Internet of vehicles ramp convergence distributed cooperative control method based on communication perception

Abstract

The invention discloses an Internet of Vehicles ramp convergence distributed cooperative control method based on communication perception. The method comprises the following steps: establishing a bicycle kinematics model based on a bicycle model and an NR-V2X mode 2 resource reservation communication model; predicting a communication conflict probability sequence in a future time domain through a Markov chain model; constructing a distributed model predictive control optimization problem introducing an adaptive communication security margin, and coupling a conflict probability as a penalty term to a target function; decomposing an original problem into a vehicle control quadratic programming sub-problem and a communication safety margin convex optimization sub-problem by adopting a block coordinate descent framework, and alternately solving to obtain optimal control input and safety margin; the vehicle executes acceleration or steering operation according to the optimal control input, and safe and smooth afflux is achieved; according to the method, the communication reliability is explicitly incorporated into a control closed loop, the communication quality change is dynamically adapted, the safety, control robustness and smoothness of ramp convergence are remarkably improved, and meanwhile, the calculation efficiency is optimized to meet the vehicle-mounted real-time control requirement.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and autonomous driving control technology, and in particular to a distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication perception. Background Technology

[0002] With the continuous growth of urban vehicle numbers, traffic congestion is worsening, and traffic safety issues are becoming increasingly prominent. In complex road networks, ramp merging areas are considered potentially high-risk zones due to frequent vehicle interactions and complex traffic dynamics. Vehicles on main roads and ramps typically need to decelerate and accelerate to adjust their positions and speeds for safe merging. This not only increases the driving burden but also significantly impacts overall traffic flow efficiency.

[0003] Autonomous driving technology offers a potential solution to these problems. However, most existing systems still rely on single-vehicle intelligence, with decisions based entirely on onboard perception, lacking information sharing and explicit collaboration. In scenarios such as ramp merging, inter-vehicle cooperation is crucial for safe and efficient operation, and the single-vehicle intelligence paradigm is no longer sufficient.

[0004] Some studies have attempted to combine V2X communication with centralized controllers for cooperative trajectory planning. While this approach is effective to some extent, centralized methods have inherent limitations in terms of scalability and real-time performance because they require solving global optimization problems and are overly reliant on infrastructure support. These drawbacks have prompted research to increasingly focus on distributed cooperative control.

[0005] To overcome the aforementioned limitations, the development of distributed cooperative control frameworks is gradually becoming a more promising direction. V2X technology based on 5G New Radio (NR) is particularly suitable for scenarios requiring real-time information exchange and collaborative decision-making due to its ultra-low latency, high reliability, and high bandwidth. NR V2X, through its sidelink interface, enables direct communication between vehicles and is one of the key technologies for achieving advanced cooperative autonomous driving. Furthermore, NR V2X Mode 2 introduces a resource reassessment mechanism, enabling vehicles to periodically reassess their reserved resources based on updated perception results and network conditions. This mechanism aims to reduce persistent resource conflicts and improve channel utilization efficiency. However, in high-density traffic environments, the distributed resource allocation mechanism of NR V2X is still prone to resource collisions, potentially leading to transmission delays or even the loss of critical safety information, thereby weakening the effectiveness of upper-layer cooperative control strategies.

[0006] Meanwhile, in the field of cooperative vehicle control, distributed model predictive control (DMCC) has become an important method for trajectory planning and coordination in autonomous vehicles due to its predictive decision-making capabilities and flexibility in handling complex constraints. However, most existing DMCC frameworks still assume ideal and latency-free inter-vehicle communication, ignoring the unavoidable communication uncertainties in real-world V2X systems, such as transmission delays, packet loss, and random interference. These idealized assumptions significantly limit the controller's ability to ensure safety and achieve optimal performance under actual communication conditions. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication awareness. This method aims to explicitly incorporate the uncertainties of NR-V2X communication into the control closed loop, thereby ensuring driving safety in environments with limited communication resources.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A distributed cooperative control method for merging vehicle-to-everything (V2X) ramps based on communication awareness includes the following steps: S1. Establish a system model for the vehicle-to-everything (V2X) ramp merging scenario, including a vehicle kinematics model based on a bicycle model and a resource reservation communication model based on NR V2X mode 2. S2. A communication conflict probability prediction model based on Markov chains is adopted. Based on the current communication status, the vehicle predicts the resource selection conflict probability sequence in the future prediction time domain. S3. Construct a communication-aware distributed model predictive control optimization problem, and introduce adaptive communication security margin. The conflict probability obtained in step S2 is then coupled into the objective function as a penalty term weight. S4. The optimization problem is solved using a block coordinate descent framework. The original problem is decomposed into a vehicle control subproblem and a communication safety margin optimization subproblem. The optimal control input and communication safety margin are calculated by alternating optimization. S5. The vehicle performs acceleration or steering operations based on the calculated optimal control input to achieve a safe and smooth ramp merge.

[0009] The present invention further defines the technical solution as follows: Preferably, in step S1, the vehicle kinematics model adopts a bicycle model, and the state variables are: , representing the lateral and longitudinal coordinates of the vehicle's rear axle center, the heading angle, and the longitudinal velocity, respectively; the control variable is , representing longitudinal acceleration and front wheel steering angle, respectively; The vehicle kinematics model is discretized using the forward Euler method, with a sampling time of [time value missing]. , No. The discrete state equation at time t is: ; in For continuous-time vehicle kinematics model, This is the discretized vehicle kinematics model.

[0010] Preferably, the Markov chain model in step S2 includes collision states. and transmission status State transition matrix Defined as: ; in, These represent the submatrices for collision state preservation, collision-to-transmission, transmission-to-collision, and transmission state preservation, respectively; the vehicle in the prediction time domain. The first Communication collision probability of step The following is obtained through recursive prediction of the state distribution: ; ; in The initial state distribution vector, This is the upper limit of the reselection counter.

[0011] Preferably, in step S3, the communication-aware penalty function that introduces adaptive communication security margin... The expression is: ; in, For the predicted probability of communication collisions, For adaptive communication security margin variables, These are the weighting coefficients. These are the parameters for regularization and normalization.

[0012] Preferably, in step S4, the vehicle control sub-problem is the fixed communication security margin. Optimize vehicle control input To meet real-time requirements, the nonlinear vehicle dynamics constraints are linearized based on the nominal trajectory, transforming the vehicle control subproblem into a quadratic programming problem to be solved. in The coefficient matrix consists of the weight matrix and the linearization parameters. , To control input constraints. The communication security margin optimization subproblem is a fixed control input. Optimize communication security margin sequence This subproblem is a convex optimization problem, and its objective function is to minimize the sum of communication-aware penalty terms for all steps in the prediction time domain. in, It is the prediction time domain length of the model predictive control, where, Based on the safe distance, For vehicles With vehicles The Euclidean distance function between them can be obtained from the double-circle model; The vehicle state is defined by this constraint, which ensures that the actual distance between the two vehicles is not less than the sum of the basic safety distance and the adaptive communication safety margin.

[0013] Preferably, the submatrices of the state transition matrix are defined as follows: Collision state preservation and transition submatrix This submatrix describes the vehicle's current collision state. The behavior of transitioning to the collision state in the next moment includes maintaining the collision state. Or reset to the initial collision state , from Transfer to probability for: ; when When the resource reservation conflict persists, the vehicle will remain in a collision state until the reselection counter reaches zero, therefore the transition probability is 1; when When, transition probability Defined as: ; from Transfer to probability for: ; in ; Collision transfer transmission matrix This submatrix describes the vehicle's state after a collision. Restore to transmission state ; behavior; from Transfer to probability for: ; in ; from Transfer to probability for ; in The This indicates a situation where the interfering vehicle reselects another resource, while the current vehicle continues to use the current resource; Transmission to Collision Submatrix This submatrix describes the vehicle's transmission state. Entering a collision state If the vehicle reselects resources and collides with another vehicle, it transitions from the transmission state to the collision state, with a transition probability of... = ; Transmission state preservation submatrix This submatrix describes the behavior of the vehicle in maintaining the transmission state, from Transfer to probability for: ; in This represents the probability that the reselection counter has not yet returned to zero; from Transfer to probability for: ; in ,and ; in, This is the lower limit of the reselection counter. To maintain probability for resources, The sum of the probabilities of reserving resources for both vehicles or for one vehicle to reserve resources. The probability of reselection, For resource reselection collision probability, , , , These are the calculated probability values ​​for the corresponding transition behaviors.

[0014] Preferably, in step S3, the constraints of the distributed model predictive control optimization problem include: Control constraints: ; Dynamic constraints: ; Communication sensing safety distance constraints: ; Margin range constraints: . Beneficial effects

[0015] Compared with existing technologies, it has the following advantages: The Markov model established in this invention can accurately characterize the dynamic characteristics of NR-V2X channels. Through a communication-aware distributed cooperative control framework, the safety distance is dynamically adjusted according to the predicted communication quality. When communication is unreliable, the distance is actively increased to reserve a safety buffer, effectively preventing collisions caused by information loss and significantly improving security. This invention decomposes a non-convex coupled problem into two convex subproblems using the BCD framework and combines linearization techniques. By combining linearization techniques, the computational complexity is reduced from cubic to near quadratic, ensuring that the algorithm can run in real time on an on-board computing platform. It has extremely high computational efficiency, and the computation time growth rate is significantly lower than that of existing methods (such as IPOPT and traditional QP methods) when the number of vehicles increases. It takes into account both the scalability and real-time performance of multi-vehicle collaboration. Meanwhile, compared with traditional model predictive control that assumes ideal communication, this invention effectively reduces control behaviors such as rapid acceleration, sudden braking, and sharp steering by optimizing the balance between the communication perception penalty term and the margin cost in the objective function. This makes the changes in vehicle acceleration and steering angle smoother, which not only improves the comfort of drivers and passengers, but also avoids the disturbance of the overall traffic flow caused by the sharp operation of a single vehicle. It helps to maintain the smoothness of traffic in the ramp merging area and achieve the dual goals of safe merging and efficient passage. Attached Figure Description

[0016] Figure 1 is a schematic diagram of the scenario of the present invention; Figure 2 For the kinematic model of the vehicle; Figure 3 The vehicle is approximated by a double circle model; Figure 4 For the resource selection model in NR V2X; Figure 5 This is a communication analysis model based on Markov chains; Figure 6 The figure shows a comparison between the simulation results and the analysis results of the communication interruption probability.

[0017] Figure 7 This is a graph showing the changes in steering angle and acceleration of vehicles at different times during the merging process on the ramp; Figure 8 A comparison chart showing the computation time of different algorithms during a single ramp merging process; Figure 9 This chart compares the average computation time of each algorithm for different numbers of vehicles. Detailed Implementation

[0018] 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, and 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. Example

[0019] This embodiment provides a distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication awareness. This method explicitly integrates the reliability prediction of 5G NR-V2X communication into a distributed model predictive control framework. By adaptively adjusting the safety margin between vehicles, it solves the control safety problem caused by communication resource conflicts. Specifically, it includes the following steps: S1. Establish a system model for the vehicle-to-everything (V2X) ramp merging scenario, including vehicle kinematics modeling and geometric approximation modeling; S101. Establish the vehicle kinematics model: such as Figure 2 As shown, the kinematic model of the vehicle adopts the bicycle model, and the state variables are... , representing the lateral and longitudinal coordinates of the vehicle's rear axle center, the heading angle, and the longitudinal velocity, respectively; the control variable is , representing longitudinal acceleration and front wheel steering angle respectively; the control variables are , where represents longitudinal acceleration and front wheel steering angle, respectively. The sideslip angle is the angle at the vehicle's center of gravity. and These are the distances from the vehicle's center of gravity to the front and rear axles, respectively. To facilitate application in digital controllers, the above model is discretized using the forward Euler method, with a sampling time of [missing information]. , No. The discrete state equation at time t is: ; in For continuous-time vehicle kinematics model, This is the discretized vehicle kinematics model; S102. Establish a geometric approximation model of the double-circle vehicle: In order to efficiently calculate collision constraints in the optimization problem, such as... Figure 3 As shown, the rectangular outline of the vehicle is approximated by two circles of radius r; the vehicle is defined. The positions of the two center points, the front center point is marked as The center position of the circle is marked as Any two cars and The center-to-center distance vector between them The calculation is as follows: ; in, The distance from the center of the circle to the geometric center of the vehicle. The geometric condition for vehicles to avoid collision is that the Euclidean distance between the centers of any two circles is greater than a safety threshold. S2. Using a Markov chain-based communication conflict probability prediction model, the channel state changes under the resource reservation and reassessment mechanism in NR V2X mode 2 are quantified, specifically including the following sub-steps. Re-evaluation mechanisms in S201 and NR V2X: In complex traffic scenarios, vehicle topology changes frequently. Once a resource is selected by a traditional semi-persistent scheduling resource selection mechanism, conflicts may occur at subsequent times due to environmental changes. This re-evaluation mechanism allows vehicles to periodically evaluate the effectiveness and channel quality of their currently selected resources at the MAC layer and trigger resource reselection when necessary, thereby proactively mitigating resource conflicts and link degradation.

[0020] The core logic of the re-evaluation mechanism is to perform another check before sending data, such as... Figure 4 As shown, the specific steps are as follows: Assume the vehicle initially selects a time slot. Sending data, it will be done before sending. Moment, that is Perform a new resource-aware operation. At that moment, the vehicle judged Check if the resource is still available or occupied. If the originally selected resource is found to be invalid, abandon the resource, and the vehicle will appear in a new selection window. Reselect resources within; S202. Calculation of resource conflict probability under the reassessment mechanism: Although the reassessment mechanism aims to reduce conflict, it can still fail in certain situations, mainly in the following two ways: If the reserved resources are located in the non-reassessment period, that is, the processing time after the sensing ends for the resource location. The vehicle could not complete the inspection within the specified timeframe. Defined as: ; in, Select the total window length for the resource. and These represent the start time and processing time limit of the selection window, respectively. Assume the resource selection window length for the interfering vehicle is... The length of the reselection window generated after this vehicle triggers a reassessment is [length missing]. Define the resource reservation interval as RRI. For a collision to occur, the selection windows of the two vehicles must overlap in the time domain. The probability of the two vehicles' selection windows overlapping. The calculation is as follows: ; Here, the denominator represents the total number of possible combinations of the two car window positions, and the numerator represents the number of combinations where the two car windows do not overlap at all. If the windows overlap, the degree of overlap needs to be further quantified, i.e., the number of overlapping time slots. Let the overlap length be a random variable $O$. According to the expected tail sum formula, the expected value of the overlap length is... Represented as: ; In order to calculate First calculate their complementary events. : ; in, Indicates that the overlap length does not exceed The number of window position combinations. This item consists of two parts: the case of no overlap. The number of combinations is ,in Secondly, the overlap length is exactly 1 / 2. situation Each length The corresponding number of combinations is .therefore, The form is: ; This formula provides the statistical average of the number of time slots that overlap in the time domain between the two vehicle selection windows; Given the expected value of the overlap length between the windows of the two vehicles Under the given conditions, calculate the conditional probability that both vehicles will choose to land on resources within the overlapping area. Assume each time slot contains Sub-channels, and Let be the total number of available resources in the selection windows of the vehicle and the interfering vehicle, respectively. The probability that both vehicles simultaneously select resources within the overlapping area is: ; In the above formula, the numerator represents the square of the total number of resources in the overlapping region, and the denominator represents the product of the total available resources for each vehicle. Combining the above steps, consider the size of the candidate resource set. and And the number of resources excluded by the two vehicles. The final reselection collision probability The calculation is as follows: ; This formula combines the probability of window overlap, the probability of resources falling into the overlap area, and the probability of selecting the same resource block in the candidate resource set, and is used to quantify the risk of newly selected resources clashing again under the re-evaluation mechanism.

[0021] S203. To describe the dynamic characteristics of the resource reservation process, the system state space is defined to include two types of states, such as... Figure 5 As shown, where This indicates that the vehicle is in a collision. This indicates that the vehicle is in a transmission state. Status indicator. The value of the corresponding reselection counter ranges from 1 to 1. ,in and These represent the lower and upper bounds of the reselection counter, respectively. From this, we can construct the state transition matrix, which consists of four sub-matrices: ; The state transition matrix Used to describe the complete state transition behavior of a vehicle within one SPS cycle, its dimensions are: The following details the definition and derivation of each submatrix.

[0022] Collision state preservation and transition submatrix This submatrix describes the vehicle's current collision state. The behavior of transitioning to the collision state in the next moment includes maintaining the collision state. Or reset to the initial collision state , from Transfer to probability for: ; when When the resource reservation conflict persists, the vehicle will remain in a collision state until the reselection counter reaches zero, therefore the transition probability is 1; when When, transition probability Defined as: ; The It consists of two parts: the first part represents the probability that both the reselection counters of this vehicle and the interfering vehicle are greater than zero; the second part represents the probability that the interfering vehicle's reselection counter has been reduced to zero, but is maintained at the resource retention probability. The probability of continuing to occupy the same resource.

[0023] from Transfer to probability for: ; in In this formula, This represents the sum of the following two scenarios: (i) both car reselection counters return to zero and both are set to a probability. (ii) Reserve resources, interfere with the vehicle reselection counter being non-zero while the vehicle uses probability. Reserve resources. The calculation formula is: ; Second item This indicates that the vehicle will reselect resources, but still based on probability. The circumstances in which a collision occurs, among which The probability of reselection: ; In summary, submatrix Represented as: ; Collision transfer transmission matrix This submatrix describes the vehicle's state after a collision. Restore to transmission state ; behavior; from Transfer to probability for: ; in The first term in the formula represents the vehicle retaining the current resource while interfering with other vehicles reselecting other resources, thereby resolving the conflict; the second term represents the vehicle reselecting a new resource and successfully avoiding a collision; therefore, the submatrix... Represented as: ; Transmission to Collision Submatrix This submatrix describes the vehicle's transmission state. Entering a collision state The behavior of this vehicle. If this vehicle reselects resources and collides with another vehicle, it transitions from the transmission state to the collision state, with a transition probability. equal Submatrix Represented as: ; from Transfer to probability for ; in The This indicates a situation where the interfering vehicle reselects another resource, while the current vehicle continues to use the current resource; Transmission to Collision Submatrix This submatrix describes the vehicle's transmission state. Entering a collision state If the vehicle reselects resources and collides with another vehicle, it transitions from the transmission state to the collision state, with a transition probability of... = ; Transmission State Preservation Submatrix This submatrix describes the behavior of the vehicle in maintaining the transmission state, from Transfer to probability for: ; in This represents the probability that the reselection counter has not yet returned to zero; from Transfer to probability for: ; in ,and ; in, This is the lower limit of the reselection counter. To maintain probability for resources, The sum of the probabilities of reserving resources for both vehicles or reserving resources for only one vehicle. For the probability of reselection, For resource reselection collision probability, , , , These are the calculated probability values ​​for the corresponding transition behaviors. Submatrix Represented as: ; S204, at each moment of the model predictive control controller Based on the currently observed communication feedback and the value of the reselection counter, initialize the state distribution vector. If current communication fails, initialize to a collision state; otherwise, initialize to a transmission state. Utilize the state transition matrix. For future prediction time domain Every moment inside Perform recursive deduction: ; Finally, the first Predicting the probability of communication collisions in steps It is the sum of the probabilities of all collision states in the distribution vector at that moment: ; This probability sequence will be output as a weight parameter to the optimization objective function in step S3; S3. Construct a communication-aware distributed model predictive control optimization problem, and introduce adaptive communication security margin. Specifically: S301, at the time Vehicles are based on the prediction time domain Constructing nominal trajectories Simultaneously, the predicted state distribution generated by the Markov chain in step S2 is utilized. Calculate the probability sequence of communication conflicts for each future step. ; S302. Construct the following nonlinear predictive control problem, where the optimization variables include the control input sequence. and communication security margin sequence : ; The constraints include: Control constraints: ; Dynamic constraints: ; Communication sensing safety distance constraints: This constraint requires the Euclidean distance between the two workshops. It must be greater than the basic safety distance With adaptive margin sum; Margin range constraints: ; S4. The optimization problem is solved using a block coordinate descent framework. The original problem is decomposed into a vehicle control subproblem and a communication safety margin optimization subproblem. The optimal control input and communication safety margin are calculated through alternating optimization, specifically as follows: S401, Vehicle Control Optimization, in this subproblem, a fixed communication security margin sequence. Only optimize control input To improve computational efficiency, a linearization technique is used to transform the problem into a quadratic programming problem, utilizing the nominal trajectory from the previous time step. For nonlinear vehicle dynamics equations Performing a first-order Taylor expansion, we obtain the linear state-space equations: ; in and For Jacobian matrices, This is the affine offset term. For non-convex distance functions... A first-order Taylor expansion is performed, and slack variables are introduced to handle constraint violations, transforming the safety constraints into linear inequality constraints. The linearized objective function and constraints are then rearranged into a standard quadratic programming form: ; The optimal control increment can be quickly obtained using the OSQP solver. S402, Communication security margin optimization, in this subproblem, fixed control input Only optimize communication security margin Only retain with The relevant term is minimizing the sum of the communication penalty cost and the margin size cost: ; Calculate the Hessian matrix of the objective function. Due to the different time steps... They are mutually decoupled, and the Hessian matrix is ​​a diagonal matrix. It is derived that the diagonal elements are always greater than 0 within the domain: ; This proves that the subproblem is a strictly convex optimization problem and has a unique global optimal solution; S403, the controller executes the following loop at each time step $k$: a. Initialization: Receive the nominal trajectory of the neighboring vehicle and initialize the state of the current vehicle; b. Fixed Solve QP(S401) and update the control input. and predicted trajectory ; c. Update the Markov chain state distribution based on the new predicted trajectory. ,fixed Solve the convex optimization (S402), and update. ; d. The first control variable after convergence Send to the vehicle for execution; S5. The vehicle performs acceleration or steering operations based on the calculated optimal control input to achieve a safe and smooth ramp merge.

[0024] like Figure 6As shown, the results demonstrate a high degree of consistency between the theoretical analysis and simulation results, where the communication failure probability represents the probability that a vehicle fails to successfully transmit a data packet. With increasing vehicle density, the communication failure probability rises significantly due to intensified resource competition in high-density traffic scenarios. Under different Resource Reservation Interval (RRI) settings, the communication failure probability also exhibits a clear increasing trend, with this effect being particularly pronounced when RRI = 20 ms. This is because a shorter RRI results in a higher transmission frequency within the same time interval, leading to longer resource occupancy time and intensified resource competition, ultimately increasing the communication failure probability. Overall, both the theoretical analysis and simulation results fall within one standard deviation, verifying the accuracy and reliability of the proposed Markov model in characterizing communication collision probabilities.

[0025] like Figure 7 The figure illustrates the changes in acceleration and steering angle over time during the entire ramp merging process under different algorithms. The methods compared include OSQP and IPOPT. OSQP represents a standard linear model predictive control method that does not consider communication awareness, while IPOPT, although considering communication awareness, does not linearize the problem. As shown in the figure, the results obtained by the proposed method are generally consistent with those of IPOPT, with only minor differences. In comparison, the proposed method achieves a lower and smoother acceleration curve compared to OSQP, thus improving ride comfort. Regarding the steering angle, although the proposed method produces a slightly higher steering angle, its overall change is more stable, while the OSQP method exhibits more pronounced fluctuations. The fundamental reason for these differences lies in the introduction of a communication cost term into the objective function of the proposed method, allowing the controller to explicitly consider potential communication interruptions during optimization, resulting in a more stable control input. Even under deteriorating channel conditions, this method maintains strong robustness and overall system stability.

[0026] like Figure 8 As shown, the computation time distribution of different algorithms at each time step is illustrated in the ramp merging scenario. Each subgraph in the figure corresponds to one algorithm, each row represents one vehicle, and the time increases from left to right. The color intensity of each small rectangle indicates the computation time at the corresponding moment. It can be seen that the proposed method achieves computation time comparable to the OSQP method while maintaining superior control performance, and significantly outperforms the IPOPT method. This advantage is mainly due to the linearization processing of the original problem near the nominal trajectory, which transforms the optimization problem into a quadratic programming form, thereby significantly improving computational efficiency.

[0027] like Figure 8As shown, the comparison results of the average computation time of various algorithms under different vehicle numbers in the ramp merging scenario are presented. It can be clearly observed that as the number of vehicles involved in control increases, the computation time of each algorithm generally increases. This phenomenon indicates that as the system scales up, the dimensionality of the optimization problem increases accordingly, and the number of constraints and variables that need to be processed increases significantly, leading to a heavier computational load and a longer computation time. This also reflects the sensitivity of multi-vehicle cooperative optimization problems to computational resources in practical applications.

[0028] like Figure 9 As shown, further comparison of algorithm performance reveals that when the number of vehicles is 6, the average computation time of the proposed algorithm is slightly higher than that of the OSQP method; however, as the number of vehicles increases, the rate of increase in computation time slows down significantly, eventually falling below that of the OSQP method. Throughout the entire range of vehicle numbers, the proposed algorithm consistently maintains a significantly lower computation time than the IPOPT method. This indicates that the algorithm effectively controls computational overhead while maintaining optimized performance, ensuring real-time control capabilities.

[0029] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0030] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A distributed cooperative control method for merging on vehicle-to-everything (V2X) ramps based on communication sensing, characterized in that, Includes the following steps: S1. Establish a system model for the vehicle-to-everything (V2X) ramp merging scenario, including a vehicle kinematics model based on a bicycle model and a resource reservation communication model based on NR V2X mode 2. S2. A communication conflict probability prediction model based on Markov chains is adopted. Based on the current communication status, the vehicle predicts the resource selection conflict probability sequence in the future prediction time domain. S3. Construct a communication-aware distributed model predictive control optimization problem, and introduce adaptive communication security margin. The conflict probability obtained in step S2 is then coupled into the objective function as a penalty term weight. S4. The optimization problem is solved using a block coordinate descent framework. The original problem is decomposed into a vehicle control subproblem and a communication safety margin optimization subproblem. The optimal control input and communication safety margin are calculated by alternating optimization. S5. The vehicle performs acceleration or steering operations based on the calculated optimal control input to achieve a safe and smooth ramp merge.

2. The distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication sensing as described in claim 1, characterized in that, In step S1, the vehicle kinematics model adopts a bicycle model, and the state variables are: , representing the lateral and longitudinal coordinates of the vehicle's rear axle center, the heading angle, and the longitudinal velocity, respectively; the control variable is , representing longitudinal acceleration and front wheel steering angle, respectively; The vehicle kinematics model is discretized using the forward Euler method, with a sampling time of [time value missing]. , No. The discrete state equation at time t is: ； in For continuous-time vehicle kinematics model, This is the discretized vehicle kinematics model.

3. The distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication sensing as described in claim 1, characterized in that, The Markov chain model in step S2 includes collision states. and transmission status State transition matrix Defined as: ; in, These represent the submatrices for collision state preservation, collision-to-transmission, transmission-to-collision, and transmission state preservation, respectively; the vehicle in the prediction time domain. The first Communication collision probability of step The following is obtained through recursive prediction of the state distribution: ; ; in The initial state distribution vector, This is the upper limit of the reselection counter.

4. The distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication sensing as described in claim 1, characterized in that, In step S3, the communication-aware penalty function that introduces adaptive communication security margin is... The expression is: ; in, For the predicted probability of communication collisions, For adaptive communication security margin variables, These are the weighting coefficients. These are the parameters for regularization and normalization.

5. The distributed cooperative control method for merging vehicle-to-everything (V2X) ramps based on communication sensing according to claim 1, characterized in that, In step S4, the vehicle control sub-problem is the fixed communication security margin. Optimize vehicle control input To meet real-time requirements, the nonlinear vehicle dynamics constraints are linearized based on the nominal trajectory, transforming the vehicle control subproblem into a quadratic programming problem to be solved. in The coefficient matrix consists of the weight matrix and the linearization parameters. , To control input constraints. The communication security margin optimization subproblem is a fixed control input. Optimize communication security margin sequence This subproblem is a convex optimization problem, and its objective function is to minimize the sum of communication-aware penalty terms for all steps in the prediction time domain. in, It is the prediction time domain length of the model predictive control, where, Based on the safe distance, For vehicles With vehicles The Euclidean distance function between them can be obtained from the double-circle model; The vehicle state is defined by this constraint, which ensures that the actual distance between the two vehicles is not less than the sum of the basic safety distance and the adaptive communication safety margin.

6. The distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication sensing according to claim 3, characterized in that, The submatrices of the state transition matrix are defined as follows: Collision state preservation and transition submatrix This submatrix describes the vehicle's current collision state. The behavior of transitioning to the collision state in the next moment includes maintaining the collision state. Or reset to the initial collision state , from Transfer to probability for: ； when When the resource reservation conflict persists, the vehicle will remain in a collision state until the reselection counter reaches zero, therefore the transition probability is 1; when When, transition probability Defined as: ； from Transfer to probability for: ； in ; Collision transfer transmission matrix This submatrix describes the vehicle's state after a collision. Restore to transmission state ; behavior; from Transfer to probability for: ； in ; from Transfer to probability for ； in The This indicates a situation where the interfering vehicle reselects another resource, while the current vehicle continues to use the current resource; Transmission to Collision Submatrix This submatrix describes the vehicle's transmission state. Entering a collision state If the vehicle reselects resources and collides with another vehicle, it transitions from the transmission state to the collision state, with a transition probability of... = ; Transmission state preservation submatrix This submatrix describes the behavior of the vehicle in maintaining the transmission state, from Transfer to probability for: ; in This represents the probability that the reselection counter has not yet returned to zero; from Transfer to probability for: ; in ,and ; in, This is the lower limit of the reselection counter. To maintain probability for resources, The sum of the probabilities of reserving resources for both vehicles or for one vehicle to reserve resources. The probability of reselection, For resource reselection collision probability, , , , These are the calculated probability values ​​for the corresponding transition behaviors.

7. The distributed cooperative control method for vehicle-to-everything (V2X) ramp merging based on communication sensing according to claim 4, characterized in that, In step S3, the constraints of the distributed model predictive control optimization problem include: Control constraints: ; Dynamic constraints: ; Communication sensing safety distance constraints: ; Margin range constraints: .

Images