Self-triggering unmanned vehicle distributed formation control method based on double triggering mechanism

Abstract

A self-triggering unmanned vehicle distributed formation control method based on a double-triggering mechanism belongs to the field of unmanned vehicle distributed cooperative formation control, and comprises the following steps: S1, constructing a multi-agent consistency cooperative control framework; s2, constructing an intelligent vehicle cooperative control system driven by a dynamic situation field; s3, carrying out communication topology modeling between the vehicles; s4, a self-triggering mechanism is introduced, and correct triggering of communication is ensured by designing a double-triggering threshold function; s5, establishing a prediction model of the vehicle formation; and S6, the distributed model prediction controller controls accurate implementation of a formation transformation strategy. According to the invention, through combination of a self-triggering mechanism and distributed model prediction control, the control efficiency and precision of the unmanned vehicle formation are significantly improved; compared with the traditional method, the communication frequency is reduced, and the communication burden is reduced by 5%; when the vehicle runs on a curve, the vehicle track error is reduced and is 4% more accurate than that of a traditional method.

Description

Technical Field

[0001] This invention belongs to the field of distributed cooperative formation control of unmanned vehicles, specifically a self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism. Background Technology

[0002] The self-triggering mechanism is a strategy based on the real-time state of the system to dynamically decide on communication and control updates. Compared with the traditional fixed-period triggering mechanism, its core is to trigger communication or control actions only when necessary, thereby reducing resource consumption and improving system efficiency.

[0003] Despite significant progress in resource optimization, self-triggered mechanisms still have the following limitations in coupling dynamic potential fields with multi-objective optimization: Dynamic potential field adaptability: Traditionally, it is assumed that the potential field is static and cannot adapt to the dynamic changes of obstacles in real time, which may lead to obstacle avoidance failure.

[0004] Multi-objective optimization conflict: There may be conflicts between multiple objectives (such as communication frequency and security), making it difficult to achieve the optimal result simultaneously.

[0005] Distributed Model Predictive Control (DMPC) is a distributed optimization control method for multi-agent systems, proposed by researchers such as Mohammadi. It aims to achieve the global control objective of complex systems through local information interaction and collaborative optimization.

[0006] Traditional DMPC uses a fixed communication cycle, requiring vehicles to exchange information periodically regardless of significant changes in system status. This results in a large amount of redundant communication (the necessity of information exchange is low when the platoon is running stably). Furthermore, the frequent information exchange between vehicles places high demands on communication bandwidth, posing numerous challenges when handling large-scale platoons, including but not limited to difficulties in dynamic path planning and obstacle avoidance, high communication bandwidth requirements, large consumption of computing resources, and stringent real-time requirements.

[0007] The combination of the self-triggering mechanism and DMPC provides a systematic solution to the aforementioned problems: the self-triggering mechanism dynamically adjusts the communication frequency, significantly reducing the communication burden; simultaneously, DMPC performs distributed rolling optimization based on the latest information at the trigger moment, ensuring control accuracy and formation stability. This collaborative architecture not only optimizes communication resources and computational efficiency but also addresses the limitations of traditional methods in adaptability to dynamic environments and multi-objective conflicts through dynamic potential field driving and multi-objective game mechanisms, providing a flexible and efficient solution for the reliable application of unmanned vehicle formations in complex scenarios. Summary of the Invention

[0008] This invention provides a self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism to address the shortcomings of existing technologies.

[0009] This invention is achieved through the following technical solution: The self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism includes the following steps: S1: Construct a multi-agent consensus and collaborative control framework. This can be achieved by equipping the platooning vehicles with sensory and communication modules and deploying a distributed model prediction controller to recognize obstacles, calculate platooning errors, and facilitate information exchange between the vehicles. S2: Construct an intelligent vehicle collaborative control system driven by a dynamic potential field; S3: Inter-vehicle communication topology modeling; S4: Introduce a self-triggering mechanism and ensure correct communication triggering by designing a dual triggering threshold function; S5: Establish a prediction model for vehicle formations. By establishing a relative model of vehicle formation and a nonlinear kinematic model, a prediction model for vehicle formations can be established. S6: The distributed model predictive controller accurately implements the formation transformation strategy.

[0010] As described above, the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism includes a multi-agent consensus collaborative control framework in step S1, comprising a vehicle interaction layer and a vehicle node layer. The vehicle interaction layer uses the lead vehicle as the global coordination node, dynamically selecting the optimal link through the V2X / RS485 dual-mode communication protocol, and broadcasting global path planning, target formation configuration, and LiDAR mapping data to following vehicles. The vehicle node layer includes a distributed model prediction controller that operates independently for each vehicle, fusing YOLOv8 visual perception data and LiDAR SLAM mapping information, and combining its own nonlinear kinematic model, the predicted state of adjacent vehicles, and formation errors to generate local trajectory control inputs. A self-triggered mechanism is introduced, triggering data interaction only when the formation error exceeds a threshold or the predicted state deviates significantly, thereby reducing communication frequency.

[0011] As described above, in the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism, the operation of constructing an intelligent vehicle cooperative control system driven by a dynamic potential field in step S2 is as follows: Based on lane lines, obstacle categories, and location data detected in real-time by YOLOv8, a dynamic repulsive field is constructed; simultaneously, a global gravitational field is generated through LiDAR SLAM mapping; combined with a vehicle motion prediction model, the vehicle pose is corrected in real-time, and a safe distance threshold is calculated, dynamically adjusting the range of the potential field to ensure a balance between attraction and repulsion. Establish an objective function for the obstacle distance penalty, with the objective function expression as follows: Among them, λ is dynamically adjusted based on the real-time safe distance of the LiDAR.

[0012] In the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism as described above, the specific operation of vehicle communication topology modeling in step S3 is as follows: S3-1: Communication topology modeling uses directed graphs Describes the information flow between vehicles, where the vertex set Vehicle nodes are represented as follows: 0 represents the lead vehicle, responsible for guiding the entire convoy; 1, 2, ..., N represent following vehicles, which follow the lead vehicle's movement; edge sets. The information exchange relationship between vehicles is described as follows: ; S3-2: Adjacency Matrix The communication permissions are described. The traction matrix P identifies the direct control relationship between the following vehicle and the lead vehicle. The expression for the communication permissions between the following vehicles is: ,in Adjacency matrix , =1 indicates that vehicle j can receive information from vehicle i; =0 indicates that vehicle j cannot receive information from vehicle i; the expression for the traction matrix P, which identifies the direct control relationship between the following vehicle and the lead vehicle, is: ,in, =1 indicates that vehicle i directly receives information from the lead vehicle; =0 indicates that vehicle i does not directly receive information from the navigator vehicle.

[0013] As described above, in the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism, the dual-trigger threshold function in step S4 includes a local state error trigger threshold function and a formation deformation trigger threshold function. The expression for the difference trigger threshold function is as follows: ,in, Let j be the tracking error of vehicle j. As an adjustable parameter, the expression for the formation deformation triggering threshold function is: ,in, For vehicle spacing deviation, This is an adjustable parameter.

[0014] As described above, in the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism, the desired relative positions of adjacent vehicles i−1 and i in the vehicle formation relative model in step S5 are determined by the target spacing. and opposite angles The actual relative position error is defined as: in: ; The nonlinear kinematic model of the vehicle in step S5 is established based on the ISO8855 vehicle body coordinate system, and the vehicle kinematic differential equations are: Where L is the wheelbase, v is the longitudinal speed, and δ is the steering angle; The vehicle platooning prediction model in step S5 is based on a nonlinear kinematic model of vehicles. The state of each vehicle is discretized, and the state update equation is: in, for ; The relative positional error between vehicle i and its neighboring vehicle j: in, For the target formation spacing, The relative orientation angle of the target. The nonlinear discrete kinematic prediction model for the entire formation can be expressed as: The vehicle platooning prediction model is implemented based on a distributed predictive controller, with each vehicle independently solved and optimized. Its expression is: The constraints of the vehicle platooning prediction model include: vehicle kinematics model, control input limit, and safety distance. The expression for the vehicle kinematics model is as follows: The expression for the control input limit is: The expression for the safety distance is: (障碍物密度) .

[0015] As described above, the distributed platooning control method for self-triggered unmanned vehicles based on a dual-trigger mechanism includes a distributed model predictive controller controlling the platooning transformation strategy in step S6, comprising a dynamic safety distance adjustment stage, a multi-vehicle cooperative lateral migration stage, and a platooning stability enhancement stage. The dynamic safety distance adjustment stage aims to dynamically adjust the longitudinal distance between vehicles to build a safe foundation for subsequent cooperative lane changes and adapt to the needs of complex scenarios such as high speeds. The multi-vehicle cooperative lateral migration stage enables multi-vehicle lateral cooperative lane changes to avoid path conflicts. The platooning stability enhancement stage aims to maintain platooning stability and lane centering.

[0016] As described above, in the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism, the dynamic adjustment phase of the safety distance uses a priority set generation mechanism to define a vehicle lane-change priority set based on the longitudinal positional relationship between the initial formation configuration and the target configuration. The expression for this set is: Among them, priority , Let i be the initial longitudinal position of vehicle i, with the lead vehicle having the highest priority. Real-time calculation of safe vehicle distance in a scene based on LiDAR SLAM data: ,in, For obstacle density, Based on the safe distance, Here, ρ is the obstacle density unit, representing the increment of safe distance per unit obstacle density. ρ is an adjustment coefficient. Obstacle density ρ is a dimensionless normalized value, ranging from [0,1], representing the density of obstacles within a unit area. It is typically calculated using LiDAR SLAM data: ρ = ; Then, the vehicle speed is optimized using a distributed model predictive controller, with the objective function expression as follows: This ensures the actual spacing. Let j be the actual distance between vehicles. For buffer distance, Let j be the acceleration of vehicle j, when all vehicles satisfy Then we will move on to the next stage.

[0017] 9. The self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism according to claim 7, characterized in that: in the multi-vehicle cooperative lateral migration stage, a fifth-order polynomial lateral trajectory is generated for the vehicle requiring lane change, the expression of which is: Among them, coefficient ,..., The boundary conditions are determined by the starting and ending positions, velocity, and acceleration. The constraints include that the velocity and acceleration are continuous at the starting and ending points. Subsequently, a polyhedral safety region is used to define the collision avoidance boundary, the range of which is expressed as follows: in, For vehicle length, For vehicle width, This is for buffer time; Then, combine it with V2X broadcast synchronized lane change commands; With a distributed model predictive controller and lateral tracking control, the objective function expression is: ,in, For the center line of the target lane, This is the steering angle; if YOLOv8 detects a sudden obstacle, it immediately freezes the lane change command and returns to the original lane.

[0018] The self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism, as described above, includes the following steps in its formation stability enhancement stage: YOLOv8 uses its target detection capabilities to identify lane lines and outputs the pixel positions of the lane lines. Step 1: Input image preprocessing, adjusting the image captured by the camera to the input resolution of YOLOv8; Step 2: Lane line detection, YOLOv8 outputs the bounding box or key points of the lane lines; Step 3: Post-processing: Extract the pixel coordinates (u,v) of the lane lines, where u is the horizontal pixel coordinate and v is the vertical pixel coordinate; Then, a pixel coordinate to world coordinate transformation is performed, converting the lane line's pixel coordinates (u,v) to world coordinates (x,y) in the vehicle coordinate system. The transformation formula is as follows: in, Here are the pixel coordinates of the image center point, and h is the camera mounting height. f is the camera focal length; Then, the lane lines are fitted using the least squares method to fit the world coordinates (x, y), resulting in the lane line fitting centerline equation based on YOLOv8 recognition. The formula for lateral deviation correction is: ; The expression for verifying formation configuration differences is: like This will trigger a refactoring; Longitudinal displacement compensation employs smooth trajectory planning, the expression of which is: ,in, To compensate for distance, It is a time constant; Lateral offset adjustment for curves, the offset formula is: ,in, R is the adjustment factor, and R is the curve radius.

[0019] The advantages of this invention are: by combining a self-triggering mechanism and distributed model predictive control, this invention significantly improves the control efficiency and accuracy of unmanned vehicle formations; compared with traditional methods, the communication frequency is reduced, reducing the communication burden by 5%; when driving on curves, the vehicle trajectory error is reduced, and it is 4% more accurate than traditional methods. Attached Figure Description

[0020] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram of a topology communication network provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the relative model of the vehicle platoon formation of the present invention; Figure 4 This is a flowchart of the formation transformation strategy controlled by the distributed model predictive controller of the present invention; Figure 5 This is a schematic diagram illustrating the platooning effect of the vehicles in a laboratory simulation verification of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, the self-triggered distributed formation control method for unmanned vehicles based on a dual-trigger mechanism includes the following steps: S1: Construct a multi-agent consensus and collaborative control framework. This can be achieved by equipping the platooning vehicles with sensory and communication modules and deploying a distributed model predictive controller to recognize obstacles, calculate platooning errors, and facilitate information exchange between the vehicles. S2: Construct an intelligent vehicle collaborative control system driven by a dynamic potential field; S3: Inter-vehicle communication topology modeling; S4: Introduce a self-triggering mechanism and ensure correct communication triggering by designing a dual triggering threshold function; S5: Establish a prediction model for vehicle formations. By establishing a relative model of vehicle formation and a nonlinear kinematic model, a prediction model for vehicle formations can be established. S6: The distributed model predictive controller accurately implements the formation transformation strategy.

[0024] Preferably, the specific operation of step S1 in this embodiment includes the following steps: S1-1: The simulated vehicle needs to be equipped with sensors such as GPS, IMU (Inertial Measurement Unit) and LiDAR, and a distributed model predictive controller (DMPC) needs to be deployed to realize obstacle recognition, formation error calculation and information exchange between small vehicles; S1-2: The multi-agent consensus cooperative control framework is divided into a two-layer structure of vehicle interaction layer and vehicle node layer to achieve the unification of global formation stability and local dynamic obstacle avoidance capability.

[0025] Preferably, the multi-agent consensus collaborative control framework in step S1 of this embodiment includes a vehicle interaction layer and a vehicle node layer. The vehicle interaction layer uses the lead vehicle as the global coordination node, dynamically selects the optimal link through the V2X / RS485 dual-mode communication protocol, and broadcasts global path planning, target formation configuration (such as queue, diamond), and LiDAR mapping data to the following vehicles. The vehicle node layer includes a distributed model predictive controller (DMPC) that operates independently for each vehicle. It integrates YOLOv8 visual perception data and LiDAR SLAM mapping information, combines its own nonlinear kinematic model, the predicted state of adjacent vehicles, and formation error to generate local trajectory control input. A self-triggering mechanism is introduced, which triggers data interaction only when the formation error exceeds a threshold or the predicted state deviates significantly, thereby reducing the communication frequency.

[0026] Preferably, the operation of constructing the intelligent vehicle cooperative control system driven by the dynamic potential field in step S2 of this embodiment is as follows: Based on the lane lines, obstacle categories and location data detected in real time by YOLOv8, a dynamic repulsive field is constructed (the closer the obstacle, the greater the repulsive force). At the same time, a global gravitational field (target path guidance) is generated through LiDAR SLAM mapping. Combined with the vehicle motion prediction model, the vehicle pose is corrected in real time and the safe distance threshold is calculated. The range of action of the potential field is dynamically adjusted to ensure the balance between attraction and repulsion. Establish an objective function for the obstacle distance penalty, with the objective function expression as follows: Among them, λ is dynamically adjusted based on the real-time safe distance of the LiDAR.

[0027] Preferably, the specific operation of vehicle-to-vehicle communication topology modeling in step S3 of this embodiment is as follows: S3-1: Communication topology modeling uses directed graphs Describes the information flow between vehicles, where the vertex set Vehicle nodes are represented as follows: 0 represents the lead vehicle, responsible for guiding the entire convoy; 1, 2, ..., N represent following vehicles, which follow the lead vehicle's movement; edge sets. The information exchange relationship between vehicles is described as follows: For example, edge (i,j) This indicates that vehicle j can receive information from vehicle i (e.g., Figure 2 (as shown) S3-2: Adjacency Matrix The communication permissions are described. The traction matrix P identifies the direct control relationship between the following vehicle and the lead vehicle. The expression for the communication permissions between the following vehicles is: ,in Adjacency matrix , =1 indicates that vehicle j can receive information from vehicle i; =0 indicates that vehicle j cannot receive information from vehicle i; The calculations are shown in Table 1; The expression for the direct control relationship between the following vehicle and the lead vehicle, represented by the traction matrix P, is as follows: ,in, =1 indicates that vehicle i directly receives information from the lead vehicle; =0 indicates that vehicle i does not directly receive information from the navigator vehicle. The calculations are shown in Table 2.

[0028] Preferably, in step S4 of this embodiment, the dual triggering threshold function includes a local state error triggering threshold function and a formation deformation triggering threshold function, and the expression of the difference triggering threshold function is: ,in, Let j be the tracking error of vehicle j. As an adjustable parameter, the expression for the formation deformation triggering threshold function is: ,in, For vehicle spacing deviation, This is an adjustable parameter.

[0029] like Figure 3 As shown, preferably, in step S5 of this embodiment, the desired relative positions of adjacent vehicles i−1 and i in the vehicle formation relative model are determined by the target spacing. and opposite angles The actual relative position error is defined as: in: .

[0030] Preferably, in this embodiment, the vehicle nonlinear kinematic model established in step S5 is based on the ISO8855 vehicle body coordinate system, and the vehicle kinematic differential equation is: Where L is the wheelbase, v is the longitudinal speed, and δ is the steering angle.

[0031] Preferably, in step S5 of this embodiment, the vehicle platooning prediction model is established based on a nonlinear kinematic model of the vehicles, and the state of each vehicle is discretized. The state update equation is as follows: in, for ; The relative positional error between vehicle i and its neighboring vehicle j: in, For the target formation spacing, The relative orientation angle of the target. The nonlinear discrete kinematic prediction model for the entire formation can be expressed as: The vehicle platooning prediction model is implemented based on a distributed predictive controller, with each vehicle independently solved and optimized. Its expression is: The constraints of the vehicle platooning prediction model include: vehicle kinematics model, control input limit, and safety distance. The expression for the vehicle kinematics model is as follows: The expression for the control input limit is: The expression for the safety distance is: (障碍物密度) .

[0032] like Figure 4 As shown, preferably, the distributed model predictive controller's platooning change strategy in step S6 of this embodiment includes a dynamic adjustment stage for safe distance, a multi-vehicle cooperative lateral migration stage, and a platooning stability enhancement stage. The dynamic adjustment stage for safe distance takes dynamically adjusting the longitudinal distance between vehicles as its core objective, building a safe foundation for subsequent cooperative lane changes and adapting to the needs of complex scenarios such as high speeds. The multi-vehicle cooperative lateral migration stage enables multi-vehicle lateral cooperative lane changes to avoid path conflicts. The platooning stability enhancement stage aims to maintain platooning stability and lane centering.

[0033] Preferably, in this embodiment, the dynamic adjustment phase of the safety distance uses a priority set generation mechanism to define a vehicle lane-changing priority set based on the longitudinal positional relationship between the initial formation configuration and the target configuration. The expression for this set is: Among them, priority , Let i be the initial longitudinal position of vehicle i, with the lead vehicle having the highest priority. Real-time calculation of safe vehicle distance in a scene based on LiDAR SLAM data: ,in, For obstacle density, Based on the safe distance, Here, ρ is the obstacle density unit, representing the increment of safe distance per unit obstacle density. ρ is an adjustment coefficient. Obstacle density ρ is a dimensionless normalized value, ranging from [0,1], representing the density of obstacles within a unit area. It is typically calculated using LiDAR SLAM data: ρ = ; Then, the vehicle speed is optimized using a distributed model predictive controller, with the objective function expression as follows: This ensures the actual spacing. Let j be the actual distance between vehicles. For buffer distance, Let j be the acceleration of vehicle j, when all vehicles satisfy Then we will move on to the next stage.

[0034] Preferably, in this embodiment, the multi-vehicle cooperative lateral migration stage generates a fifth-order polynomial lateral trajectory for the vehicle requiring lane change, the expression of which is: Among them, coefficient ,..., The boundary conditions are determined by the starting and ending positions, velocity, and acceleration. The constraints include that the velocity and acceleration are continuous at the starting and ending points. Subsequently, a polyhedral safety region is used to define the collision avoidance boundary, the range of which is expressed as follows: in, For vehicle length, For vehicle width, This is for buffer time; Then, combine it with V2X broadcast synchronized lane change commands; With a distributed model predictive controller and lateral tracking control, the objective function expression is: ,in, For the center line of the target lane, This is the steering angle; if YOLOv8 detects a sudden obstacle, it immediately freezes the lane change command and returns to the original lane.

[0035] Preferably, the formation stability enhancement stage described in this embodiment is based on YOLOv8, which uses its target detection capabilities to identify lane lines and outputs the pixel positions of the lane lines. The specific operation includes the following steps: Step 1: Input image preprocessing, adjust the image captured by the camera to the input resolution of YOLOv8 (e.g., 640×640). Step 2: Lane line detection, YOLOv8 outputs the bounding box or key points of the lane lines; Step 3: Post-processing: Extract the pixel coordinates (u,v) of the lane lines, where u is the horizontal pixel coordinate and v is the vertical pixel coordinate; Then, a pixel coordinate to world coordinate transformation is performed, converting the lane line's pixel coordinates (u,v) to world coordinates (x,y) in the vehicle coordinate system. The transformation formula is as follows: in, Here, h represents the pixel coordinates of the image center point (principal point), and h represents the camera mounting height. f is the camera focal length (in pixels); Then, the lane lines are fitted using the least squares method to fit the world coordinates (x, y), resulting in the lane line fitting centerline equation based on YOLOv8 recognition. The formula for lateral deviation correction is: ; The expression for verifying formation configuration differences is: like This will trigger a refactoring; Longitudinal displacement compensation employs smooth trajectory planning, the expression of which is: ,in, To compensate for distance, It is a time constant; Lateral offset adjustment for curves, the offset formula is: ,in, R is the adjustment factor, and R is the curve radius.

[0036] Preferably, in the implementation of the formation change strategy described in this embodiment, the process of solving the fifth-order polynomial lateral trajectory during the multi-vehicle cooperative lateral migration stage is as follows: The equation of the fifth-order polynomial locus is: + + + + Its derivative is: + + + + + + + Boundary conditions include: Starting point (t=0): Position y(0) = ,speed = acceleration (0)= ; End point (t=T): Position y(T)= ,speed (T)= acceleration (T)= .

[0037] Substituting the boundary conditions into the trajectory equation and its derivative, we obtain the following system of equations: Starting point: y(0)= = Starting speed: = = Initial acceleration: (0)=2 = ⇒ = Finish line location: + + + + Final speed: + + + + Final acceleration: + + + Known coefficients , , After substitution, the remaining unknown coefficients , , Solve using the following matrix equation: = = .

[0038] Verification test The key performance indicators of this invention were verified through laboratory simulations under different weather conditions (sunny, light rain) and application scenarios (urban roads). Figure 5 The results of the verification of actual roads (the dataset used is derived from multimodal perception data collected in actual scenarios by our company) are shown in Table 3.

[0039] As shown in Table 3, the experimental data demonstrates that under clear weather conditions, the present invention can operate with a lower communication frequency and higher control accuracy on both highways and urban roads. For example, in highway scenarios, the average communication frequency is 3.0 times / second, and the lateral deviation is controlled within ±0.15 meters. As weather conditions deteriorate, the communication frequency, safety distance, and response time of the present invention increase accordingly to ensure driving safety, while the control accuracy decreases. These results demonstrate that the present invention can dynamically adjust parameters according to different environmental challenges to achieve efficient and safe unmanned vehicle platooning control.

[0040] The results of comparing the data of the present invention with those of the traditional method (fixed periodic triggering) are shown in Table 4.

[0041] As shown in Table 4, the self-triggered distributed platooning control algorithm for unmanned vehicles proposed in this invention, based on a dual-trigger mechanism, exhibits significant advantages in control accuracy, communication efficiency, and system stability. Compared to the traditional fixed-cycle triggering method, this invention reduces the control error from ±6cm to ±5.7cm, an improvement of 4%. The data interaction frequency is reduced from 50 times per second to an average of 48 times per second, reducing communication bandwidth requirements by 4%. Simultaneously, the response time is shortened from 100ms to 94ms, an improvement of 6%. By introducing DMPC and dynamic potential field collaborative optimization, the system stability and robustness are significantly improved. Its universality has been verified in complex scenarios such as multi-lane cooperative lane changing and dynamic obstacle avoidance, fully demonstrating the superiority of this algorithm in reducing communication costs, enhancing real-time control accuracy, and adapting to complex environments. It provides an efficient and reliable solution for the large-scale application of unmanned vehicle platooning.

[0042] In summary, this invention significantly improves the control accuracy of unmanned vehicle platooning by constructing a multi-agent consensus cooperative control framework combined with a multi-objective optimized distributed model predictive controller. This invention not only handles complex nonlinear kinematic models but also effectively addresses lateral and longitudinal coupling problems, ensuring that each vehicle in the platoon accurately tracks the expected path. By introducing a self-triggering mechanism, this invention reduces unnecessary data interactions, thereby lowering the demand for communication bandwidth. Each vehicle node in this invention exchanges data only under specific conditions, rather than exchanging information in every control cycle. This invention not only improves the system's efficiency and flexibility but also adapts to application requirements under different network conditions.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed formation control method for self-triggered unmanned vehicles based on a dual-trigger mechanism, characterized in that: It comprises the following steps: S1: Constructing a multi-agent consistency collaborative control framework, which can realize the identification of obstacles, the calculation of formation errors and the information interaction between vehicles by matching the formation car with sensory modules and communication modules and deploying a distributed model predictive controller; S2: Constructing an intelligent vehicle collaborative control system driven by a dynamic potential field; S3: Modeling the inter-vehicle communication topology; S4: Introducing a self-triggering mechanism to ensure correct triggering of communication by designing a double-trigger threshold function; S5: Establishing a prediction model for vehicle formation by establishing a vehicle formation shape relative model and a nonlinear kinematics model to establish a prediction model for vehicle formation; S6: Distributed model predictive controller controls the accurate implementation of formation transformation strategy. 2.The distributed formation control method based on the double trigger mechanism self-triggered unmanned vehicle according to claim 1, wherein: The multi-agent consistency collaborative control framework in step S1 comprises a vehicle interaction layer and a vehicle node layer, the vehicle interaction layer takes the lead vehicle as the global coordination node, dynamically selects the optimal link through the V2X / RS485 dual-mode communication protocol, and broadcasts the global path planning, target formation configuration and laser radar mapping data to the follower vehicle; the vehicle node layer comprises a distributed model predictive controller independently running on each vehicle, which fuses YOLOv8 visual perception data and laser radar SLAM mapping information, combines the nonlinear kinematics model of the vehicle itself, the predicted state of the adjacent vehicle and the formation error, generates a local trajectory control input, introduces a self-triggering mechanism, and only triggers data interaction when the formation error exceeds the threshold or the predicted state deviates significantly, thereby reducing the communication frequency. 3.The distributed formation control method based on the double trigger mechanism self-triggered unmanned vehicle according to claim 1, characterized in that: In step S2, the operation of constructing an intelligent vehicle collaborative control system driven by a dynamic potential field is as follows: based on the lane line, obstacle category and position data detected by YOLOv8 in real time, a dynamic repulsive force field is constructed, and a global attractive force field is generated by laser radar SLAM mapping; combined with the vehicle motion prediction model, the vehicle pose is corrected in real time and the safe vehicle distance threshold is calculated, the potential field action range is dynamically adjusted to ensure the balance of attractive force and repulsive force, An obstacle distance penalty objective function is established, and the target function expression is: Wherein, lambda is dynamically adjusted according to the real-time safe vehicle distance of laser radar.

4. The distributed platoon control method based on the double trigger mechanism self-triggering unmanned vehicle according to claim 1, characterized in that: In step S3, the specific operation of modeling the inter-vehicle communication topology is as follows: S3-1: Communication topology modeling uses a directed graph The inter-vehicle information flow is described, wherein the vertex set represents a vehicle node, 0 represents a leading vehicle responsible for guiding the entire vehicle platoon, and 1, 2, …, N represent following vehicles following the movement of the leading vehicle; the edge set The information interaction relationship between vehicles is described as: ; S3-2: Adjacency matrix The communication authority is described, and the traction matrix P identifies the direct control relationship between the following vehicles and the leading vehicle. The expression of the communication authority between the following vehicles is: wherein is the adjacency matrix , =1 indicates that vehicle j can receive information of vehicle i; =0 indicates that vehicle j cannot receive information of vehicle i; The expression of the traction matrix P identifying the direct control relationship between the following vehicles and the leading vehicle is: wherein, =1 indicates that vehicle i directly receives information of the leading vehicle; =0 indicates that vehicle i does not directly receive information of the leading vehicle.

5. The distributed platoon control method based on the double trigger mechanism self-triggering unmanned vehicle according to claim 1, characterized in that: In step S4, the double-trigger threshold function comprises a local state error trigger threshold function and a formation shape change trigger threshold function, and the difference trigger threshold function expression is: wherein, is the tracking error for vehicle j, is an adjustable parameter, and the platoon shape variation trigger threshold function expression is: wherein, is the inter-vehicle distance deviation, is an adjustable parameter.

6. The distributed platoon control method for self-triggered unmanned vehicles based on a dual-trigger mechanism according to claim 1, characterized in that: The expected relative position of the adjacent vehicles i−1 and i in the vehicle platoon formation relative model in step S5 is defined by a target spacing and a relative angle The actual relative position error is defined as: Wherein: ； The vehicle nonlinear kinematics model in step S5 is established based on the ISO8855 vehicle body coordinate system, and the vehicle kinematics differential equation is: Wherein L is the wheelbase, v is the longitudinal velocity, and δ is the steering angle; The vehicle formation prediction model in step S5 is established based on the vehicle nonlinear kinematics model, and the state update equation is: wherein is ; The relative position error of vehicle i and adjacent vehicle j is: wherein, is a target platoon spacing, is a target relative orientation angle, The nonlinear discrete kinematics prediction model of the entire formation can be expressed as: The vehicle formation prediction model is realized based on the distributed model predictive controller, and each vehicle independently solves the optimization, and the expression is: The constraint conditions of the vehicle platoon prediction model include: a vehicle kinematics model, a control input limit, and a safety gap, an expression of the vehicle kinematics model is: , an expression of the control input limit is: , and an expression of the safety gap is: (障碍物密度) . 7.The distributed formation control method based on the double trigger mechanism self-triggered unmanned vehicle according to claim 1, wherein: The distributed model predictive controller in the step S6 controls the platoon transformation strategy, including a safety distance dynamic adjustment phase, a multi-vehicle cooperative lateral migration phase, and a platoon stability reinforcement phase; the safety distance dynamic adjustment phase takes dynamic adjustment of vehicle longitudinal distance as a core target, builds a safety foundation for subsequent cooperative lane changing, and adapts to complex scene requirements such as high speed; The multi-vehicle cooperative lateral migration phase realizes multi-vehicle cooperative lateral lane changing and avoids path conflicts; and the platoon stability reinforcement phase aims to maintain platoon stability and lane-centered driving. 8.The distributed formation control method based on the double trigger mechanism self-triggered unmanned vehicle according to claim 7, wherein: The safety distance dynamic adjustment phase generates a priority set according to the longitudinal position relationship between the initial configuration and the target configuration of the platoon, and defines a vehicle lane-changing priority set, which is expressed as: wherein the priority , is the initial longitudinal position of the vehicle i, and the priority of the leading vehicle is the highest. Real-time calculation of scene safety distance based on lidar SLAM data: wherein, is the obstacle density, is the base safety distance, is the obstacle density unit, representing the increment of safety distance per unit obstacle density, is the adjustment coefficient, the obstacle density ρ is a dimensionless normalized value, ranging between [0, 1], representing the density of obstacles within a unit area, which is usually calculated through lidar SLAM data: ρ= ; Then, the vehicle speed is optimized by a distributed model predictive controller, with the objective function expressed as: , thus ensuring the actual distance between vehicles j and j + 1, is the buffer distance, is the acceleration of vehicle j, and when all vehicles satisfy , the next stage is entered. 9.The distributed formation control method based on the double-trigger mechanism self-triggered unmanned vehicle according to claim 7, wherein: The multi-vehicle cooperative lateral migration phase generates a five-degree polynomial lateral trajectory for the vehicle that needs to change lanes, and the expression is: wherein the coefficients ,..., are determined by the start and end position, velocity, acceleration boundary conditions, and the constraints include continuity of velocity and acceleration at the start and end points. Then, a polyhedral safety region is used to define the anti-collision boundary, and the range expression is: wherein, L is the vehicle length, W is the vehicle width, T is the buffer time; Then, a V2X broadcast synchronization lane changing instruction is combined; By the distributed model predictive controller and lateral tracking control, the objective function expression is: wherein, is the target lane centerline, is the steering angle; if a sudden obstacle is detected by YOLOv8, the lane changing instruction is immediately frozen and the original lane is returned.

10. The distributed platoon control method for self-triggered unmanned vehicles based on a dual-trigger mechanism according to claim 7, characterized in that: The platoon stability reinforcement phase identifies lane lines based on YOLOv8 through its target detection capability, and outputs the pixel position of the lane line, and the specific operation includes the following steps: Step 1: image preprocessing, adjusting the image captured by the camera to the input resolution of YOLOv8; Step 2: lane line detection, YOLOv8 outputs the boundary box or key points of the lane line; Step 3: post-processing: extract the pixel coordinates (u, v) of the lane line, where u is the lateral pixel coordinate and v is the longitudinal pixel coordinate; Then, the pixel coordinate to world coordinate conversion is performed to convert the pixel coordinates (u, v) of the lane line into world coordinates (x, y) in the vehicle coordinate system, and the conversion formula is: wherein, is the pixel coordinate of the image center point, h is the camera installation height, f is the focal length of the camera; Then, the lane line fitting is performed, and a least square method is used to fit the world coordinates (x, y) to obtain a lane line fitting center line equation based on the YOLOv8 recognition ; and a lateral deviation correction formula is ; The platoon configuration difference verification expression is: If , then trigger reconstruction; The longitudinal displacement compensation adopts a smooth trajectory planning, and its expression is: wherein, is a compensation distance, is a time constant; The lateral offset adjustment of the curve is given by the formula: wherein, is the adjustment coefficient and R is the radius of the curve.

Images