Multi-unmanned aerial vehicle navigation method, system, device and medium based on distributed model predictive control and virtual pipeline

Abstract

The application discloses a kind of multi-unmanned aerial vehicle navigation method, system, equipment and medium based on distributed model predictive control and virtual pipeline, and the method comprises the following steps: constructing virtual pipeline, forming the flyable area of unmanned aerial vehicle group;In each sampling period, each unmanned aerial vehicle generates the assumed trajectory in prediction time domain based on its current state using its dynamic model, and broadcasts to neighbor unmanned aerial vehicle;Obstacle intrusion judgment and virtual pipeline dynamic processing, construct linear separation hyperplane, generate local convex safety sub-pipeline constraint;Inter-machine collision risk judgment is carried out and inter-machine collision avoidance constraint is dynamically activated;Set target cost function, and construct optimization problem in combination with the above constraint;The optimal prediction control input sequence of current time is obtained by solving, and its first control instruction is executed;The above process is repeatedly executed until all unmanned aerial vehicles pass through virtual pipeline.The application can simultaneously solve multiple challenges such as dynamic environment adaptability, calculation real-time and trajectory smoothness.

Description

Technical Field

[0001] This invention relates to the field of aircraft navigation and control technology, and in particular to a multi-UAV navigation method, system, device and medium based on distributed model predictive control and virtual pipeline. Background Technology

[0002] The application prospects of drone swarm technology in logistics, inspection and other fields are broad. Its collaborative navigation needs to solve core problems such as collision avoidance between drones, environmental obstacle avoidance and dynamic constraints at the same time.

[0003] Virtual pipeline technology provides a macroscopic environmental obstacle avoidance framework for drone swarms by constructing "aerial corridors." However, most existing virtual pipeline solutions are based on pre-planning using static environmental information. When unforeseen dynamic obstacles intrude into the pipeline, these static pipelines cannot adjust their boundaries in real time, causing drones to lose effective safe zone guidance and posing a collision risk. Therefore, enabling virtual pipelines to cope with dynamic obstacles is crucial for their practical application in open and dynamic environments.

[0004] On the other hand, Distributed Model Predictive Control (DMPC) provides a feasible framework for real-time control of large-scale UAV swarms by decomposing the global optimization problem into distributed local optimization problems. However, efficiently integrating complex virtual pipeline constraints, especially dynamically changing pipeline boundaries, into the online optimization problem of DMPC is a significant challenge. Directly handling non-convex pipeline boundary constraints makes the optimization problem difficult to solve in real time.

[0005] Existing technologies lack a solution that can deeply integrate dynamic environment perception with distributed optimization control. Its shortcomings are mainly reflected in: (1) the virtual pipeline lacks a dynamic update mechanism; (2) the high-dimensional complex constraints of the dynamic pipeline are difficult to embed into the real-time distributed optimization framework. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-UAV navigation method, system, device, and medium based on distributed model predictive control and virtual pipelines. It deeply integrates dynamic virtual pipeline technology with the distributed model predictive control framework, solving the real-time processing challenges of complex pipeline constraints in optimization problems, and enabling safe and efficient collaborative navigation of multiple UAVs in dynamic environments.

[0007] Firstly, a multi-UAV navigation method based on distributed model predictive control and virtual pipelines is provided, including the following steps:

[0008] S1: Construct virtual pipelines to form a flyable area for drone swarms;

[0009] S2: In each sampling period, each UAV generates a hypothetical trajectory in the predicted time domain based on its current state using its dynamic model, and broadcasts it to neighboring UAVs.

[0010] S3: Perform obstacle intrusion detection and virtual pipeline dynamic processing, construct a linear separation hyperplane, and generate local convex safety sub-pipeline constraints;

[0011] S4: Each drone acquires the assumed trajectory information of its neighboring drones, performs an inter-drone collision risk assessment, and activates inter-drone collision avoidance constraints only when a collision risk is detected.

[0012] S5: Set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraint, inter-UAV collision avoidance constraint and UAV dynamic model to construct an optimization problem; solve to obtain the optimal predictive control input sequence for each UAV at the current moment, and execute its first optimal control command;

[0013] S6: Repeat steps S2 to S5 until all drones pass through the virtual pipeline safely and without collision.

[0014] Further, step S1 includes:

[0015] S1.1: Geometric definition of virtual pipe, virtual pipe It is defined by a path and the safe zone formed by extending the normal vector of that path.

[0016] S1.2: Definition of safe avoidance zone: A safe avoidance zone is introduced outside the boundary of the virtual pipeline. The width of the safe avoidance zone shall not be less than the minimum safe distance for the drone; virtual pipeline and safe avoidance area The union of these elements constitutes the flyable area of ​​the drone swarm. .

[0017] Further, step S2 includes:

[0018] S2.1: Prediction model construction: Based on the dynamic model of the UAV, construct the state prediction equations of each UAV in the prediction time domain;

[0019] S2.2: Assuming trajectory generation and communication, each UAV generates a hypothetical trajectory based on its current state through a state prediction equation and broadcasts it to neighboring UAVs; under the initial state of the UAV, the predicted trajectory is constructed using the initial position information as the hypothetical trajectory for the previous cycle.

[0020] Further, step S3 includes:

[0021] S3.1: Pipeline status judgment and boundary acquisition. Starting from the current position of the UAV, it is determined whether the predefined virtual pipeline is intruded by an obstacle within the current perception range of the UAV. If an obstacle intrusion is detected, the pipeline replanning mechanism is triggered. Based on the obstacle position information, the original virtual pipeline boundary is recalculated to generate a new local pipeline boundary that avoids the obstacle. The geometry of the new local pipeline boundary remains smooth and continuous with the unaffected global virtual pipeline boundary.

[0022] S3.2: Generation of a locally convex safe sub-pipeline. Combine the virtual pipe boundary obtained in step S3.1 with the current perception range of the UAV to construct a locally convex safe sub-pipeline.

[0023] S3.3: Constraint linearization, using a linear separating hyperplane, describes the boundary of the local convex safety sub-pipe as a set of linear inequality constraints.

[0024] Further, step S4 includes:

[0025] S4.1: Neighbor information exchange, each UAV receives the assumed trajectory of all neighboring UAVs in the previous cycle through the communication network, as the known input for its own optimization in the current cycle;

[0026] S4.2: Collision risk assessment. Under the distributed framework, each UAV performs time-step inter-UAV collision risk detection in the prediction time domain based on its own assumed trajectory and the assumed trajectories of neighboring UAVs.

[0027] S4.3: Collision avoidance constraint generation. Only when an inter-drone collision risk is detected, a linear separating hyperplane constraint is dynamically constructed for the neighboring drone with a collision risk. This constraint forces both drones to stay on opposite sides of the hyperplane in the prediction time domain and maintain a minimum safe distance.

[0028] Furthermore, when assessing collision risk, each drone focuses only on the time step in which a collision is detected within the prediction time domain.

[0029] Further, step S5 includes:

[0030] S5.1: Problem formulation: Each UAV predicts the time domain based on its own state prediction equation and the assumed trajectories of its neighboring UAVs. Control input sequence within Using the objective cost function, local convex safety sub-pipeline constraints, inter-machine collision avoidance constraints, and UAV dynamics model as decision variables, a constrained time-domain optimization problem is formed.

[0031] S5.2: Solve, execute and update. Solve the optimization problem for each UAV to obtain the optimal predictive control input sequence and the corresponding optimal predictive state trajectory for each UAV. Each UAV executes its first optimal control command. The optimal predictive state trajectory obtained in this solution is used as the assumed trajectory at the next sampling time.

[0032] Secondly, a multi-UAV navigation system based on distributed model predictive control and virtual pipelines is provided, including:

[0033] The pipeline construction module is used to build virtual pipelines, forming a flyable area for drone swarms;

[0034] The collaborative communication and trajectory hypothesis module is used to generate a hypothetical trajectory in the prediction time domain based on the current state of each UAV using its dynamic model in each sampling period, and broadcast it to neighboring UAVs.

[0035] The Local Convex Safe Sub-Pipe Constraint Generation Module is used to perform obstacle intrusion judgment and virtual pipeline dynamic processing, construct a linear separating hyperplane, and generate local convex safe sub-pipe constraints.

[0036] The inter-drone collision avoidance constraint activation module is used for each drone to obtain the assumed trajectory information of neighboring drones, to judge the inter-drone collision risk, and to activate the inter-drone collision avoidance constraint only when a collision risk is detected.

[0037] The local prediction and rolling optimization module is used to set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraints, inter-UAV collision avoidance constraints and UAV dynamics model to construct an optimization problem; solve for the optimal predictive control input sequence of each UAV at the current time, and each UAV executes its first optimal control command;

[0038] The collaborative communication and trajectory assumption module, the local convex safety sub-pipe constraint generation module, the inter-machine collision avoidance constraint activation module, and the local prediction and rolling optimization module are executed in a loop until all UAVs pass through the virtual pipe safely and without collision.

[0039] Thirdly, an electronic device is provided, comprising:

[0040] A memory on which computer programs are stored;

[0041] A processor is used to load and execute the computer program to implement the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described above.

[0042] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described above.

[0043] This invention proposes a multi-UAV navigation method, system, device, and medium based on distributed model predictive control and virtual pipelines. Compared with existing technologies, it can simultaneously solve multiple challenges such as dynamic environment adaptability, real-time computation, and trajectory smoothness. Specifically, it is reflected in:

[0044] (1) Through the dynamic pipeline update mechanism, the UAV cluster can actively and quickly respond to dynamic obstacles intruding into the pipeline, which breaks through the limitations of static pipeline in dynamic open environment and ensures the safe navigation of the cluster in changing environment.

[0045] (2) The on-demand collision avoidance strategy based on collision risk identification is adopted. The inter-machine collision avoidance constraint is activated only when a collision risk is detected, which reduces redundant calculations, significantly improves the solution efficiency, and ensures the real-time performance of the system.

[0046] (3) It provides an integrated solution that balances safety and performance. By using a unified optimization framework to collaboratively handle pipeline constraints, dynamic obstacles and multi-machine collision avoidance, it achieves smooth and efficient collaborative navigation while ensuring strict safety. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a multi-UAV navigation method based on distributed model predictive control and virtual pipeline provided in an embodiment of the present invention;

[0049] Figure 2 This is a detailed flowchart of the multi-UAV navigation method based on distributed model predictive control and virtual pipeline provided in the embodiments of the present invention;

[0050] Figure 3 This is a schematic diagram of a virtual pipeline in an obstacle environment provided in an embodiment of the present invention;

[0051] Figure 4 These are schematic diagrams of the flight of a drone swarm inside a pipeline provided in an embodiment of the present invention, wherein (a), (b), (c), and (d) are schematic diagrams of the flight of a drone swarm inside a pipeline at different times;

[0052] Figure 5This is a minimum distance map between the UAV and the virtual pipeline boundary provided in the embodiment of the present invention, wherein (a) and (b) are the minimum distance maps between the UAV and the left and right boundaries of the virtual pipeline, respectively;

[0053] Figure 6 This is a minimum distance map between drones provided in an embodiment of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0055] The purpose of this invention is to overcome the shortcomings of existing static virtual pipelines in dealing with dynamic obstacles. This invention provides a multi-UAV navigation method, system, device, and medium based on distributed model predictive control and virtual pipelines, enabling safe and efficient collaborative navigation of multiple UAVs in dynamic environments. Its core lies in constructing a dynamically updated virtual pipeline as the global flyable area of ​​the UAV swarm. Under the DMPC framework, each UAV independently performs rolling optimization based on its own state and assumed trajectory information shared by neighboring UAVs, solving the real-time processing challenge of complex pipeline constraints in optimization problems. The technical solution of this invention will be described in detail below with reference to specific embodiments.

[0056] like Figure 1 As shown, this embodiment of the invention provides a multi-UAV navigation method based on distributed model predictive control and virtual pipelines, including the following steps:

[0057] S1: Construct virtual pipelines to form a flyable area for drone swarms.

[0058] Specifically, step S1 includes:

[0059] S1.1: Geometric definition of virtual pipe, virtual pipe Virtual pipelines are defined as strip-shaped or tubular flyable areas in two-dimensional or three-dimensional space. It is defined by a path and the safe zone formed by extending the path's normal vector direction; the path can be a set of discrete path points or a continuous centerline.

[0060] S1.2: Definition of safe avoidance zone: A safe avoidance zone is introduced outside the boundary of the virtual pipeline. The width of the safe avoidance zone shall not be less than the minimum safe distance of the drone. During flight, the drone's position is restricted to prevent it from crossing the safe avoidance zone; virtual pipeline and safe avoidance area The union of these elements constitutes the flyable area of ​​the drone swarm. , means as follows:

[0061] ;

[0062] In the specific implementation, the drone's position constraint will be set to strictly within... internal.

[0063] S2: During each sampling period, each UAV generates a hypothetical trajectory in the prediction time domain based on its current state using its dynamic model, and broadcasts it to neighboring UAVs.

[0064] Specifically, step S2 includes:

[0065] S2.1: Prediction model construction: Based on the dynamic model of the UAV, construct the state prediction equations of each UAV in the prediction time domain;

[0066] Typically, the state prediction equations for each UAV can be expressed as follows:

[0067] ;

[0068] Where A and B represent the system matrix and control input matrix, respectively; and Let k represent the state vector and control input vector at discrete time step k, respectively. The control input vector is composed of the UAV's kinematic states (such as position and velocity). This refers to the physical quantity driving the state change (such as acceleration or velocity). The UAV dynamics model and state prediction equation are mature technologies and will not be elaborated upon here. In the prediction time domain, this state prediction equation is recursively used to calculate the UAV's future state sequence;

[0069] S2.2: Assuming trajectory generation and communication, each UAV generates an assumed state sequence based on its current state (such as position and speed) through a state prediction equation and broadcasts it to neighboring UAVs; under the initial state of the UAV, the predicted trajectory is constructed using the initial position information as the assumed trajectory for the previous cycle.

[0070] S3: Perform obstacle intrusion judgment and virtual pipeline dynamic processing, construct a linear separation hyperplane, and generate local convex safety sub-pipeline constraints.

[0071] Step S3 is an environment-aware, condition-triggered safe zone generation process, which specifically includes:

[0072] S3.1: Pipeline status judgment and boundary acquisition. Starting from the current position of the drone, determine whether the predefined virtual pipeline is intruded by obstacles within the current perception range of the drone.

[0073] If no obstacle intrudes into the predefined global virtual pipeline, the segment related to the current field of view of the drone is directly extracted from the boundary of the predefined virtual pipeline as the basis for processing.

[0074] If an obstacle is detected encroaching on a predefined global virtual pipeline, a pipeline replanning mechanism is triggered. This mechanism recalculates the original virtual pipeline boundary based on the obstacle's location information, generating a new local pipeline boundary that avoids the obstacle. The new local pipeline boundary generated by this mechanism is local, only adjusting the area adjacent to the obstacle, and the geometry of the new local pipeline boundary remains smooth and continuous with the unaffected global virtual pipeline boundary, ensuring that the UAV's heading command does not undergo abrupt changes.

[0075] S3.2: Generation of a locally convex safe sub-pipeline. Combine the virtual pipe boundary (whether predefined or replanned) obtained in step S3.1 with the current perception range of the UAV to construct a locally convex safe sub-pipeline.

[0076] Specifically, based on the virtual pipeline boundary obtained in step S3.1, spatial points within the perception range are sampled and the convex hull is calculated. The convex hull encloses a convex polyhedron (a convex polygon in a two-dimensional plane). This convex hull region should satisfy the flyable area of ​​the UAV swarm defined in S1.2. This forms a new locally safe sub-pipeline; simultaneously, the region of this locally safe sub-pipeline is updated as the UAV's position changes. In some preferred embodiments, the convex hull vertices can be simplified to reduce the number of constraints and improve computational efficiency.

[0077] S3.3: Constraint linearization. Using a linear separating hyperplane, the boundary of the local convex safe sub-pipe is described as a set of linear inequality constraints. This constraint is embedded in the optimization problem of distributed model predictive control, defining a safe flight region for the UAV that can be directly processed by the optimizer. Specifically, a normal vector can be calculated for each face (or each edge in two dimensions) of the convex safe sub-pipe. and scalar This allows the drone on the allowed side of the face (edge ​​in 2D) to perform actions at each time step in the prediction time domain. The predicted locations are all constrained within the set of all such inequalities:

[0078] ;

[0079] In the formula, Indicates the time step of the i-th drone. The predicted location.

[0080] S4: Each UAV acquires the assumed trajectory information of its neighboring UAVs, performs an inter-UAV collision risk assessment, and activates inter-UAV collision avoidance constraints only when a collision risk is detected.

[0081] Specifically, step S4 includes:

[0082] S4.1: Neighbor information exchange, each UAV receives the assumed trajectory of all neighboring UAVs in the previous cycle through the communication network, as the known input for its own optimization in the current cycle;

[0083] S4.2: Collision risk assessment. Under the distributed framework, each UAV performs time-step-by-time collision risk detection in the prediction time domain based on its own predicted trajectory and the assumed trajectories of neighboring UAVs.

[0084] The method for assessing the risk of inter-drone collision is as follows: Unmanned Aerial Vehicles (UAVs) For neighboring drones If the assumed trajectory is less than a safe distance at a certain time step within the prediction time domain... If so, it is judged that a collision will occur in the future;

[0085] In some preferred embodiments, to make the most efficient use of computing resources, when assessing collision risk, each UAV only focuses on the time step in which a collision is first detected within the prediction time domain. And prioritize resolving this most pressing conflict;

[0086] S4.3: Collision avoidance constraint generation. Additional constraints are dynamically generated for neighboring drones at risk of collision only when an inter-drone collision risk is detected. Specifically, for neighboring drones at risk of collision, at the collision time step... Then, for each time step in the prediction time domain, a linear separating hyperplane constraint (i.e., inter-machine collision avoidance constraint) is dynamically constructed. This constraint forces both parties to remain on opposite sides of the hyperplane in the prediction time domain and maintain a minimum safe distance.

[0087] More specifically, the hyperplane is designed between the predicted positions of both parties (the current drone and a neighboring drone at risk of collision), and its normal vector... Typically, the constraints generated along the line connecting the two machines require both parties to be within the same time step. The positions are maintained on both sides of the hyperplane, and at least a minimum safe buffer distance is maintained. The buffer distance should take into account the relative speed and should not be less than the minimum safe distance of the UAV. The collision avoidance constraints between machines are expressed as follows:

[0088] ;

[0089] In the formula, and They represent drones For neighboring drones At time step The predicted location.

[0090] S5: Set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraint, inter-UAV collision avoidance constraint and UAV dynamic model to construct an optimization problem; solve to obtain the optimal predictive control input sequence for each UAV at the current time, and each UAV executes its first optimal control command.

[0091] Specifically, step S5 includes:

[0092] S5.1: Problem formulation: Each UAV predicts the time domain based on its own state prediction equation and the assumed trajectories of its neighboring UAVs. Control input sequence within Using the objective cost function as the decision variable, the local convex safety sub-pipeline constraint, inter-machine collision avoidance constraint, and UAV dynamics model are integrated to form a constrained time-domain optimization problem. In this embodiment, the objective cost function includes a trajectory tracking error term, a control input energy term, and a control increment smoothing term. It should be understood that the above-mentioned objective cost function is exemplary and not a limitation on the scope of protection. The construction of the objective function can be flexibly designed according to specific task requirements. Other common construction methods include, but are not limited to, energy efficiency terms and formation maintenance terms.

[0093] S5.2: Solve, execute and update. Solve the optimization problem for each UAV to obtain the optimal predictive control input sequence and the corresponding optimal predictive state trajectory for each UAV. Each UAV executes its first optimal control command. The optimal predictive state trajectory obtained in this solution is used as the assumed trajectory at the next sampling time.

[0094] S6: Repeat steps S2 to S5 to form a rolling time-domain closed-loop control until all drones pass through the virtual pipeline safely and without collision.

[0095] The following section provides further explanation using a simulation experiment of multi-UAV navigation based on distributed model predictive control and virtual pipelines. The detailed process is as follows: Figure 2 As shown.

[0096] Step 1: System initialization and environment setup.

[0097] like Figure 3 As shown, in a two-dimensional simulation scene, a line of length is constructed. A virtual pipeline 1 is defined with a width of 1.8 meters and a safety buffer zone of 0.4 meters. The central path of the pipeline consists of a set of discrete path points. A cluster of 12 drones 3 is set up, and each drone 3 is configured with the same dynamic and control parameters.

[0098] The UAV adopts a discrete-time double integrator dynamic model, and the state prediction equation is expressed as:

[0099] ;

[0100] in, Let be the state vector of drone i at time t. Let i be the control input vector of UAV i at time t. and These represent the system matrix and control input matrix under the dual integrator dynamics model, respectively. The simulation parameters for the UAV system are set as follows: planning control period h = 0.2 seconds; UAV physical radius... meters, safety radius needs meters, therefore the minimum safe distance meters; maximum speed m / s, maximum acceleration meters per second ² .

[0101] The parameters of the distributed model prediction controller are set as follows: prediction time domain K=15 steps, and the weight matrix of each item in the objective cost function and inequality constraints is defined.

[0102] To simulate a dynamic environment, circular and elliptical dynamic obstacles 4 were introduced into the simulation. Their movement paths intersect with the pipes and they can randomly invade the pipe space.

[0103] Step 2: Construction of objective cost function and constraints.

[0104] In each control cycle, each UAV performs the following core operations to construct its local optimization problem.

[0105] Complete the construction of local pipeline constraints. Based on real-time perception information of their current location, each UAV constructs a locally convex safety sub-pipeline (e.g., based on the pipeline replanning mechanism provided by this invention). Figure 3 (Ref. 5 in the text). Specifically, by sampling and calculating the convex hull and transforming its boundary into a linear inequality constraint through a linear separating hyperplane, the complex non-convex pipe constraint is transformed into a constraint that can be processed in real time.

[0106] Collision avoidance constraint activation. Collision risk detection is performed based on the assumed trajectory obtained from neighboring UAVs. In this simulation experiment, only when a collision risk is predicted, a linear separating hyperplane constraint based on buffered Voronoi units is dynamically generated according to the collision avoidance strategy described in this invention.

[0107] Objective cost function and optimization problem construction. A local model prediction optimization problem is constructed for the UAV within a finite prediction time domain K. The objective cost function is... The design is as follows:

[0108] a. Trajectory tracking error term: Penalizes the positional deviation between the predicted trajectory and the forward-looking target point, causing the UAV to advance along the virtual pipeline; represented as follows:

[0109] ;

[0110] in, This represents the trajectory tracking error term for drone i. These are the weighting coefficients for the trajectory tracking error term. Represents time The predicted location of the drone at time step k. Represents time The target point position at the set time step k.

[0111] b. Input energy control: Penalty control of acceleration amplitude to avoid excessive energy consumption and actuator load; expressed as follows:

[0112] ;

[0113] in, The control input energy term for drone i, and These are the weight matrices for the control input sequence and the control input energy term, respectively.

[0114] c. Control Increment Smoothing Term: Penalizes the change amplitude between adjacent control steps to ensure continuous and smooth UAV movement; expressed as follows:

[0115] ;

[0116] in, For the control increment smoothing term of UAV i, To control the weighting coefficient of the incremental smoothing term, It is in time The input vector is controlled at the predicted time step k.

[0117] The objective cost function is based on the above three factors: .

[0118] Other UAV dynamic constraints involved in the optimization problem constructed in this simulation experiment include, but are not limited to, the following:

[0119] Velocity amplitude constraint: This restricts the speed of the UAV at any given moment from exceeding its maximum allowable value, ensuring that the generated trajectory is dynamically feasible and conforms to the physical limits of the UAV platform, while also contributing to flight stability; expressed as follows:

[0120] ;

[0121] in, Indicates that drone i is in time The velocity vector is composed of velocity components in various directions. This vector represents the maximum permissible velocity value for the drone, and is composed of the maximum velocity values ​​in each direction. It is a velocity weighting coefficient vector, which consists of the weighting coefficients of velocities in each direction.

[0122] Control input amplitude constraint: This restricts the control input of the UAV at any time step, meaning the acceleration magnitude must not exceed its maximum allowable value; expressed as follows:

[0123] ;

[0124] in, Indicates that drone i is at time The control input consists of control input components in each direction. This represents the maximum control input amplitude vector that the actuator can provide, composed of the maximum control input amplitudes in each direction. The input weight coefficient vector is composed of weight coefficients that control the input in each direction.

[0125] Step 3: Optimize the solution and implement closed-loop control.

[0126] Each UAV integrates the aforementioned objective optimization function and all constraints into a quadratic programming problem and calls the solver to solve it independently. After obtaining the optimal predictive control input sequence, it executes its first control command and updates its own state. In the next sampling period, steps 2 and 3 are repeated to form a rolling optimization closed-loop control until all UAVs reach the target point.

[0127] After following the steps above, the following phenomena and data were observed during the simulation of this embodiment.

[0128] (1) Successful obstacle avoidance: When a dynamic obstacle approaches, the affected UAV automatically and smoothly adjusts its trajectory based on the local convex safety sub-pipeline mechanism to form a one-way flow, effectively avoiding collisions with the obstacle. See [reference] Figure 4 .

[0129] (2) Strict boundary adherence: Data shows that the position of all drones throughout the entire flight process remained strictly within the total allowable area of ​​the virtual pipeline. Within, and without crossing the boundary safety buffer zone, see [reference]. Figure 5 .

[0130] (3) Collision-free passage: The minimum distance between any two drones in the cluster is always maintained above 0.4 meters, meeting the set minimum safe distance, proving the effectiveness of the collision avoidance constraint. See [reference needed]. Figure 6 .

[0131] This simulation experiment demonstrates that the method described in this invention can effectively coordinate dynamic pipeline constraints and on-demand collision avoidance strategies through a distributed model predictive control framework, ultimately achieving safe, efficient, and smooth navigation of UAV swarms in dynamic and uncertain environments.

[0132] This invention also provides a multi-UAV navigation system based on distributed model predictive control and virtual pipelines, comprising:

[0133] The pipeline construction module is used to build virtual pipelines, forming a flyable area for drone swarms;

[0134] The collaborative communication and trajectory hypothesis module is used to generate a hypothetical trajectory in the prediction time domain based on the current state of each UAV using its dynamic model in each sampling period, and broadcast it to neighboring UAVs.

[0135] The Local Convex Safe Sub-Pipe Constraint Generation Module is used to perform obstacle intrusion judgment and virtual pipeline dynamic processing, construct a linear separating hyperplane, and generate local convex safe sub-pipe constraints.

[0136] The inter-drone collision avoidance constraint activation module is used for each drone to obtain the assumed trajectory information of neighboring drones, to judge the inter-drone collision risk, and to activate the inter-drone collision avoidance constraint only when a collision risk is detected.

[0137] The local prediction and rolling optimization module is used to set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraints, inter-UAV collision avoidance constraints and UAV dynamics model to construct an optimization problem; solve for the optimal predictive control input sequence of each UAV at the current time, and each UAV executes its first optimal control command;

[0138] The collaborative communication and trajectory assumption module, the local convex safety sub-pipe constraint generation module, the inter-machine collision avoidance constraint activation module, and the local prediction and rolling optimization module are executed in a loop until all UAVs pass through the virtual pipe safely and without collision.

[0139] It should be understood that the functional unit modules in the various embodiments of the present invention can be concentrated in one processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated into one unit module, and can be implemented in hardware or software.

[0140] This invention also provides an electronic device, comprising:

[0141] A memory on which computer programs are stored;

[0142] A processor is used to load and execute the computer program to implement the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described above.

[0143] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described above.

[0144] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0145] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0149] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A multi-UAV navigation method based on distributed model predictive control and virtual pipeline, characterized in that, Includes the following steps: S1: Construct virtual pipelines to form a flyable area for drone swarms; S2: In each sampling period, each UAV generates a hypothetical trajectory in the predicted time domain based on its current state using its dynamic model, and broadcasts it to neighboring UAVs. S3: Perform obstacle intrusion detection and virtual pipeline dynamic processing, construct a linear separation hyperplane, and generate local convex safety sub-pipeline constraints; Step S3 includes: S3.1: Pipeline status judgment and boundary acquisition. Starting from the current position of the UAV, it is determined whether the predefined virtual pipeline is intruded by an obstacle within the current perception range of the UAV. If an obstacle intrusion is detected, the pipeline replanning mechanism is triggered. Based on the obstacle position information, the original virtual pipeline boundary is recalculated to generate a new local pipeline boundary that avoids the obstacle. The geometry of the new local pipeline boundary remains smooth and continuous with the unaffected global virtual pipeline boundary. S3.2: Generation of a locally convex safe sub-pipeline. Combine the virtual pipe boundary obtained in step S3.1 with the current perception range of the UAV to construct a locally convex safe sub-pipeline. S3.3: Constraint linearization, using a linear separating hyperplane, describes the boundary of the local convex safety sub-pipe as a set of linear inequality constraints; S4: Each drone acquires the assumed trajectory information of its neighboring drones, performs an inter-drone collision risk assessment, and activates inter-drone collision avoidance constraints only when a collision risk is detected. Step S4 includes: S4.1: Neighbor information exchange, each UAV receives the assumed trajectory of all neighboring UAVs in the previous cycle through the communication network, as the known input for its own optimization in the current cycle; S4.2: Collision risk assessment. Under the distributed framework, each UAV performs time-step inter-UAV collision risk detection in the prediction time domain based on its own assumed trajectory and the assumed trajectories of neighboring UAVs. S4.3: Collision avoidance constraint generation. Only when an inter-drone collision risk is detected, a linear separating hyperplane constraint is dynamically constructed for the neighboring drones with collision risk. This constraint forces both drones to stay on opposite sides of the hyperplane in the prediction time domain and maintain a minimum safe distance. S5: Set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraint, inter-UAV collision avoidance constraint and UAV dynamic model to construct an optimization problem; solve to obtain the optimal predictive control input sequence for each UAV at the current time, and each UAV executes its first optimal control command; S6: Repeat steps S2 to S5 until all drones pass through the virtual pipeline safely and without collision.

2. The multi-UAV navigation method based on distributed model predictive control and virtual pipeline according to claim 1, characterized in that, Step S1 includes: S1.1: Geometric definition of virtual pipe, virtual pipe It is defined by a path and the safe zone formed by extending the normal vector of that path. S1.2: Definition of safe avoidance zone: A safe avoidance zone is introduced outside the boundary of the virtual pipeline. The width of the safe avoidance zone shall not be less than the minimum safe distance for the drone; virtual pipeline and safe avoidance area The union of these elements constitutes the flyable area of ​​the drone swarm. .

3. The multi-UAV navigation method based on distributed model predictive control and virtual pipeline according to claim 1, characterized in that, Step S2 includes: S2.1: Prediction model construction: Based on the dynamic model of the UAV, construct the state prediction equations of each UAV in the prediction time domain; S2.2: Assuming trajectory generation and communication, each UAV generates a hypothetical trajectory based on its current state through a state prediction equation and broadcasts it to neighboring UAVs; under the initial state of the UAV, the predicted trajectory is constructed using the initial position information as the hypothetical trajectory for the previous cycle.

4. The multi-UAV navigation method based on distributed model predictive control and virtual pipeline according to claim 1, characterized in that, When assessing collision risk, each drone focuses only on the time step in which a collision is detected within the prediction time domain.

5. The multi-UAV navigation method based on distributed model predictive control and virtual pipeline according to claim 1, characterized in that, Step S5 includes: S5.1: Problem formulation: Each UAV predicts the time domain based on its own state prediction equation and the assumed trajectories of its neighboring UAVs. Control input sequence within Using the objective cost function, local convex safety sub-pipeline constraints, inter-machine collision avoidance constraints, and UAV dynamics model as decision variables, a constrained time-domain optimization problem is formed. S5.2: Solve, execute and update. Solve the optimization problem for each UAV to obtain the optimal predictive control input sequence and the corresponding optimal predictive state trajectory for each UAV. Each UAV executes its first optimal control command. The optimal predictive state trajectory obtained in this solution is used as the assumed trajectory at the next sampling time.

6. A multi-UAV navigation system based on distributed model predictive control and virtual pipeline, characterized in that, For implementing the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described in any one of claims 1 to 5, the system comprises: The pipeline construction module is used to build virtual pipelines, forming a flyable area for drone swarms; The collaborative communication and trajectory hypothesis module is used to generate a hypothetical trajectory in the prediction time domain based on the current state of each UAV using its dynamic model in each sampling period, and broadcast it to neighboring UAVs. The Local Convex Safe Sub-Pipe Constraint Generation Module is used to perform obstacle intrusion judgment and virtual pipeline dynamic processing, construct a linear separating hyperplane, and generate local convex safe sub-pipe constraints. The inter-drone collision avoidance constraint activation module is used for each drone to obtain the assumed trajectory information of neighboring drones, to judge the inter-drone collision risk, and to activate the inter-drone collision avoidance constraint only when a collision risk is detected. The local prediction and rolling optimization module is used to set the target cost function for each UAV, and combine the target cost function with the local convex safety sub-pipe constraints, inter-UAV collision avoidance constraints and UAV dynamics model to construct an optimization problem; solve for the optimal predictive control input sequence of each UAV at the current time, and each UAV executes its first optimal control command; The collaborative communication and trajectory assumption module, the local convex safety sub-pipe constraint generation module, the inter-machine collision avoidance constraint activation module, and the local prediction and rolling optimization module are executed in a loop until all UAVs pass through the virtual pipe safely and without collision.

7. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor is configured to load and execute the computer program to implement the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-UAV navigation method based on distributed model predictive control and virtual pipeline as described in any one of claims 1 to 5.

Images