Control information value evaluation method based on model prediction for networked automobiles

Abstract

The invention discloses a networked automobile-oriented control information value evaluation method based on model prediction, and relates to the technical field of vehicle control, and the method mainly comprises the steps: obtaining control information which is sent by a front vehicle and comprises a plurality of future time slot prediction state sequences of the front vehicle; taking the control information as known input, solving a model prediction control problem taking minimization of a preset objective function as a core, obtaining a first optimal control sequence, and calculating a first control performance index; under the condition that the control information is assumed to be missing, generating an alternative control information sequence through extrapolation based on the historical control information; taking the alternative control information sequence as known input, solving the model prediction control problem to obtain a second optimal control sequence, and calculating to obtain a second control performance index; and calculating a difference value between the first control performance index and the second control performance index, and defining the difference value as an information value of the control information. According to the invention, dynamic quantification of control information value and intelligent scheduling of communication resources are realized.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, and specifically to a model-predictive-based method for evaluating the value of control information for connected vehicles. Background Technology

[0002] With the evolution of cellular vehicle-to-everything (C-V2X) technology, connected autonomous vehicles (CAVs) can exchange a large amount of real-time data, including cooperative perception information (CAM) and cooperative perception information (CPM), providing an information foundation for collaborative control. Model predictive control (MPC), as an advanced control method, is widely used in fleet collaborative operation. It can utilize this shared information to continuously optimize future control behavior while meeting vehicle dynamics and safety constraints, thereby improving the stability and safety of fleet operation. However, in real-world high-density traffic scenarios, vehicle-to-everything (V2X) communication resources are extremely limited. Bandwidth constraints, channel congestion, and interference inevitably lead to latency and packet loss in communication links. Currently, the resource allocation mechanism of V2X systems mainly relies on fixed rules defined by standards or scheduling strategies based on communication quality (such as channel busy rate). While these methods can reduce some redundant transmissions, they have a fundamental flaw: a lack of control perception capability, meaning they cannot dynamically quantify the intrinsic value of each piece of information from the perspective of control performance optimization. This means that high-value information crucial for collaborative control (e.g., the predicted trajectory of the vehicle ahead's emergency braking) is treated equally with routine updates in resource competition, and the system cannot intelligently prioritize the transmission of critical control information during congestion. The direct consequence is that when communication load increases, the control system may fail to obtain the most critical status updates in a timely manner, leading to increased following errors, decreased queue stability, and even safety issues. Therefore, existing technologies struggle to consistently ensure the low latency and high reliability required for collaborative control in resource-constrained and complex vehicle-to-everything (V2X) environments. A new mechanism is urgently needed to accurately assess the contribution of information to control performance and use this information to guide communication decisions. Summary of the Invention

[0003] To better assess the contribution of information to control performance, this invention proposes a model-predictive-based control information value assessment method for connected vehicles. The specific steps for assessing the value of the target vehicle's own control information include the following: S1: Obtain control information sent by the preceding vehicle, which contains a sequence of predicted states for several future time slots of the preceding vehicle; S2: Using control information as known input, solve the model predictive control problem with minimizing the preset objective function as the core, obtain the first optimal control sequence and calculate the first control performance index; S3: In the case of missing control information, generate a sequence of alternative control information by extrapolation based on historical control information; S4: Using the alternative control information sequence as known input, solve the model predictive control problem to obtain the second optimal control sequence and calculate the second control performance index; S5: Calculate the difference between the first control performance index and the second control performance index, and define the difference as the information value of the control information.

[0004] Furthermore, the preset objective function is used to evaluate the control cost of a single time slot, and its components include at least one of the following: the state error between the target vehicle and the preceding vehicle, the absolute value of the target vehicle's acceleration, and the rate of change of the target vehicle's acceleration.

[0005] Furthermore, the expression for the preset objective function is: In the formula, The loss value for a single time slot. The predicted state of the target vehicle. For the control information of the vehicle in front, The preset state error vector, To predict acceleration, To predict the rate of change of acceleration, These are the weighting coefficients. For maximum acceleration, It is a single time slot interval.

[0006] Furthermore, the first control performance index With the second control performance index The calculation method is as follows: In the formula, This represents the total number of time slots. Let be the objective function value for the k-th time slot.

[0007] Furthermore, the value assessment method is performed in a vehicle queue based on a forward topology, where each vehicle acts as the preceding vehicle to its following vehicle and sends its own predicted state sequence to its following vehicle.

[0008] Furthermore, both the predicted state sequence and the alternative control information sequence contain predicted positions and predicted velocities for several future time slots.

[0009] Furthermore, the solution to the model predictive control problem is subject to terminal constraints, which are used to force the target vehicle to maintain a preset distance and the same speed as the preceding vehicle at the end of several future time slots.

[0010] Furthermore, step S5 is followed by the following step: S6: Compare the information value with a preset threshold. When the information value is higher than the preset threshold, the target vehicle sends its control information to the following vehicle.

[0011] Compared with the prior art, the present invention has at least the following beneficial effects: (1) The present invention proposes a model prediction-based control information value assessment method for connected vehicles, which deeply integrates the performance indicators of model prediction control with the information value theory, thereby providing a precise and direct decision basis for the allocation of vehicle network resources. (2) By simulating the difference in control performance under two hypothetical scenarios of having and lacking control information of the preceding vehicle, the importance of specific information for achieving safe, efficient and comfortable control of the vehicle can be dynamically and quantitatively evaluated. (3) By prioritizing the transmission of high-value information and suppressing low-value redundant information, channel congestion can be significantly reduced under limited communication bandwidth, the overall efficiency of information interaction can be improved, key control information can be guaranteed, and the tracking accuracy, queue stability and driving comfort of the vehicle cooperative control system can be effectively improved, thereby enhancing the robustness and safety of the autonomous driving system in complex traffic environments. Attached Figure Description

[0012] Figure 1 A flowchart illustrating the steps of a model-predictive-based control information value assessment method for connected vehicles; Figure 2 A comparison of the control performance of MPC and DDPG, which have good training results; Figure 3 A comparison of the control performance of MPC and DDPG, which have poor training results; Figure 4 Communication strategies under two different control methods; Figure 5 The control performance under two communication strategies. Detailed Implementation

[0013] This invention aims to provide a practically implementable control information value assessment scheme. To make the above-mentioned objectives, features, and advantages of this invention more apparent and easier to understand, the following will elaborate on the control information value assessment method for connected vehicles based on model prediction, in conjunction with specific embodiments and simulation experimental results. This embodiment will systematically demonstrate the complete implementation process of the method in a typical multi-vehicle platoon following scenario, including the establishment of its system model, the specific execution steps of the distributed model predictive control algorithm, the core calculation process of information value assessment, and the final communication decision made based on the value assessment results. It should be understood that the specific embodiments and simulation parameters described herein are only for explaining this invention and are not intended to limit the scope of protection of this invention. Under the guidance of the inventive concept, those skilled in the art can make various adjustments and transformations to the method and parameters, and all such adjustments and transformations should fall within the scope of protection defined by the claims of this invention. Figure 1 As shown, the present invention proposes a model-predictive-based control information value assessment method for connected vehicles. The specific steps for assessing the value of the target vehicle's own control information include the following: S1: Obtain control information sent by the preceding vehicle, which contains a sequence of predicted states for several future time slots of the preceding vehicle; S2: Using control information as known input, solve the model predictive control problem with minimizing the preset objective function as the core, obtain the first optimal control sequence and calculate the first control performance index; S3: In the case of missing control information, generate a sequence of alternative control information by extrapolation based on historical control information; S4: Using the alternative control information sequence as known input, solve the model predictive control problem to obtain the second optimal control sequence and calculate the second control performance index; S5: Calculate the difference between the first control performance index and the second control performance index, and define the difference as the information value of the control information.

[0014] Specifically, this invention considers a vehicle queue scenario containing N vehicles, i.e. Where index 0 represents the lead vehicle, and indices 1 to N-1 represent following vehicles. Each vehicle is equipped with a communication unit and a high-precision positioning system. The convoy adopts a forward topology communication mode, meaning each vehicle only transmits relevant control information to the vehicles behind it. The control information in this invention is for future... The predicted state information for each control time slot includes predicted position and predicted speed. After receiving control information from the preceding vehicle, the following vehicle obtains the optimal control input through model predictive control, thereby achieving safe and efficient following behavior. The core problem this invention aims to solve is how the target connected autonomous vehicle defines the importance of its own generated control information relative to the control performance of its target receiving vehicle.

[0015] To establish an accurate mathematical model, this invention employs a linearly discretized forward Euler vehicle dynamics model to describe the vehicle's motion state. This model is represented by the following formula: It describes the location update; It describes the speed update; It describes the acceleration update; in, Representing vehicles In control time slot Position, velocity, and acceleration It is the time delay coefficient representing the dynamics of the transmission system. It is the time interval of the discrete control time slot. For vehicles In the time slot The control input (such as the desired acceleration). Based on this, any vehicle In the time slot The state is denoted as The corresponding status output information is recorded as This output information is the control information used for transmission. For ease of subsequent description, the above vehicle dynamics model can be uniformly represented as: , , in, , , Represented as: .

[0016] To ensure driving safety and comfort, this invention imposes constraints on control inputs: ,in and These are the upper and lower boundaries of the control input. The convoy control objective in this application scenario is to ensure that each following vehicle maintains the same speed as the lead vehicle while maintaining an ideal distance from the vehicle in front, i.e.: , in, Let i be the ideal distance between the target vehicle i and the preceding vehicle i-1. This invention employs a constant distance strategy, i.e. .

[0017] To achieve the aforementioned vehicle control objectives, during the control time slot Any vehicle Both require the use of model predictive control methods to solve for the current optimal control input. For ease of description, this invention also distinguishes three types of control inputs: Predictive control input Optimal control input and assumptions about control input ,in This indicates that the predicted time slot range is . For all feasible solutions that satisfy the constraints, It is the optimal solution among all feasible solutions; It is the optimal solution among all feasible solutions; Assuming future control inputs, used to generate relevant control information, derived from the optimal control inputs. Launch, specifically: , Accordingly, based on the dynamic model, the three control sequences correspond to three state trajectory sequences: Predicted trajectory Optimal trajectory and hypothetical trajectory Assuming the trajectory For vehicles In control time slot The generated control information is used to transmit to the target receiving vehicle. .

[0018] The model predictive control method in this invention employs rolling time-domain control. During the control time slot... For any vehicle The specific steps for obtaining optimal control using the MPC algorithm are as follows: First, the vehicle... Get your current state Then, the vehicle According to the received information from the vehicle Relevant control information To predict the future trajectory changes of the vehicle in front. Then, the vehicle... The optimal control input sequence is determined based on the current state and the received control information. , … The optimization control problem is formulated as the problem... In time slots For each Minimize the objective function: The optimization problem must satisfy the following constraints: State update equation Output equation Initial state constraints Control input constraints Terminal state constraints Force vehicles to be at the end of the prediction time domain ( The state of the vehicle in front should maintain an ideal distance from the state of the vehicle in front; terminal control input constraints. This forces the vehicle to have a zero rate of acceleration change at the final moment to ensure smooth operation. For vehicles In control input The corresponding state; Indicates the predicted preceding vehicle Trajectory changes (i.e., vehicle) (Transmitted control information). , indicating car With the car The ideal state difference vector between them.

[0019] And optimization problem functions in For vehicles The control performance objective function is defined as: , The objective function consists of state error, acceleration, and rate of change of acceleration, aiming to achieve a following effect while also considering the smoothness and comfort of vehicle operation. These are the weighting coefficients for each item. For vehicles The rate of change of acceleration is expressed as: .

[0020] By solving this optimization problem, the vehicle It is possible to obtain the current time slot Initial optimal control input sequence , … Subsequently, the vehicle Only the first control variable in the sequence is executed, i.e. After completing the control actions, the vehicle Need to be behind its vehicle Preparing for the next time slot The control information to be sent. Therefore, based on the optimal control sequence obtained in this solution, it calculates the assumed control input sequence for the next time slot. and the corresponding hypothetical state output trajectory This hypothetical trajectory This constitutes a vehicle In the time slot Control information is sent to the vehicle following it. The entire DMPC process follows the time slots. The process proceeds in a rolling fashion, enabling distributed and collaborative control of the fleet.

[0021] It should be noted that in this invention, any vehicle All of these require solving for optimal control according to the above process. Therefore, the MPC in the proposed scheme of this invention is a distributed model predictive control, which specifically includes the following two stages.

[0022] Initialization phase: At that time, it is assumed that all vehicles are moving at a constant speed, and the initialized control input and corresponding states are as follows: , in Obtained through iterative calculation: .

[0023] DMPC Iteration Phase: When At that time, any vehicle Perform the following steps in sequence: (1) Solving the optimization problem The optimal control sequence is obtained. ; (2) Using the optimal control sequence Derive the optimal state sequence within the prediction time slot range: ; (3) Using the optimal control sequence Derivation of the control input hypothesis : , At the same time, the corresponding hypothetical state output is obtained: , , ; (4) Vehicles Transmission control information Give the target receiving vehicle (The following vehicle); (5) Vehicle i uses the first term of the optimal control sequence as the control input, i.e. ; (6) Finally, time update And repeat the above iterative process.

[0024] The core of this invention's technical solution lies in quantitatively answering the following questions based on the aforementioned DMCP framework: vehicles Send to vehicle Control information What exactly is the impact on the vehicle? To what extent is the control performance improved? Therefore, the core idea of ​​this invention is: by comparing vehicle... Under the hypothetical scenarios of "having" the latest control information and "lacking" the latest control information, the difference in control performance achievable by applying the MPC strategy is used to quantify the value of this information. The specific evaluation algorithm is executed according to the following steps.

[0025] The first step is to assess the control performance when information is available. (Vehicle) In the time slot Normally receive and utilize the latest control information sent by the preceding vehicle. The optimization problem is solved according to the aforementioned DMPC process, and the first optimal control sequence is obtained. and its corresponding optimal state trajectory At this point, control performance indicators under information-rich scenarios are available. Defined as the negative value of the total performance loss function corresponding to the optimal control sequence, i.e.: , The larger this value, the better the control performance with the support of this information.

[0026] The second step is to evaluate the control performance when information is missing. (Vehicle) In the same decision-making time slot It performs a hypothetical calculation: it assumes that the preceding vehicle was not received in the time slot. Latest control information sent In this situation, the vehicle Unable to know the latest future movements of the vehicle ahead, we can only rely on the previous (i.e., time slot) Received control information Extrapolation prediction is performed using the vehicle dynamics model to generate an alternative control information sequence. Then, the vehicle Replace this sequence Using the known input, we solve the same MPC optimization problem again to obtain the second optimal control sequence. and their corresponding state trajectories Accordingly, control performance indicators under scenarios with missing information Defined as: .

[0027] The third step is to calculate the value of the information. This is done within the control time slot. Send to vehicle Control information Information Value (VoI) That is, the difference in control performance indices under the two scenarios mentioned above is obtained by calculating: , The difference It intuitively reflects the improvement in control performance resulting from acquiring the latest control information. The larger the value, the more important the control information is for the vehicle. The more critical the achievement of excellent vehicle following control, the higher its value, and therefore, its transmission should be given priority when communication resources are scarce.

[0028] To verify the effectiveness of the proposed solution, a simulation was conducted in another embodiment (Leader and Follower). The simulation first acquired a next-generation traffic simulation dataset and extracted car-following events using a car-following filter. In the experiment, the speed trajectory of the leader vehicle in each round corresponds to the speed data of the vehicle preceding it in a specific car-following event, thus simulating a real-world road environment with uncertainties. The simulation progressed with the control time slots. The update (with a control time slot of 50ms) is performed, and the specific simulation implementation process is as follows: First, the parameters are initialized, and the simulation initial time is set to... The simulation process begins by initializing the states and hypothetical state outputs of all vehicles. Then, iterates through each vehicle, checking if it has received control information from the preceding vehicle. If so, each vehicle performs MPC (Multi-Process Control) based on the preceding vehicle's hypothetical state output to obtain the optimal control sequence and records its control performance. If not, each vehicle extends and completes the previously received control information, then performs MPC again to obtain the control sequence with missing information and records its control performance. The control performance in both cases is then compared to evaluate the VoI (Voice of Information). Afterward, each vehicle updates its hypothetical state output according to the obtained optimal control sequence. Subsequently, each vehicle uses the first element of the optimal control sequence as its control input to update its vehicle state (the leading vehicle updates according to a predetermined trajectory in the data). The simulation time is incremented. Determine if the maximum simulation duration has been reached. If so, end the simulation for that scenario; otherwise, return to the judgment step.

[0029] This simulation was conducted using the Monte Carlo method, and all simulation results are based on statistical results from multiple experiments. The simulation parameters were set as follows: control time slot interval 0.05s, number of time slots per round 100, and time delay coefficient... The maximum acceleration is 0.3. The value is 6 m / s², and the weighting coefficient is... All are 1, prediction steps It is 20.

[0030] Figure 2 and Figure 3 The differences in control performance between MPC and DDPG with different training effects are shown respectively. Figure 2 The data results show that although both schemes have a certain following effect, the MPC has smaller fluctuations in position and speed errors, and its acceleration changes are more consistent with the lead car. Therefore, the MPC's control performance is superior to the DDPG. Figure 3 The data shows that when the DDPG algorithm has poor training results, various errors increase significantly and the control performance decreases greatly. However, the MPC control strategy adopted in this invention still maintains low error and good following performance, which indicates that the MPC control strategy adopted in this invention has good stability.

[0031] Figure 4 and Figure 5 This paper presents two different communication strategies and their corresponding control performance obtained through Dueling DQN training under non-ideal communication environments. The two strategies are based on MPC and TD3 (a deep reinforcement learning algorithm, a variant of DDPG), respectively. Figure 4 As shown (1 indicates sending, 0 indicates not sending), the communication strategy under the VoI driver proposed in this invention sends slightly fewer messages and occupies fewer communication resources, but from... Figure 5 It is easy to observe that it still maintains superior control performance, indicating that the VoI evaluation scheme proposed in this invention is highly efficient.

[0032] Simulation results show that the MPC method proposed in this invention outperforms the DDPG method in terms of position error, velocity error, and acceleration smoothness. Especially when DDPG training is insufficient, MPC can still maintain low tracking error and stable following behavior. Furthermore, in non-ideal communication environments, the VoI-based communication strategy can effectively reduce the number of information transmissions, lower channel load, and maintain superior control performance.

[0033] In summary, the technical solution of this invention combines the performance indicators of model predictive control with the information value theory, thereby realizing the dynamic quantification of control information value and intelligent scheduling of communication resources, significantly improving the efficiency, safety, and comfort of vehicle platoon collaborative control.

[0034] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0035] Furthermore, in this invention, descriptions involving terms such as "first," "second," and "a" are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0036] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0037] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

Claims

1. A method for evaluating the value of control information based on model prediction for connected vehicles, characterized in that, The value assessment of the target vehicle's control information includes the following steps: S1: Obtain control information sent by the preceding vehicle, which contains a sequence of predicted states for several future time slots of the preceding vehicle; S2: Using control information as known input, solve the model predictive control problem with minimizing the preset objective function as the core, obtain the first optimal control sequence and calculate the first control performance index; S3: In the case of missing control information, generate a sequence of alternative control information by extrapolation based on historical control information; S4: Using the alternative control information sequence as known input, solve the model predictive control problem to obtain the second optimal control sequence and calculate the second control performance index; S5: Calculate the difference between the first control performance index and the second control performance index, and define the difference as the information value of the control information.

2. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 1, characterized in that, The preset objective function is used to evaluate the control cost of a single time slot, and its components include at least one of the following: the state error between the target vehicle and the preceding vehicle, the absolute value of the target vehicle's acceleration, and the rate of change of the target vehicle's acceleration.

3. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 2, characterized in that, The expression for the preset objective function is: In the formula, The loss value for a single time slot. The predicted state of the target vehicle. For the control information of the vehicle in front, The preset state error vector, To predict acceleration, To predict the rate of change of acceleration, These are the weighting coefficients. For maximum acceleration, It is a single time slot interval.

4. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 3, characterized in that, The first control performance index With the second control performance index The calculation method is as follows: In the formula, This represents the total number of time slots. Let be the objective function value for the k-th time slot.

5. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 1, characterized in that, The value assessment method is performed in a vehicle queue based on a forward topology, where each vehicle acts as the preceding vehicle of its following vehicle and sends its own predicted state sequence to its following vehicles.

6. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 1, characterized in that, Both the predicted state sequence and the alternative control information sequence contain the predicted position and predicted velocity for several future time slots.

7. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 1, characterized in that, The solution to the model predictive control problem is subject to terminal constraints, which are used to force the target vehicle to maintain a preset distance and the same speed as the preceding vehicle at the end of several future time slots.

8. The method for evaluating the value of control information based on model prediction for connected vehicles as described in claim 1, characterized in that, The step S5 is followed by the following step: S6: Compare the information value with a preset threshold. When the information value is higher than the preset threshold, the target vehicle sends its control information to the following vehicle.

Images