A distributed safety reconfiguration control method and system for a multi-unmanned vehicle inspection system triggered by a specific region.

Abstract

This application discloses a distributed safety reconfiguration control method and system for a multi-unmanned vehicle (UAV) inspection system triggered by a region, belonging to the field of multi-UAV platoon control technology. The method includes: determining a reference path and a triggering region within the reference path; obtaining a matrix of the desired formation based on the desired pose of the follower relative to the virtual navigator; tracking the reference path through the virtual navigator and detecting the triggering region; in response to the virtual navigator being located in the triggering region, obtaining the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation; obtaining the MPC controllers of the virtual navigator and the follower; solving the MPC controllers using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence; controlling the follower's actions using the desired pose and the optimal control sequence to reconfigure the multi-UAV inspection system to the desired formation and maintain the desired formation during inspection.

Description

Technical Field

[0001] This application belongs to the field of multi-unmanned vehicle platoon control technology, specifically involving a distributed safety reconfiguration control method and system for a region-triggered multi-unmanned vehicle inspection system. Background Technology

[0002] With the rapid development of unmanned systems, their working mode has evolved from a single system to a multi-node collaborative mode. Compared with traditional single mobile robots, swarm formations have many advantages, such as reducing overall system cost, improving system robustness and efficiency, wide application scenarios, and higher environmental adaptability. They have received considerable attention in recent years. Research on the collaborative formation control of multiple mobile robots is particularly important in fire search and rescue operations, where a single robot struggles to efficiently complete complex tasks such as multi-target search and real-time information transmission.

[0003] Multi-robot collaborative formation offers numerous advantages and has a wide range of applications, extending beyond mere formation changes to be more closely connected to real-life situations. This design focuses on safety inspections. Traditional inspection markers are typically installed manually at fixed points, resulting in limited flexibility and fixed locations. In actual inspection operations, personnel need to identify, locate, and detect inspection markers at different distances and orientations to meet the requirements of both inspection efficiency and recognition accuracy.

[0004] Over the past decade, researchers have developed various formation control strategies for the cooperative control of multiple mobile robots, such as the leader-follower method, the virtual structure method, behavior-based methods, and consistency-based methods. Each method has its advantages and disadvantages. The leader-follower method has been the most extensively studied; however, if the leader fails, the entire system will collapse. The virtual structure method treats the entire formation as a whole, with each vehicle corresponding to a node; however, this method cannot flexibly change formations. For behavior-based methods, effectively avoiding obstacles becomes a core issue during target tracking. While consistency-based methods are robust and theoretically well-proven, they are not suitable for fast-response scenarios. Therefore, researchers typically choose the appropriate control method based on the specific circumstances.

[0005] In summary, there is an urgent need for a formation control strategy that integrates high efficiency, strong robustness, and low computational complexity to address the current problems in multi-mobile robot formation control, such as the risk of single-point failure of the navigator, insufficient flexibility in formation changes, difficulties in cooperative obstacle avoidance, and heavy computational burden of centralized control. Especially in dynamic application scenarios such as safety inspection and marker placement, a collaborative control method that balances formation stability, environmental adaptability, and real-time response performance is needed to achieve high-precision and high-efficiency collaborative movement of multiple vehicles in complex terrains. Summary of the Invention

[0006] Purpose of the invention: This application develops a distributed safety reconfiguration control method and system for a multi-unmanned vehicle inspection system triggered by a region, aiming to solve the technical problems in the prior art.

[0007] Technical Solution: In a first aspect, embodiments of this application provide a distributed safety reconfiguration control method for a region-triggered multi-unmanned vehicle inspection system, applied to a multi-unmanned vehicle inspection system including a virtual navigator and multiple followers, the control method comprising:

[0008] Determine a reference path, and then determine the trigger area within that reference path;

[0009] The matrix for obtaining the desired formation is obtained based on the desired pose of the follower relative to the virtual navigator;

[0010] The virtual navigator tracks the reference path and performs trigger area detection.

[0011] In response to the virtual navigator being located in the triggering area, the desired pose of the follower is obtained based on the real-time pose of the virtual navigator and the matrix of the desired formation;

[0012] Obtain the MPC controller of the virtual navigator and the MPC controller of the follower;

[0013] The optimal control sequence is obtained by solving the MPC controller using the IPOPT nonlinear optimization problem solver.

[0014] By controlling the follower's desired pose and the optimal control sequence, the follower's actions are reconstructed into the desired formation of the multi-unmanned vehicle inspection system, and the inspection is carried out while maintaining the desired formation.

[0015] In some embodiments, the step of obtaining the follower's MPC controller includes:

[0016] Obtain the state constraint function before the follower reaches the predicted time domain endpoint, and the terminal constraint function when the follower reaches the predicted time domain endpoint;

[0017] Establish collision avoidance soft constraints based on the real-time distance and safe distance between the followers;

[0018] In response to maintaining the desired formation, the MPC controller of the follower is obtained based on the state constraint function and the terminal constraint function;

[0019] In response to reconstructing the desired formation, the MPC controller of the follower is obtained based on the state constraint function, the terminal constraint function, and the collision avoidance soft constraint.

[0020] In some embodiments, the characterization formula of the follower's MPC controller includes:

[0021] ;

[0022] in, The total cost function of the follower's MPC controller; The state constraint function is defined for the follower before it reaches the predicted time domain endpoint. The terminal constraint function is the function used when the follower reaches the predicted time domain endpoint. The cost function for maintaining the desired formation for the followers. This refers to the collision avoidance soft constraint; The parameters are transformed by the cost function. This indicates that the desired formation should be maintained. This indicates the desired formation to be reconstructed.

[0023] In some embodiments, the characterization formula of the state constraint function includes:

[0024] ;

[0025] in, The state constraint function; The state error at time step k; This is the weight matrix for the state error; This is the control input at time step k; The weight matrix for controlling the input; To predict the end point in the time domain;

[0026] The characterization formula of the terminal constraint function includes:

[0027] ;

[0028] in, The terminal constraint function; To predict the endpoint in the time domain Terminal state error; This is the weight matrix for the terminal state error.

[0029] In some embodiments, the characterization formula of the collision avoidance soft constraint includes:

[0030] ;

[0031] in, This refers to the collision avoidance soft constraint; The real-time distance between the followers; The safe distance between the followers; The weighting coefficient for obstacle avoidance penalty is such that the larger the value, the greater the constraint, and vice versa.

[0032] In some embodiments, the MPC controller of the virtual navigator is obtained based on the state constraint function and the terminal constraint function, and its characterization formula includes:

[0033] ;

[0034] in, The total cost function of the MPC controller of the virtual navigator.

[0035] In some embodiments, the step of obtaining the optimal control sequence includes:

[0036] Obtain the input linear velocity boundary and input angular velocity boundary of the follower, and construct physical constraints:

[0037] ;

[0038] in, , The solver for the first The control inputs for the linear velocity and angular velocity of the first follower are used to characterize the first... The expected linear velocity and expected angular velocity of each follower; , For the first The minimum and maximum expected linear velocities of each follower; , For the first The minimum and maximum expected angular velocities of the autonomous vehicle;

[0039] The optimal control sequence is obtained based on the physical constraints:

[0040] ;

[0041] in, To control the time domain; In order to be in Always The predicted result of the linear velocity output at time t; In order to be in Always The predicted result of the angular velocity output at time t.

[0042] In some embodiments, the representation formula for the matrix of the desired formation includes:

[0043] ;

[0044] in, Let be the matrix of the desired formation, where the first row is the formation parameters of the virtual navigator, and the remaining rows are the formation parameters of the followers; i = 2…n, where i is the follower, and 2…n are the identification markers of the follower i; The expected distance between the follower i and the virtual navigator; Let i be the desired angle between the follower i and the virtual navigator.

[0045] In some embodiments, a discrete model description is performed based on the forward Euler method to obtain the kinematic model of the multi-unmanned vehicle inspection system. The representation formula of the kinematic model includes:

[0046] ;

[0047] in, For the unmanned vehicle in the first Position coordinates at each moment; For the unmanned vehicle in the first Position coordinates at each moment; For driverless cars in the first The orientation angle at any given moment; Indicates that driverless cars are in the first The linear velocity at time t; For driverless cars in the first angular velocity at time t; This indicates the interval between sampling times.

[0048] Secondly, embodiments of this application also provide a region-triggered distributed safety reconfiguration control system for a multi-unmanned vehicle inspection system, applied to a multi-unmanned vehicle inspection system including a virtual navigator and multiple followers, the system comprising:

[0049] A path acquisition module, which is used to determine a reference path and a trigger area within the reference path;

[0050] A formation acquisition module, which is used to acquire a matrix of desired formations based on the desired pose of the followers relative to the virtual navigator;

[0051] A navigator motion module, which is used to track the reference path through the virtual navigator and perform trigger area detection;

[0052] The follower pose module is used to obtain the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation in response to the virtual navigator being located in the trigger area.

[0053] A controller acquisition module is used to acquire the MPC controller of the virtual navigator and the MPC controller of the follower.

[0054] A control sequence acquisition module is used to solve the MPC controller using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence.

[0055] The formation control module is used to control the movement of the follower according to the desired pose of the follower, reconstruct the multi-unmanned vehicle inspection system to the desired formation, and maintain the desired formation for inspection.

[0056] Beneficial Effects: Compared with the prior art, the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system with regional triggering provided in this application includes determining a reference path and a triggering region within the reference path; obtaining a matrix of the desired formation based on the desired pose of the follower relative to the virtual navigator; tracking the reference path through the virtual navigator and detecting the triggering region; in response to the virtual navigator being located in the triggering region, obtaining the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation; obtaining the MPC controller of the virtual navigator and the MPC controller of the follower; solving the MPC controller using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence; controlling the follower's actions using the desired pose of the follower and the optimal control sequence to reconfigure the multi-unmanned vehicle inspection system to the desired formation and maintain the desired formation for inspection. This application employs a multi-unmanned vehicle (UAV) platooning control framework based on "distributed model prediction + region triggering." It specifically addresses challenges such as difficulty in forming, maintaining, and reconfiguring multi-UAV platoons under different environments, weak collaborative collision avoidance among multiple vehicles, and single-point failure risks, insufficient constraints, and weak environmental adaptability in traditional lead-follow platooning. This framework improves platooning coordination accuracy and safety, operational reliability, and reduces the computational load of platooning control. Simultaneously, a virtual lead-follow platooning algorithm is used to construct the required formation, reducing formation transformation time. Distributed model predictive control is adopted, with each vehicle activating its own MPC control node to solve problems in parallel for each vehicle, shortening computation time. A region triggering mechanism is designed to make the increase in distance constraints controllable, reducing the computational load on individual robots. This significantly improves the efficiency of multi-UAV platooning formation, maintenance, and reconfiguration, as well as its safety and reliability during operation. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart illustrating the steps of the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in an embodiment of this application.

[0059] Figure 2 A flowchart illustrating the steps of obtaining the MPC controller of the follower in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this embodiment of the application.

[0060] Figure 3 A flowchart illustrating the steps for obtaining the optimal control sequence in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this embodiment of the application.

[0061] Figure 4 This is a module connection diagram of the distributed safety reconfiguration control system for a multi-unmanned vehicle inspection system triggered by a region, as provided in an embodiment of this application.

[0062] Figure 5 A schematic diagram illustrating the structure of the virtual navigation and following formation in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by the region provided in this application embodiment;

[0063] Figure 6 A schematic diagram of the formation library in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in the embodiments of this application;

[0064] Figure 7 A Gazebo simulation scene diagram showing the initial stage where four vehicles maintain a straight line and the signpost is lowered in a controllable manner.

[0065] Figure 8 A Gazebo simulation scene diagram to trigger the switch of the autonomous vehicle cluster from a straight formation to a parallelogram formation in Phase 1 of the region;

[0066] Figure 9 This is a diagram showing the switching positions of the parallelogram.

[0067] Figure 10 A Gazebo simulation scene diagram to trigger the transition of the autonomous vehicle cluster from a straight formation to a trapezoidal formation in Phase 2 of the region;

[0068] Figure 11 This is a diagram showing the trapezoidal switching position.

[0069] Figure 12A Gazebo simulation scene diagram showing the transition of the autonomous vehicle cluster from a straight formation to a line formation in the third stage of the trigger area.

[0070] Figure 13 This is a diagram showing the location of the linear switching pattern;

[0071] Attached reference numerals: 10, Path acquisition module; 20, Formation acquisition module; 30, Navigator motion module; 40, Follower pose module; 50, Controller acquisition module; 60, Control sequence acquisition module; 70, Formation control module. Detailed Implementation

[0072] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0073] With the rapid development of unmanned systems, their working mode has evolved from a single system to a multi-node collaborative mode. Compared with traditional single mobile robots, swarm formations have many advantages, such as reducing overall system cost, improving system robustness and efficiency, wide application scenarios, and higher environmental adaptability. They have received considerable attention in recent years. Research on the collaborative formation control of multiple mobile robots is particularly important in fire search and rescue operations, where a single robot struggles to efficiently complete complex tasks such as multi-target search and real-time information transmission.

[0074] Multi-robot collaborative formation offers numerous advantages and has a wide range of applications, extending beyond mere formation changes to be more closely connected to real-life situations. This design focuses on safety inspections. Traditional inspection markers are typically deployed manually at fixed points, resulting in fixed locations and limited flexibility. In actual inspection operations, personnel need to identify, locate, and detect inspection markers at different distances and orientations to meet the requirements of both inspection efficiency and recognition accuracy. Based on these needs, this design proposes using a four-wheeled differential speed mobile robot equipped with controllable inspection markers as a platform, applying multi-vehicle formation control technology to an intelligent inspection system.

[0075] Over the past decade, researchers have developed various formation control strategies for the cooperative control of multiple mobile robots, such as the leader-follower method, the virtual structure method, behavior-based methods, and consistency-based methods. Each method has its advantages and disadvantages. The leader-follower method has been the most extensively studied; however, if the leader fails, the entire system will collapse. The virtual structure method treats the entire formation as a whole, with each vehicle corresponding to a node; however, this method cannot flexibly change formations. For behavior-based methods, effectively avoiding obstacles becomes a core issue during target tracking. While consistency-based methods are robust and theoretically well-proven, they are not suitable for fast-response scenarios. Therefore, researchers typically choose the appropriate control method based on the specific circumstances.

[0076] Model Predictive Control (MPC) is further divided into centralized MPC and distributed MPC (DMPC). Centralized MPC considers the entire system and constructs performance index functions for all robots. Although centralized MPC can solve some formation problems, it requires complex calculations for each rolling time-domain update. Distributed MPC (DMPC), on the other hand, solves multiple small-dimensional distributed optimization problems. Each robot is equipped with an independent MPC controller, making the calculations for each rolling time-domain update relatively simple.

[0077] In summary, there is an urgent need for a formation control strategy that integrates high efficiency, strong robustness, and low computational complexity to address the current problems in multi-mobile robot formation control, such as the risk of single-point failure of the navigator, insufficient flexibility in formation changes, difficulties in cooperative obstacle avoidance, and heavy computational burden of centralized control. Especially in dynamic application scenarios such as safety inspection and signage erection, a collaborative control method that balances formation stability, environmental adaptability, and real-time response performance is needed to achieve high-precision and high-efficiency collaborative movement of multiple vehicles in complex terrains.

[0078] In view of this, this application provides a distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, including determining a reference path and a triggering region within the reference path; obtaining a matrix of the desired formation based on the desired pose of the follower relative to the virtual navigator; tracking the reference path through the virtual navigator and detecting the triggering region; in response to the virtual navigator being located in the triggering region, obtaining the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation; obtaining the MPC controllers of the virtual navigator and the follower; solving the MPC controllers using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence; controlling the follower's actions using the desired pose and the optimal control sequence to reconfigure the multi-unmanned vehicle inspection system to the desired formation and maintain the desired formation during inspection. This application employs a multi-unmanned vehicle (UAV) platooning control framework based on "distributed model prediction + region triggering." It specifically addresses challenges such as difficulty in forming, maintaining, and reconfiguring multi-UAV platoons under different environments, weak collaborative collision avoidance among multiple vehicles, and single-point failure risks, insufficient constraints, and weak environmental adaptability in traditional lead-follow platooning. This framework improves platooning coordination accuracy and safety, operational reliability, and reduces the computational load of platooning control. Simultaneously, a virtual lead-follow platooning algorithm is used to construct the required formation, reducing formation transformation time. Distributed model predictive control is adopted, with each vehicle activating its own MPC control node to solve problems in parallel for each vehicle, shortening computation time. A region triggering mechanism is designed to make the increase in distance constraints controllable, reducing the computational load on individual robots. This significantly improves the efficiency of multi-UAV platooning formation, maintenance, and reconfiguration, as well as its safety and reliability during operation.

[0079] In some embodiments, the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by the region provided in this application is applied to a multi-unmanned vehicle inspection system including a virtual navigator and multiple followers. Specifically, this application defines the multi-unmanned vehicle system as "1 virtual navigator + 3 followers". All unmanned vehicles are four-wheeled differential cars with controllable steers, with consistent specifications and performance, focusing on the formation control of the unmanned vehicle system and collision avoidance between systems.

[0080] In some embodiments, this application uses the forward Euler method to perform discrete model description to obtain the kinematic model of the multi-unmanned vehicle inspection system. The representation formula of the kinematic model includes:

[0081] ;

[0082] in, For driverless cars in the first Position coordinates at each moment; For driverless cars in the first Position coordinates at each moment; For driverless cars in the first The orientation angle at any given moment; Indicates that driverless cars are in the first The linear velocity at time t; For driverless cars in the first angular velocity at time t; This indicates the interval between sampling times.

[0083] In some embodiments, to avoid control misalignment, this application considers the actual situation of the vehicle, limits linear velocity and angular velocity, and constructs system constraints, the characterization formulas of which include:

[0084] ;

[0085] in, , The solver for the first The control inputs for the linear velocity and angular velocity of the first follower are used to characterize the first... The expected linear velocity and expected angular velocity of each follower; , For the first The minimum and maximum expected linear velocities of each follower; , For the first The minimum and maximum expected angular velocities of the autonomous vehicle.

[0086] In some embodiments, please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this application embodiment. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this application embodiment, is specifically implemented through steps 100 to 700:

[0087] Step 100: Determine the reference path and identify the trigger area within the reference path.

[0088] In some embodiments, this application establishes a region triggering mechanism, specifying several trigger regions in the reference path to trigger formation switching and changes in the controllable beacon state. The virtual navigator in this application only tracks the path, and the representation formula of the reference path includes:

[0089] ;

[0090] in, This is the reference path after convergence; This is a reference path that has not converged; For path parameters, it is expressed by differential equations describe, It is not a priori known quantity; it does not contain time information, and its reference value needs to be calculated at each sampling time.

[0091] Furthermore, this application designs an MPC controller that enables the virtual navigator's pose state to converge to the reference path. :

[0092] ;

[0093] Understandably, after the virtual navigator reaches the designated trigger area, a formation change is triggered. At this time, the desired position of the followers will change, guiding them to move towards the desired position. The designed MPC controller for the followers needs to consider the distance constraints of adjacent vehicles to ensure safety during the formation change process. After the area is triggered... The parameter values ​​will change, and the distance constraints will be dynamically introduced, thereby ensuring collision avoidance between vehicles and improving the speed of MPC solution.

[0094] Step 200: Obtain the matrix of the desired formation based on the desired pose of the follower relative to the virtual navigator.

[0095] In some embodiments, this application considers a multi-vehicle system consisting of n four-wheel differential vehicle robots (UGVs), whose index is The position of the virtual navigator R0 is determined by the reference trajectory point, and the other robots act as followers R1, R2, ..., Rn. The four-wheeled vehicle model is constrained by its own mechanical kinematics; its inputs are only angular velocity and linear velocity, and its controlled degrees of freedom are... , , This is a non-integrity constraint. The pose of the vehicle in the inertial coordinate system is defined as... The kinematic model is as follows;

[0096] ;

[0097] Due to the physical constraints of the UGV, it is difficult to install the positioning sensor at its geometric center. This application assigns an offset point along the x-axis to the i-th UGV. Distance from the geometric center The location indicates its actual position; Indicates the first The angle of the small car in the global coordinate system Indicates the first The speed of the small car in the global coordinate system. Indicates the first The desired position of the small car. Indicates the first Taiwanese cars and virtual navigators Expected relative distance Indicates the first Taiwanese cars and virtual navigators The relative angle of expectation;

[0098] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure analysis of the virtual navigator following formation in the distributed safety reconfiguration control method of the multi-unmanned vehicle inspection system with regional triggering provided in the embodiments of this application. It defines the distance and angle between the follower and the virtual navigator (in...). (in coordinate system) For followers To Virtual Navigator The relative distance between them, i.e. reference point to The distance between the center points of the two vehicles is such that, in order to prevent a collision between the two vehicles, . yes and The difference in the direction angle of motion describes the relative position and direction between the two robots;

[0099] ;

[0100] ;

[0101] ;

[0102] ;

[0103] Define the desired distance of the follower in the virtual navigator coordinate system as: , angle is . No. The desired position of the car is ;

[0104]

[0105] The virtual navigator, based on an algorithm, employs a "distance-angle" strategy and obtains the matrix definition of the desired formation through geometric analysis:

[0106] ;

[0107] ;

[0108] in, The matrix represents the desired formation; A unique ID for virtual navigators or followers; The expected distance between follower i and the virtual navigator; The expected perspective of the follower i and the virtual navigator.

[0109] Furthermore, the formula for representing the matrix of the desired formation includes:

[0110] ;

[0111] in, To construct the matrix of the desired formation, this application overlaps the virtual navigator R0 with the follower R1 to build a formation library, where the first row is the formation parameters of the virtual navigator and the remaining rows are the formation parameters of the follower; i=2…n, where i is the follower and 2…n are the identification markers of follower i; The expected distance between follower i and the virtual navigator; The expected perspective of the follower i and the virtual navigator.

[0112] Specifically, please refer to Figure 6 , Figure 6 This is a schematic diagram of the formation library in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, provided in an embodiment of this application. In this application, the virtual navigator R0 and follower R1 are overlapped to construct the formation library, including straight lines, parallelograms, trapezoids, and linear formations.

[0113] ;

[0114] ;

[0115] ;

[0116] .

[0117] Step 300: Track the reference path using the virtual navigator and perform trigger area detection.

[0118] Step 400: In response to the virtual navigator being located in the trigger area, obtain the desired pose of the followers based on the matrix of the virtual navigator's real-time pose and desired formation.

[0119] Step 500: Obtain the MPC controller for the virtual navigator and the MPC controller for the follower.

[0120] In some embodiments, this application adopts a distributed MPC controller design. During the task, the virtual navigator is only responsible for tracking the reference path and detecting the trigger area, without considering collision avoidance between vehicles; the followers are responsible for maintaining, reconstructing and restoring the formation, and need to consider vehicle collisions.

[0121] In some embodiments, please refer to Figure 2 , Figure 2This is a flowchart illustrating the steps of obtaining the MPC controller of the follower in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this application embodiment. The method for obtaining the MPC controller of the follower in this application is specifically implemented through steps 510 to 540:

[0122] Step 510: Obtain the state constraint function before the follower reaches the end of the prediction time domain, and the terminal constraint function when the follower reaches the end of the prediction time domain.

[0123] In some embodiments, the characterization formula of the state constraint function includes:

[0124] ;

[0125] in, Here are the state constraint functions; The state error at time step k; This is the weight matrix for the state error; This is the control input at time step k; The weight matrix for controlling the input; To predict the end point in the time domain.

[0126] In some embodiments, this application considers the state of the robot reaching the endpoint, which can enable the system to eventually converge to the desired terminal state, thereby ensuring the transformation and preservation of the formation without significant distortion. The representation formula for the terminal constraint function includes:

[0127] ;

[0128] in, For terminal constraint functions; To predict the endpoint in the time domain Terminal state error; This is the weight matrix for the terminal state error;

[0129] Step 520: Establish collision avoidance soft constraints based on the real-time distance and safe distance between followers.

[0130] In some embodiments, the followers are responsible for maintaining, reconfiguring, and restoring the formation. Vehicle collisions need to be considered. This application designs soft collision avoidance constraints to ensure the safe operation of the system.

[0131] ;

[0132] ;

[0133] in, To avoid collisions with soft constraints; The weighting coefficient for obstacle avoidance penalty is such that the larger the value, the greater the constraint, and vice versa. It is a barrier function; To maintain a safe distance between followers; This represents the real-time distance between followers.

[0134] Furthermore, follower autonomous vehicles exist Time and the Real-time Euclidean distance of an autonomous vehicle , is represented as:

[0135] ;

[0136] in, For the first The first follower The position at that moment; For the first The first driverless car The position at any given moment; only when the real-time distance between the two vehicles is less than [the specified value]. Only then will it be subject to distance constraints from adjacent vehicles. The overall soft collision avoidance constraint is characterized by the following formulas:

[0137] ;

[0138] in, To avoid collisions with soft constraints; Real-time distance between followers; To maintain a safe distance between followers; The weighting coefficient for obstacle avoidance penalty is such that the larger the value, the greater the constraint, and vice versa.

[0139] Step 530: In response to maintaining the desired formation, obtain the MPC controller for the follower based on the state constraint function and the terminal constraint function.

[0140] Step 540: In response to reconstructing the desired formation, obtain the MPC controller for the follower based on the state constraint function, terminal constraint function, and collision avoidance soft constraint.

[0141] In some embodiments, the characterization formula for the follower's MPC controller includes:

[0142] ;

[0143] in, The total cost function of the MPC controller for the follower; The state constraint function is defined for the follower before it reaches the predicted time domain endpoint. This is the terminal constraint function when the follower reaches the predicted time domain endpoint. The cost function for maintaining the desired formation for followers. To avoid collisions with soft constraints; The parameters are transformed by the cost function. This indicates maintaining the desired formation. This indicates the desired reconfiguration formation.

[0144] In some embodiments, since the virtual navigator only considers the tracking of the reference path and does not consider collisions between vehicles, the representation formula of the virtual navigator's cost function includes:

[0145] ;

[0146] in, The total cost function for the MPC controller of the virtual navigator.

[0147] Step 600: Solve the MPC controller using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence.

[0148] In some embodiments, the overall objective functions of the virtual navigator and the followers differ; the overall objective function of the virtual navigator is to obtain... The overall objective function of the followers is to obtain .

[0149] In some embodiments, please refer to Figure 3 , Figure 3 This is a flowchart illustrating the steps for obtaining the optimal control sequence in the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as provided in this application embodiment. The method for obtaining the optimal control sequence in this application is specifically implemented through steps 610 to 620:

[0150] Step 610: Obtain the input linear velocity boundary and input angular velocity boundary of the follower, and construct physical constraints.

[0151] In some embodiments, the formula for characterizing physical constraints includes:

[0152] ;

[0153] in, , The solver for the first The control inputs for the linear velocity and angular velocity of the first follower are used to characterize the first... The expected linear velocity and expected angular velocity of each follower; , For the first The minimum and maximum expected linear velocities of each follower; , For the first The minimum and maximum expected angular velocities of the autonomous vehicle;

[0154] Step 620: Obtain the optimal control sequence based on physical constraints.

[0155] In some embodiments, the formula for representing the optimal control sequence includes:

[0156] ;

[0157] in, To control the time domain; In order to be in Always The predicted result of the linear velocity output at time t; In order to be in Always The predicted result of the angular velocity output at time t.

[0158] Step 700: Control the follower's actions by using the follower's desired pose and optimal control sequence, reconstruct the multi-unmanned vehicle inspection system to the desired formation, and maintain the desired formation for inspection.

[0159] For example, this application uses the Ubuntu 20.04 operating system, the ROS Noetic framework, and the Gazebo simulation platform to build the experimental environment. A four-wheel differential belt-mounted unmanned vehicle is selected, with the virtual navigator vehicle labeled UGV1 and the three follower vehicles labeled UGV2, UGV3, and UGV4 respectively. The experimental parameters are configured as shown in Table 1, and the core parameters are as follows:

[0160] The initial pose of the virtual navigator UGV1 autonomous vehicle is set with preset initial x and y coordinates and angles in the global coordinate system; the initial poses of the follower UGV2-UGV4 autonomous vehicles are set to their initial positions around the virtual navigator, ensuring no collisions in the initial state; three area trigger points are set, all within the reference path; the number of predicted steps is also specified. Safety radius Sampling time Number of neighbors of driverless cars and the duration of formation All settings were configured according to the experimental requirements to ensure a balance between formation maintenance and collision avoidance priorities within the system.

[0161] Table 1 Parameters of the Example

[0162]

[0163] Please see Figure 7 , Figure 7 This is a Gazebo simulation scene diagram showing the initial stage where four vehicles maintain a straight line and the signpost is lowered in a controllable manner; please refer to [link / reference]. Figure 8, Figure 8 To trigger the transition of the autonomous vehicle swarm from a straight formation to a parallelogram formation in Phase 1 of the region, a marker is erected during the transition; please refer to the Gazebo simulation scenario. Figure 9 , Figure 9 This is a diagram showing the switching positions of the parallelogram; please refer to [link / reference]. Figure 10 , Figure 10 To trigger the transition of the autonomous vehicle cluster from a straight formation to a trapezoidal formation in Phase 2 of the region, a marker is erected during the transition; please refer to the Gazebo simulation scenario. Figure 11 , Figure 11 This is a diagram showing the trapezoidal switching position; please refer to [link / reference]. Figure 12 , Figure 12 To trigger the transition of the autonomous vehicle cluster from a straight-line formation to a linear formation in the Phase 3 area, a marker is erected during the transition; please refer to the Gazebo simulation scenario. Figure 13 , Figure 13 This is a diagram showing the location of the linear switching.

[0164] Understandably, the distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system with regional triggering provided in this application includes: determining a reference path and identifying a triggering region within the reference path; obtaining a matrix of the desired formation based on the desired pose of the follower relative to the virtual navigator; tracking the reference path through the virtual navigator and detecting the triggering region; in response to the virtual navigator being located in the triggering region, obtaining the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation; obtaining the MPC controllers of the virtual navigator and the follower; solving the MPC controllers using the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence; controlling the follower's actions using the desired pose and the optimal control sequence to reconfigure the multi-unmanned vehicle inspection system to the desired formation and maintain the desired formation during inspection. This application employs a multi-unmanned vehicle (UAV) platooning control framework based on "distributed model prediction + region triggering." It specifically addresses challenges such as difficulty in forming, maintaining, and reconfiguring multi-UAV platoons under different environments, weak collaborative collision avoidance among multiple vehicles, and single-point failure risks, insufficient constraints, and weak environmental adaptability in traditional lead-follow platooning. This framework improves platooning coordination accuracy and safety, operational reliability, and reduces the computational load of platooning control. Simultaneously, a virtual lead-follow platooning algorithm is used to construct the required formation, reducing formation transformation time. Distributed model predictive control is adopted, with each vehicle activating its own MPC control node to solve problems in parallel for each vehicle, shortening computation time. A region triggering mechanism is designed to make the increase in distance constraints controllable, reducing the computational load on individual robots. This significantly improves the efficiency of multi-UAV platooning formation, maintenance, and reconfiguration, as well as its safety and reliability during operation.

[0165] Accordingly, embodiments of this application also provide a distributed safety reconfiguration control system for a region-triggered multi-unmanned vehicle inspection system. Please refer to [link to relevant documentation]. Figure 4 , Figure 4 This is a module connection diagram of the distributed safety reconfiguration control system for a multi-unmanned vehicle inspection system triggered by a region, provided in this embodiment of the application. The distributed safety reconfiguration control system for a multi-unmanned vehicle inspection system triggered by a region, provided in this embodiment of the application, is applied to a multi-unmanned vehicle inspection system including a virtual navigator and multiple followers. The system includes:

[0166] The path acquisition module 10 is used to determine the reference path and the trigger area in the reference path;

[0167] The formation acquisition module 20 is used to acquire the matrix of the desired formation based on the expected pose of the follower relative to the virtual navigator.

[0168] Navigator motion module 30 is used to track the reference path through a virtual navigator and perform trigger area detection;

[0169] Follower pose module 40 is used to obtain the desired pose of the follower in response to the virtual navigator being located in the trigger area, based on the real-time pose of the virtual navigator and the matrix of the desired formation.

[0170] The controller acquisition module 50 is used to acquire the MPC controller of the virtual navigator and the MPC controller of the follower.

[0171] The control sequence acquisition module 60 is used to solve the MPC controller through the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence.

[0172] The formation control module 70 is used to control the follower's movements according to the follower's desired pose, reconstruct the multi-unmanned vehicle inspection system to the desired formation, and maintain the desired formation for inspection.

[0173] This application has provided a detailed description of a distributed safety reconfiguration control method and system for a multi-unmanned vehicle inspection system with regional triggering, as provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, characterized in that, The control method, applied to a multi-unmanned vehicle inspection system including a virtual navigator and multiple followers, includes: Determine a reference path, and then determine the trigger area within that reference path; The matrix for obtaining the desired formation is obtained based on the desired pose of the follower relative to the virtual navigator; The virtual navigator tracks the reference path and performs trigger area detection. In response to the virtual navigator being located in the triggering area, the desired pose of the follower is obtained based on the real-time pose of the virtual navigator and the matrix of the desired formation; Obtain the MPC model predictive control controller of the virtual navigator and the MPC controller of the follower: Obtain the state constraint function before the follower reaches the predicted time domain endpoint, and the terminal constraint function when the follower reaches the predicted time domain endpoint; Establish collision avoidance soft constraints based on the real-time distance and safe distance between the followers; In response to maintaining the desired formation, the MPC controller of the follower is obtained based on the state constraint function and the terminal constraint function; In response to reconstructing the desired formation, the follower's MPC controller is obtained based on the state constraint function, the terminal constraint function, and the collision avoidance soft constraint. The characterization formula of the follower's MPC controller includes: ； in, The total cost function of the follower's MPC controller; The state constraint function is defined for the follower before it reaches the predicted time domain endpoint. The terminal constraint function is the function used when the follower reaches the predicted time domain endpoint. The cost function for maintaining the desired formation for the followers. This refers to the collision avoidance soft constraint; The parameters are transformed by the cost function. This indicates that the desired formation should be maintained. This indicates the reconstruction of the desired formation; The MPC controller of the virtual navigator is obtained based on the state constraint function and the terminal constraint function, and its characterization formula includes: ； in, The total cost function of the MPC controller of the virtual navigator; The optimal control sequence is obtained by solving the MPC controller using the IPOPT nonlinear optimization problem solver. By controlling the follower's desired pose and the optimal control sequence, the follower's actions are reconstructed into the desired formation of the multi-unmanned vehicle inspection system, and the inspection is carried out while maintaining the desired formation.

2. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as described in claim 1, is characterized in that... The characterization formula of the state constraint function includes: ； in, The state constraint function; The state error at time step k; This is the weight matrix for the state error; This is the control input at time step k; The weight matrix for controlling the input; To predict the end point in the time domain; The characterization formula of the terminal constraint function includes: ； in, The terminal constraint function; To predict the endpoint in the time domain Terminal state error; This is the weight matrix for the terminal state error.

3. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as described in claim 1, is characterized in that... The characterization formula for the collision avoidance soft constraint includes: ； in, This refers to the collision avoidance soft constraint; The real-time distance between the followers; The safe distance between the followers; The weighting coefficient for obstacle avoidance penalty is such that the larger the value, the greater the constraint, and vice versa.

4. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region, as described in claim 1, is characterized in that... The steps for obtaining the optimal control sequence include: Obtain the input linear velocity boundary and input angular velocity boundary of the follower, and construct physical constraints: ； in, , The solver for the first The control inputs for the linear velocity and angular velocity of the first follower are used to characterize the first... The expected linear velocity and expected angular velocity of each follower; , For the first The minimum and maximum expected linear velocities of each follower; , For the first The minimum and maximum expected angular velocities of the autonomous vehicle; The optimal control sequence is obtained based on the physical constraints: ； in, To control the time domain; In order to be in Always The predicted result of the linear velocity output at time t; In order to be in Always The predicted result of the angular velocity output at time t.

5. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region as described in claim 1, characterized in that, The formula for representing the matrix of the desired formation includes: ； in, Let be the matrix of the desired formation, where the first row is the formation parameters of the virtual navigator, and the remaining rows are the formation parameters of the followers; i = 2…n, where i is the follower, and 2…n are the identification markers of the follower i; The expected distance between the follower i and the virtual navigator; Let i be the desired angle between the follower i and the virtual navigator.

6. The distributed safety reconfiguration control method for a multi-unmanned vehicle inspection system triggered by a region as described in claim 1, characterized in that, The kinematic model of the multi-unmanned vehicle inspection system is obtained by using the forward Euler method for discrete model description. The representation formula of the kinematic model includes: ； in, For the unmanned vehicle in the first Position coordinates at each moment; For the unmanned vehicle in the first Position coordinates at each moment; For driverless cars in the first The orientation angle at any given moment; Indicates that driverless cars are in the first The linear velocity at time t; For driverless cars in the first angular velocity at time t; This indicates the interval between sampling times.

7. A distributed safety reconfiguration control system for a multi-unmanned vehicle inspection system triggered by a region, characterized in that, An inspection system for multiple unmanned vehicles, including a virtual navigator and multiple followers, comprising: The path acquisition module (10) is used to determine a reference path and to determine a trigger area in the reference path; The formation acquisition module (20) is used to acquire a matrix of desired formations based on the desired pose of the follower relative to the virtual navigator. Navigator motion module (30), the navigator motion module (30) is used to track the reference path through the virtual navigator and perform trigger area detection; Follower pose module (40), the follower pose module (40) is used to obtain the desired pose of the follower based on the real-time pose of the virtual navigator and the matrix of the desired formation in response to the virtual navigator being located in the trigger area; The controller acquisition module (50) is used to acquire the MPC model predictive control controller of the virtual navigator and the MPC controller of the follower. Obtain the state constraint function before the follower reaches the predicted time domain endpoint, and the terminal constraint function when the follower reaches the predicted time domain endpoint; Establish collision avoidance soft constraints based on the real-time distance and safe distance between the followers; In response to maintaining the desired formation, the MPC controller of the follower is obtained based on the state constraint function and the terminal constraint function; In response to reconstructing the desired formation, the follower's MPC controller is obtained based on the state constraint function, the terminal constraint function, and the collision avoidance soft constraint. The characterization formula of the follower's MPC controller includes: ； in, The total cost function of the follower's MPC controller; The state constraint function is defined for the follower before it reaches the predicted time domain endpoint. The terminal constraint function is the function used when the follower reaches the predicted time domain endpoint. The cost function for maintaining the desired formation for the followers. This refers to the collision avoidance soft constraint; The parameters are transformed by the cost function. This indicates that the desired formation should be maintained. This indicates the reconstruction of the desired formation; The MPC controller of the virtual navigator is obtained based on the state constraint function and the terminal constraint function, and its characterization formula includes: ； in, The total cost function of the MPC controller of the virtual navigator; The control sequence acquisition module (60) is used to solve the MPC controller through the IPOPT nonlinear optimization problem solver to obtain the optimal control sequence; Formation control module (70) is used to control the movement of the follower by the desired pose of the follower, reconstruct the multi-unmanned vehicle inspection system to the desired formation, and maintain the desired formation for inspection.

Images