributed cluster self-organizing model as described above, and the system includes:

[0094] Preprocessing module: obtain the task environment and preprocess it;

[0095] Feature description module: describes the unmanned cluster platform using features;

[0096] Cluster internal and external interaction framework rule design module: Based on the unmanned cluster platform described by characteristics, the cluster internal and external interaction framework rule design based on Boyd ring is carried out, so that each unmanned platform can adjust its own motion state according to the motion state of the surrounding platforms and external environment information perceived by the sensors on board;

[0097] Flight status update module: Based on the interactive framework criterion, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and uses the fourth-order Runge-Kutta method to update the flight status; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute the interactive framework criterion design module inside and outside the cluster until it reaches the target point.

[0098] The beneficial effects of the present invention are:

[0099] The present invention can realize the coordinated autonomous navigation behavior of the cluster in static obstacle and dynamic obstacle environments.

[0100] The self-organizing model in the present invention has good scalability and can reduce the dependence on the communication network by means of local perception and event-triggered communication.

[0101] The asynchronous model parameter online adjustment mechanism in the present invention has high computational real-time performance and can give individuals in the cluster higher autonomy so as to adapt to a changing external environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 It is a flow chart of the method of the present invention.

[0103] Figure 2It is a schematic diagram of determining the local perception neighbor set (left) and the key neighbor set (right) of the present invention.

[0104] Figure 3 It is a three-dimensional schematic diagram of the collaborative navigation of the present invention.

[0105] Figure 4 It is a three-dimensional schematic diagram of collaborative navigation of the traditional Flocking model.

[0106] Figure 5 It is a three-dimensional schematic diagram of MPC collaborative navigation.

[0107] Figure 6 It is a three-dimensional schematic diagram of parameter adjustment collaborative navigation.

[0108] Figure 7 It is a schematic diagram of the relative distance curve between cluster members of the present invention and different methods.

[0109] Figure 8 It is a schematic diagram of the relative distance curves between cluster members and environmental obstacles of the present invention and different methods.

[0110] Fig. 9 It is a schematic diagram of the speed change curve of cluster members according to the present invention and different methods.

[0111] Fig.10 It is a schematic diagram of the angular velocity variation curves of cluster members according to the present invention and different methods.

[0112] Fig.11 This is a screenshot of the motion state of the present invention in a dynamic environment at t=30s.

[0113] Fig.12 This is a screenshot of the motion state of the present invention in a dynamic environment at t=70s.

[0114] Fig.13 This is a screenshot of the motion state of the present invention in a dynamic environment at t=110s.

[0115] Fig.14 This is a screenshot of the motion state of the present invention in a dynamic environment at t=150s.

[0116] Fig.15 This is a screenshot of the motion state of the present invention in a dynamic environment at t=160s.

[0117] Fig.16 This is a screenshot of the motion state of the present invention in a dynamic environment at t=200s.

[0118] Fig.17 It is a schematic diagram of the relative distance change curve within the cluster under a dynamic environment of the present invention.

[0119] Fig.18It is a schematic diagram of the curve of the relative distance change outside the cluster under a dynamic environment of the present invention. DETAILED DESCRIPTION

[0120] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0121] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0122] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0123] The following is attached to this application specification Figure 1-18 , the technical solutions in the embodiments of the present application are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0124] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0125] Implementation Method 1

[0126] This embodiment provides an unmanned cluster collaborative navigation method based on a distributed cluster self-organizing model. This method is first based on Boyd's OODA loop (observation, orientation, decision, action), combined with a local limited perception model, to establish a description of the interactive decision-making process between a single unmanned platform and its external environment. The roles of different UAV platforms in the cluster are further classified, and the efficiency of collaborative flight is improved through the guiding behavior of key nodes. Each UAV utilizes a dynamic adjustment mechanism of control parameters to allow adaptive adjustment based on local flight status. In addition, each UAV is equipped with a model predictive control (MPC) controller, which provides feasible control inputs to ensure robust and reliable operation in complex and dynamic scenarios. By integrating the above-mentioned multiple modules, the distributed cluster autonomous collaborative navigation method of the present invention is finally formed. The method flow chart is shown as follows. Figure 1 As shown, the method comprises the following steps:

[0127] Step 1: Obtain the task environment and preprocess it;

[0128] Furthermore, the step 1 specifically includes the following steps:

[0129] Step 1.1: Consider an unknown three-dimensional task environment, which can be obtained by processing existing elevation DEM data or by digital computer simulation. The shape of the obstacle is approximated by its circumscribed circle, whose radius can be expressed as T r Therefore, obstacles and no-fly zones in the environment can be represented by sets. In the present invention, a cylindrical model with randomly distributed positions is used to simulate obstacles that unmanned swarms may encounter during low-altitude flight.

[0130] Step 1.2: For non-convex obstacles that may appear in the environment, the unmanned swarm can transform them into convex obstacles by adjusting the flight altitude and further approximate them using the circumscribed cylinder in step 1.1;

[0131] Step 2: Describe the unmanned cluster platform using characteristics;

[0132] Furthermore, the step 2 specifically includes the following steps:

[0133] Step 2.1: Simplify the platform kinematic model;

[0134] Specifically, in the task scenario involved in the present invention, it is assumed that the unmanned clusters participating in the task are marked as N={1,...,N}. For convenience, the following simplified motion model is adopted for each unmanned platform i, which can be expressed as formula (1):

[0135]

[0136] Among them, x i ,y i ,h i is the three-dimensional coordinate of the unmanned platform i, and are the control inputs of the position holding, heading holding and altitude holding autopilots of platform i respectively; τ v τ ψ , τ h and τ λ are the time constants of the three autopilots;

[0137] After determining the time constant, the model is further simplified to a second-order integrator model, as shown in formula (2):

[0138]

[0139] Among them, p i =(x i ,y i ,h i ) T represents the position vector, represents the velocity vector, represents the control input of each platform; the motion state and acceleration input are further restricted to ensure that they can be processed by subsequent actuators; at each time step k, set v min ≤v i (k)≤v max ,u min ≤u i (k)≤u max ; Based on the control input, the fourth-order Runge-Kutta method is used to update the motion state of each unmanned platform at each moment;

[0140] Step 2.2: Based on the simplified platform kinematic model in step 2.1, divide the platform task roles;

[0141] Specifically, the step 2.2 includes dividing the unmanned platform into an informed member or an uninformed member according to whether the unmanned platform has global target information;

[0142] Among them, the informed unmanned platform has global navigation information and can guide the uninformed unmanned platform to fly to the target;

[0143] In contrast, an uninformed unmanned platform can only perceive the local state changes of its neighbors and use this information to determine its state at the next time step;

[0144] During the initialization process, informed individuals are randomly assigned to the population according to the specified proportion Δ.

[0145] Step 3: For the unmanned swarm platform described in step 2, an interaction framework criterion based on the Boyd ring is designed inside and outside the swarm. In this framework, each unmanned platform adjusts its own motion state according to the motion state of the surrounding platforms and the external environment information perceived by the onboard sensors, thereby achieving safe collaborative navigation from the initial position to the target.

[0146] Furthermore, the step 3 specifically includes the following steps:

[0147] Step 3.1: Each unmanned platform observes the motion state of the unmanned platforms in the adjacent airspace based on the sensors it carries and obtains its predicted motion state at the current moment and in the future with the help of the corresponding filtering prediction algorithm;

[0148] Furthermore, the step 3.1 is specifically as follows: step 3.1.1: at the current speed v i (k) and A reference coordinate system is established for the reference direction, and counterclockwise rotation is defined as positive; on this basis, the field of view (FOV) is defined as where α max for the maximum viewing angle; then Represents the local perception range R of unmanned platform i sen In order to describe unmanned platforms at different locations, the field of view is divided into Q equal parts; the individuals located in the qth angular interval It can be expressed as formula (3), where ∠p ij (k) represents the relative angle between unmanned platform i and unmanned platform j, ||p ij (k)||2 represents the relative distance between unmanned platform i and unmanned platform j;

[0149]

[0150] In order to reduce the complexity of perception, formula (4) is used to calculate the nearest perceptible individual in each subinterval and form a neighbor set Specifically,

[0151]

[0152] Since the relative positions between unmanned platforms will not change suddenly, when platform i detects platform j, it will create and store the state information structure of platform j locally. In practical applications, methods such as Kalman filtering can be used to estimate the motion state of platform j.

[0153] Step 3.1.2: If the current unmanned platform calculates that it is at the front of the cluster flight based on the relative position and relative speed information, if the unmanned platform is currently in the role of a non-informed person, it will apply to the informed unmanned platform in the group through the communication network to obtain target information to lead the cluster to fly towards the target. Once it no longer occupies a dominant position, it will automatically switch to the follower role; this mechanism ensures that at any time, there is at least one leader guiding the cluster to move towards the target; it can be seen that the information flow only includes the target and the corresponding sign information, and does not involve the exchange of real-time status information; this event-triggered mechanism effectively reduces the frequency of information exchange, thereby reducing the communication burden; if the current unmanned platform is in the role of an informed person, skip this step;

[0154] Step 3.1.3: If the current unmanned platform detects an external obstacle and needs to take obstacle avoidance action, define the critical neighbor set under the obstacle detection condition in formula (5): where <·> represents the angle between two vectors and λ represents the critical neighbor angle range; by narrowing the set of critical neighbors, the platform can be more purposefully guided to the neighbors in the forward direction;

[0155]

[0156] It can be seen that the key neighbor set is the set of sensing neighbors A subset of , the establishment of this relationship does not require a communication link; it should be pointed out that when the key neighbor set of an individual is an empty set, it uses the communication network to obtain global target information from the locally grouped insider platform and acts as an "informed person" to continue flying towards the target and restores to its original identity after returning to the cluster;

[0157] If the current unmanned platform does not detect any obstacles, it continues to use the perceived neighbor set. Perform subsequent state parameter calculation updates.

[0158] Step 3.2: Each unmanned platform makes local decisions based on the state observations of adjacent platforms and environmental observation information to achieve dynamic adaptive adjustment of key parameters;

[0159] Further, the step 3.2 includes the following steps:

[0160] Step 3.2.1: If the current flight status triggers the optimization flag (internal safety distance is too small or external obstacle is encountered) and reaches the optimization interval T0, go to step 3.2.2;

[0161] If the current flight status is safe, the current parameter set is maintained and the flight continues. When the fixed optimization interval T1 is reached, the process also goes to step 3.2.2.

[0162] At other times, the flight state is updated according to step 3.3; it is worth noting that the present invention does not assume that the optimization process between platforms is synchronized; the described method does not require explicit exchange of motion states through a communication network, thereby eliminating the need for strict clock synchronization between platforms during parameter optimization; this asynchronous method can adapt to external dynamic environments more flexibly and in real time;

[0163] Step 3.2.2: Design of interaction criteria between members within the cluster;

[0164] Furthermore, the step 3.2.2 is specifically as follows:

[0165] Step 3.2.2.1: Calculate the relative velocity alignment term to prevent the cluster from dispersing due to local velocity inconsistency during flight, which can be expressed as formula (6);

[0166]

[0167] Considering that cluster members tend to align their states with nearby individuals, the relative distance between the neighboring members and the unmanned platform i is first calculated as Considering that individuals closer to each other have a greater impact on speed consistency, the normalized vector is obtained It is used as a local weight to calculate the total alignment term, as shown in formula (7):

[0168]

[0169] in, is a constant that needs to be determined. It reflects the influence of the neighbors of unmanned platform i on its speed at the current time k;

[0170] Step 3.2.2.2: Calculate member local cohesion to ensure perception range The members in the cluster are kept close to prevent the unmanned cluster from decomposing, which can be expressed as formula (8):

[0171]

[0172] Among them, d coh It is the minimum distance between unmanned platforms to produce cohesion. is the unit vector pointing from unmanned platform i to unmanned platform j. Considering that cluster members tend to prioritize the cohesion effect with individuals at the farthest distance within the sensing range, the relative distance weighted method is used here to calculate the local overall cohesion term as shown in formula (9):

[0173]

[0174] in, is a cohesion term coefficient that needs to be determined dynamically;

[0175] Step 3.2.2.3: Calculate the local repulsion term to avoid collision between members and ensure that the members always maintain the minimum safe distance; the repulsion effect is defined as formula (10)

[0176]

[0177] Among them, d rep is the maximum interaction range where unmanned platforms begin to exclude each other; is the avoidance direction from unmanned platform j to unmanned platform i; considering that individuals give priority to avoiding collisions with their neighbors, the relative distance is used as the weighted weight of the avoidance term; the total repulsion term of unmanned platform i and its adjacent individuals can be expressed as formula (11):

[0178]

[0179] in, is the weight parameter of the exclusion term;

[0180] Step 3.2.2.4: In order to maintain stability during flight, the change in the speed of the unmanned platform will not exceed the performance limit; therefore, a speed maintenance term is added, as shown in formula (12)

[0181]

[0182] in, is the reference speed, usually represented by the average performance of unmanned platforms; is the coefficient, is the average flight direction of the aircraft, as shown in formula (13), and a relative distance weighting method is also used to ensure that the unmanned platform is aligned with distant individuals first, thereby reducing the probability of splitting:

[0183]

[0184] in, is the unit vector of unmanned platform j;

[0185] Through the above analysis, the definition To characterize all interactions within the swarm, as shown in formula (14):

[0186]

[0187] Step 3.2.3: Based on the interaction criteria in step 3.2.2, further define the information interaction items between the unmanned platform and the environment;

[0188] Further, the step 3.2.3 includes the following steps:

[0189] Step 3.2.3.1: Design the goal-oriented interaction term for the informed individuals with global navigation information in the cluster, as shown in formula (15);

[0190]

[0191] Among them, p g is the target coordinate, C tar is the attraction term gain, Ω is the set of informed unmanned platforms, and this step is skipped for uninformed individuals;

[0192] Step 3.2.3.2: Further, based on the relative interaction between the platform and the environmental obstacles, the obstacle avoidance interaction term is defined. Assume that the field of view of the unmanned platform i is α i (k); After the obstacle enters the detection range, its boundary intersection point P can be calculated l =(x l ,y l ), P r =(x r ,y r ) ; use the law of cosines to calculate the angle θ it occupies i,o (k) Assume that unmanned platform i detects an obstacle at time k Then the feasible angle of unmanned platform i can be expressed as formula (16):

[0193]

[0194] In order to prevent the unmanned platform from splitting due to obstacle avoidance behavior based on local perception, the feasible angle A i (k) is divided into W sub-intervals A i,w (k); Then, evaluate the angular difference between the center line of each subinterval and the heading of the current local neighbor, and select the subinterval with the smallest relative angular difference as the local obstacle avoidance direction; it can be calculated as formula (17);

[0195]

[0196] It is worth noting that when calculating the obstacle avoidance direction, the key neighbor set is used This helps the unmanned platform select individuals that can guide it forward, allowing for a more targeted response to external obstacles. On this basis, the external obstacle avoidance behavior term is defined as shown in formula (18):

[0197]

[0198] in, and Represents the boundary angle of each sub-interval; By analyzing the navigation direction, the expected navigation item of the unmanned platform i is defined, where represents the gain factor; this factor allows to adapt its heading by adaptively modifying its navigation gain, thus adapting to the different requirements of the task in different situations; similarly, define To represent the sum of the interactions between the unmanned platform and the external environment, as shown in formula (19):

[0199]

[0200] Step 3.2.4: Aggregate the various control items to obtain the overall control input, as shown in formula (20); this comprehensive input integrates the effects of neighboring speeds, obstacle avoidance, and target attraction, enabling the unmanned swarm to coordinate safe navigation in complex environments;

[0201] Step 3.2.5: After obtaining the virtual control input, combine the motion equation to predict its next T p Step status in represents the predicted state at each time step; based on the predicted state of itself and its neighbors, a local performance evaluation function is defined for each unmanned platform, as shown in formula (21);

[0202] Each term in the function is normalized to facilitate optimization.

[0203]

[0204] in represents the local neighbor position concentration of unmanned platform i. When the sensed neighbor set is non-empty, is calculated based on the distance between their predicted and current positions, where The furthest neighbor. δ It is a parameter used to adjust the rate of change of the indicator.

[0205]

[0206] in, represents the degree of directional consistency among the local neighbors of unmanned platform i; represents the collision cost between the unmanned platform and its neighbors within the sensing range, where d rep Represents the predetermined safety radius of unmanned platform i. represents the collision cost between the unmanned platform and external obstacles, where r o A preset safe distance for each obstacle; by limiting the local objective function, each unmanned platform can independently optimize its parameters and flexibly adjust its local behavior.

[0207] Step 3.2.6: Based on the above local optimization indicators, the local optimization target of each unmanned platform is defined as formula (26):

[0208]

[0209] Among them, γ∈(0,1) is the state decay factor that characterizes the importance between different times. In essence, this is achieved by adjusting the weight strategy set To maximize T p To facilitate deployment, we use the same particle swarm solver for each unmanned platform to achieve real-time online optimization.

[0210] Further, the step 3.2.6 includes the following steps,

[0211] Step 3.2.6.1: Initialize the meta-heuristic optimization algorithm, including the feasible range of each parameter, the number of populations, the maximum number of iterations, and the set of factors to be optimized can be expressed as Represents the weights of each item in the virtual control input;

[0212] Step 3.2.6.2: Substitute the set of variables to be optimized into formula (26) to calculate the function value of the current individual motion state in the future period of time;

[0213] Step 3.2.6.3: Perform multiple iterations of optimization to obtain the control parameter set at the future moment

[0214] Step 3.3: After obtaining the optimized parameter set Then, the motion model in step 2.1 is used to predict T p The future state of the time step, serving as a reference trajectory To ensure the stability of the input, a decoupled MPC controller is used for the horizontal and vertical directions in equation (27), where is based on the reference trajectory before the current state optimization; this method provides the final control input, which is then used to update the state at the next time step;

[0215]

[0216] Step 4: Based on the interactive framework criteria designed in step 3, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and updates the flight status using the fourth-order Runge-Kutta method; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute step 3 until it reaches the target point;

[0217] Specifically, assume that there are M randomly distributed obstacles in the 2×2×2km mission area, where the radius of each obstacle is randomly distributed within 3%-6% of the mission area size. The initial position of the unmanned cluster is randomly generated in the specified area and the relative distance between them is guaranteed to be safe. The initial heading is randomly generated within the range of ±30° on the left and right sides of the line connecting the starting point and the end point, and the informed individuals are randomly selected and initialized according to Δ=20%. Our goal is to achieve autonomous collaborative navigation of the cluster from the starting point to the end point in an obstacle environment. The following table gives the relevant parameter ranges.

[0218] Table 1 Key simulation parameters

[0219]

[0220]

[0221] from Figure 3-Figure 6 It can be seen that the proposed method can achieve collision-free collaborative navigation behavior from the starting point to the end point, meeting the task requirements.

[0222] Figure 7 and Figure 8 The curve reflects the change of the relative distance between the inside and outside of the cluster during the flight, where the legend 0 represents this method, and 1, 2, and 3 represent the other three methods. It can be seen that it can maintain a safe flight state and achieve the established mission objectives.

[0223] Fig. 9 and Fig.10 The velocity and angular velocity curves of the cluster members during flight reflect the change of the velocity and angular velocity. It can be seen that different methods can obtain control input within the specified range. In addition, it can be seen that the state quantity of method 2 changes relatively slowly in comparison, because it uses the rolling control strategy of MPC and adds energy consumption to the optimization model. However, from the final results, it can be seen that its arrival time at the end point is the longest compared to other methods.

[0224] Figure 11-Figure 16 The figure shows the verification results of the collaborative navigation behavior of this method in a dynamic obstacle environment. The red obstacle indicates that the obstacle is stationary at that moment, and the green obstacle indicates that the obstacle is moving at that moment. It can be seen that our method avoids external collision and dispersion of the cluster while maintaining the collaborative navigation behavior. This can be seen from Fig.17 and Fig.18 The distance variation curves of clusters relative to the inside and outside are verified, which proves the effectiveness of the proposed method.

[0225] Implementation Method 2

[0226] This embodiment provides an unmanned cluster collaborative navigation system based on a distributed cluster self-organizing model. The system uses the unmanned cluster collaborative navigation method based on a distributed cluster self-organizing model as described in Embodiment 1. The system includes:

[0227] Preprocessing module: obtain the task environment and preprocess it;

[0228] Feature description module: describes the unmanned cluster platform using features;

[0229] Cluster internal and external interaction framework rule design module: Based on the unmanned cluster platform described by characteristics, the cluster internal and external interaction framework rule design based on Boyd ring is carried out. In this framework, each unmanned platform adjusts its own motion state according to the motion state of the surrounding platforms and external environment information perceived by the onboard sensors, so as to achieve safe collaborative navigation from the initial position to the target;

[0230] Flight status update module: Based on the interactive framework criterion, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and uses the fourth-order Runge-Kutta method to update the flight status; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute the interactive framework criterion design module inside and outside the cluster until it reaches the target point.

Claims

1. An unmanned cluster collaborative navigation method based on a distributed cluster self-organizing model, characterized in that: The method comprises the following steps: Step 1: Obtain the task environment and preprocess it; Step 2: Describe the unmanned cluster platform using characteristics; Step 3: For the unmanned swarm platforms characterized in step 2, design the interaction framework criteria inside and outside the swarm based on the Boyd ring, so that each unmanned platform can adjust its own motion state according to the motion state of the surrounding platforms and external environment information perceived by the sensors on board; Step 4: Based on the interactive framework criteria designed in step 3, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and uses the fourth-order Runge-Kutta method to update the flight status; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute step 3 until it reaches the target point.

2. The unmanned swarm collaborative navigation method according to claim 1, characterized in that: The step 1 specifically includes the following steps: Step 1.1: Based on the existing elevation DEM data or obtained through digital computer simulation, the obstacle shape is approximated by its circumscribed circle, and its radius is expressed as T r ; A cylindrical model with randomly distributed positions is used to simulate the obstacles that unmanned swarms may encounter during low-altitude flight; Step 1.2: For non-convex obstacles that may appear in the environment, the unmanned swarm converts them into convex obstacles by adjusting the flight altitude and further approximates them using the circumscribed cylinder in step 1.

1.

3. The unmanned swarm collaborative navigation method according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1: Simplify the platform kinematic model; Specifically, step 2.1 is as follows: assuming that the unmanned clusters participating in the task are labeled as N={1,...,N}, the following simplified motion model is adopted for each unmanned platform i, expressed as formula (1): Among them, x i ,y i ,h i is the three-dimensional coordinate of the unmanned platform i, and are the control inputs of the position holding, heading holding and altitude holding autopilots of platform i respectively; τ v τ ψ , τ h and τ λ are the time constants of the three autopilots; After determining the time constant, the model is simplified to a second-order integrator model, as shown in formula (2): Among them, p i =(x i ,y i ,h i ) T represents the position vector, represents the velocity vector, represents the control input of each platform; at each time step k, set v min ≤v i (k)≤v max ,u min ≤u i (k)≤u max ; Based on the control input, the fourth-order Runge-Kutta method is used to update the motion state of each unmanned platform at each moment; Step 2.2: Based on the simplified platform kinematic model in step 2.1, divide the platform task roles; Specifically, the step 2.2 includes dividing the unmanned platform into an informed member or an uninformed member according to whether the unmanned platform has global target information; Among them, the informed unmanned platform has global navigation information and can guide the uninformed unmanned platform to fly to the target; An uninformed unmanned platform can only perceive the local state changes of its neighbors and use this information to determine its state in the next time step; During the initialization process, informed individuals are randomly assigned to the population according to the specified proportion Δ.

4. The unmanned swarm collaborative navigation method according to claim 1, characterized in that: The step 3 specifically includes the following steps: Step 3.1: Each unmanned platform observes the motion state of the unmanned platforms in the adjacent airspace based on the sensors it carries and obtains its predicted motion state at the current moment and in the future with the help of the corresponding filtering prediction algorithm; Step 3.2: Each unmanned platform makes local decisions based on the state observations of adjacent platforms and environmental observation information to achieve dynamic adaptive adjustment of key parameters; Step 3.3: After obtaining the optimized parameter set Then, the motion model in step 2.1 is used to predict T p The future state of the time step, serving as a reference trajectory A decoupled MPC controller is used for the horizontal and vertical directions, where is the reference trajectory before optimization based on the current state; it provides the final control input and is then used to update the state at the next time step.

5. The unmanned swarm collaborative navigation method according to claim 4, characterized in that: The step 3.1 is specifically as follows: step 3.1.1: at the current speed v i (k) and Establish a reference coordinate system for the reference direction, define counterclockwise rotation as positive; define the field of view as where α max for the maximum viewing angle; then Represents the local perception range R of unmanned platform i sen The detectable individuals within the qth angular interval; the field of view is divided into Q equal parts; then the individuals within the qth angular interval It can be expressed as formula (3), where ∠p ij (k) represents the relative angle between unmanned platform i and unmanned platform j, ||p ij (k)||2 represents the relative distance between unmanned platform i and unmanned platform j; Formula (4) is used to calculate the nearest perceptible individual in each subinterval and form a neighbor set Specifically, Since the relative positions between unmanned platforms will not mutate, when platform i detects platform j, it will create and store the state information structure of platform j locally; Step 3.1.2: If the current unmanned platform calculates that it is at the front of the cluster flight based on the relative position and relative speed information, if the unmanned platform is currently in the role of non-informed, it will apply to the informed unmanned platform in the group to obtain target information through the communication network to lead the cluster to fly towards the target. Once it is no longer in a dominant position, it will automatically switch to the follower role; the information flow only includes the target and the corresponding sign information, and does not involve the exchange of real-time status information; if the current unmanned platform is in the role of informed, skip this step; Step 3.1.3: If the current unmanned platform detects an external obstacle and needs to take obstacle avoidance action, define the critical neighbor set under the obstacle detection condition in formula (5): Where <·> represents the angle between two vectors, and λ represents the critical neighbor angle range; Key Neighbor Set is the set of sensing neighbors When the critical neighbor set of an individual is an empty set, it uses the communication network to obtain global target information from the local grouped insider platform and acts as an "informed person" to continue flying towards the target and restores to its original identity after returning to the cluster. If the current unmanned platform does not detect any obstacles, it continues to use the perceived neighbor set. Perform subsequent state parameter calculation updates.

6. The unmanned swarm collaborative navigation method according to claim 4, characterized in that: The step 3.2 comprises the following steps: Step 3.2.1: If the current flight status triggers the optimization flag and reaches the optimization interval T0, go to step 3.2.2; If the current flight status is safe, the current parameter set is maintained and the flight continues. When the fixed optimization interval T1 is reached, the process also goes to step 3.2.

2. At other times, the flight status is updated according to step 3.3; Step 3.2.2: Design of interaction criteria between members within the cluster; Step 3.2.3: Based on the interaction criteria in step 3.2.2, further define the information interaction items between the unmanned platform and the environment; Step 3.2.4: Aggregate the various control items to obtain the overall control input; Step 3.2.5: After obtaining the virtual control input, combine the motion equation to predict its next T p Step status in Represents the predicted state at each time step; a local performance evaluation function is defined for each unmanned platform based on its own and its neighbors’ predicted states; Step 3.2.6: Based on the above local optimization indicators, the local optimization target of each unmanned platform is defined as formula (26): Among them, γ∈(0,1) is the state decay factor that characterizes the importance between different times.

7. The unmanned swarm collaborative navigation method according to claim 6, characterized in that: The step 3.2.2 is specifically as follows: Step 3.2.2.1: Calculate the relative velocity alignment term to prevent the cluster from dispersing due to local velocity inconsistency during flight, which can be expressed as formula (6); First, the relative distance between the neighboring member and the unmanned platform i is calculated as Get the normalized vector It is used as a local weight to calculate the total alignment term, as shown in formula (7): in, is a constant that needs to be determined. It reflects the influence of the neighbors of unmanned platform i on its speed at the current time k; Step 3.2.2.2: Calculate member local cohesion to ensure perception range The members within are kept close to each other as shown in formula (8), Among them, d coh It is the minimum distance between unmanned platforms to produce cohesion. is the unit vector pointing from unmanned platform i to unmanned platform j; the local overall cohesion term is calculated using the relative distance weighted method as shown in formula (9): in, is a cohesion term coefficient that needs to be determined dynamically; Step 3.2.2.3: Calculate the local repulsion term to ensure that members always maintain the minimum safe distance; the repulsion effect is defined as formula (10) Among them, d rep is the maximum interaction range where unmanned platforms begin to exclude each other; is the avoidance direction from unmanned platform j to unmanned platform i; the total repulsion term between unmanned platform i and its adjacent individuals is expressed as formula (11): in, is the weight parameter of the exclusion term; Step 3.2.2.4: In order to maintain stability during flight, a speed maintenance term is added, as shown in formula (12) in, is the reference speed; is the coefficient, is the average flight direction of the aircraft, as shown in formula (13): in, is the unit vector of unmanned platform j; definition To characterize all interactions within the swarm, as shown in formula (14):

8. The unmanned swarm collaborative navigation method according to claim 6, characterized in that: The step 3.2.3 comprises the following steps: Step 3.2.3.1: Design the goal-oriented interaction term for the informed individuals of global navigation information, as shown in formula (15); Among them, p g is the target coordinate, C tar is the attraction term gain, Ω is the set of informed unmanned platforms, and this step is skipped for uninformed individuals; Step 3.2.3.2: Assume that the field of view of unmanned platform i is α i (k); After the obstacle enters the detection range, its boundary intersection point P can be calculated l =(x l ,y l ), P r =(x r ,y r ) ; use the law of cosines to calculate the angle θ it occupies i,o (k) Assume that unmanned platform i detects an obstacle at time k Then the feasible angle of unmanned platform i is expressed as formula (16): The feasible angle A i (k) is divided into W sub-intervals A i,w (k); Evaluate the angular difference between the center line of each subinterval and the heading of the current local neighbor, and select the subinterval with the smallest relative angular difference as the local obstacle avoidance direction; calculated as formula (17); When calculating the obstacle avoidance direction, the key neighbor set is used The external obstacle avoidance behavior term is defined as shown in formula (18): in, and Represents the boundary angle of each sub-interval; By analyzing the navigation direction, the expected navigation item of the unmanned platform i is defined, where represents the gain factor; this factor allows to adapt its heading by adaptively modifying its navigation gain, thus adapting to the different requirements of the task in different situations; similarly, define To represent the sum of the interactions between the unmanned platform and the external environment, as shown in formula (19):

9. The unmanned swarm collaborative navigation method according to claim 6, characterized in that: The step 3.2.6 includes the following steps, Step 3.2.6.1: Initialize the meta-heuristic optimization algorithm, including the feasible range of each parameter, the number of populations, the maximum number of iterations, and the set of factors to be optimized is expressed as Represents the weights of each item in the virtual control input; Step 3.2.6.2: Substitute the set of variables to be optimized into formula (26) to calculate the function value of the current individual motion state in the future period of time; Step 3.2.6.3: Perform multiple iterations of optimization to obtain the control parameter set at the future moment 10. An unmanned cluster collaborative navigation system based on a distributed cluster self-organizing model, characterized in that: The system uses the unmanned cluster collaborative navigation method based on the distributed cluster self-organizing model as described in any one of claims 1 to 9, and the system includes: Preprocessing module: obtain the task environment and preprocess it; Feature description module: describes the unmanned cluster platform using features; Cluster internal and external interaction framework rule design module: Based on the unmanned cluster platform described by characteristics, the cluster internal and external interaction framework rule design based on Boyd ring is carried out, so that each unmanned platform can adjust its own motion state according to the motion state of the surrounding platforms and external environment information perceived by the sensors on board; Flight status update module: Based on the interactive framework criterion, each UAV uses its onboard model predictive controller to solve the control input at the next moment online and uses the fourth-order Runge-Kutta method to update the flight status; each UAV uses its onboard sensors to update the flight status of the surrounding unmanned platforms online and returns to execute the interactive framework criterion design module inside and outside the cluster until it reaches the target point.

Images