A carbon flow optimization method, device and medium under vehicle-road cooperation edge computing

Abstract

The application discloses a kind of carbon flow optimization methods, equipment and medium under edge computing of vehicle-road cooperation, it is related to traffic carbon emission technical field.The method includes: obtaining the real-time state parameter sequence of vehicle and environmental information, at least two candidate driving strategies are generated by built-in carbon emission digital twin model;For candidate driving strategy, by querying the emission factor mapping table of vehicle specific power and integration, the corresponding expected carbon emission is calculated;Candidate driving strategy and corresponding expected carbon emission are encapsulated as bidding package, and sent to the edge computing node of roadside;Based on the received bidding package, initiate carbon right reverse auction, solve the combinatorial optimization problem using heuristic search algorithm, determine the winning strategy set;According to the winning strategy set, generate cooperative driving strategy, and issue to corresponding vehicle end node.The application realizes the suppression information distortion by the above method, realizes the direct, real-time optimization control to regional total carbon emission.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic carbon emission, and particularly relates to a carbon flow optimization method, equipment and medium for edge computing under vehicle-road cooperation. BACKGROUND

[0002] With the rapid development of intelligent transportation and vehicle-road cooperation technology, optimizing traffic flow to achieve energy saving and emission reduction has become an important research direction. The existing technical solutions mainly focus on two levels: at the macro level, traffic signal timing optimization, path induction and other means are used to improve road traffic efficiency, thereby indirectly reducing the additional emissions caused by congestion; at the micro level, the vehicle's own energy-saving driving assistance system can optimize the driving behavior of a single vehicle, but its decision is often limited to local information, lacking cooperation with surrounding vehicles and roadside systems, which may lead to local optimization but negative impact on overall traffic flow. In recent years, although some research has tried to introduce game theory or simple incentive mechanisms under the framework of vehicle-road cooperation to guide vehicles, these solutions still mainly focus on improving traffic efficiency or safety, and have not taken the total physical carbon emissions of a region as a directly quantifiable and real-time optimized control target.

[0003] Further, the existing technology has several inherent defects. First, they generally lack an effective verification mechanism for the authenticity of the information provided by vehicles, and the system cannot identify and suppress the malicious false reporting behavior of vehicles, resulting in a serious information asymmetry problem. Second, the optimization vision of these solutions is usually limited to a single intersection or road section, which is an isolated optimization, and downstream nodes cannot obtain the carbon emission characteristics of upstream vehicles, so they cannot achieve forward-looking carbon flow regulation at the road network level. Finally, existing methods treat traffic flow as a homogeneous whole and do not fully utilize the differences in vehicle type, load and energy consumption characteristics to guide in a refined and differentiated manner, resulting in one-size-fits-all control measures and the optimization potential not being fully released.

[0004] Through the above analysis, the problems and defects of the existing technology are:

[0005] The vehicle-road cooperation technology in the prior art does not first take the total physical carbon emissions of a region as a directly quantifiable and real-time optimized control target, and thus cannot directly, real-time, reliably and refinedly optimize the instantaneous physical carbon emissions of vehicles in the region. SUMMARY

[0006] The embodiments of the present application provide a carbon flow optimization method, equipment and medium for edge computing under vehicle-road cooperation, which can solve the problem that the vehicle-road cooperation technology does not first take the total physical carbon emissions of a region as a directly quantifiable and real-time optimized control target, and thus cannot directly, real-time, reliably and refinedly optimize the instantaneous physical carbon emissions of vehicles in the region.

[0007] In a first aspect, the embodiments of the present application provide a carbon flow optimization method for vehicle-road cooperation edge computing, characterized in that the method comprises: obtaining a real-time state parameter sequence based on a vehicle and environment information through a vehicle end node, generating at least two candidate driving strategies through a built-in carbon emission digital twin model; for the candidate driving strategies, calculating the corresponding expected carbon emissions by querying the emission factor mapping table of the specific power of the vehicle and integrating; encapsulating the candidate driving strategies and the corresponding expected carbon emissions into a bidding package, and sending it to the edge computing node of the roadside; based on the received bidding package, initiating a carbon right reverse auction, using a heuristic search algorithm to solve the combinatorial optimization problem, and determining the winning strategy set; generating a cooperative driving strategy according to the winning strategy set, and delivering it to the corresponding vehicle end node.

[0008] In an implementation manner of the present application, after the cooperative driving strategy is generated according to the winning strategy set and delivered to the corresponding vehicle end node, the method further comprises: obtaining the actual carbon emissions of the vehicle after the vehicle executes the cooperative driving strategy; comparing the actual carbon emissions with the expected carbon emissions to calculate the prediction error; based on the prediction error, updating the reputation score of the vehicle using a preset update rule, and the reputation score and the reputation weight coefficient are preset fixed mapping rules.

[0009] In an implementation manner of the present application, the real-time state parameter sequence and the environment information of the vehicle are obtained through the vehicle end node, and at least two candidate driving strategies are generated through the built-in carbon emission digital twin model, which specifically comprises: obtaining real-time environment information through the vehicle end node, the environment information including rainfall, visibility, road wet slip coefficient and environmental temperature; according to the rainfall, visibility and road wet slip coefficient, the safety distance is increased, and the deceleration curve is planned according to the safety distance; according to the environmental temperature, the emission correction factor is called to calibrate the output result of the carbon emission digital twin model.

[0010] In an implementation manner of the present application, the method further comprises: generating at least two candidate driving strategies based on the carbon emission digital twin model, the two candidate driving strategies including a first candidate driving strategy and a second candidate driving strategy; the first candidate driving strategy takes minimizing the impact degree in the acceleration and deceleration process of the vehicle as the optimization target, and obtains a smooth speed curve combined with the deceleration curve; the second candidate driving strategy takes minimizing the driving time through the front road section as the optimization target, and obtains a minimum time speed curve.

[0011] In an implementation form of the present application, the candidate driving strategy and the corresponding expected carbon emission are packaged as a bidding package and sent to an edge computing node at the roadside, specifically comprising: based on the edge computing node including an upstream edge computing node and a downstream edge computing node; obtaining a carbon emission inertia index and an average reputation score after the upstream edge computing node completes the candidate driving strategy; packaging the carbon emission inertia index and the average reputation score into a carbon wave protocol data packet and sending it to the downstream edge computing node; the downstream edge computing node uses the carbon wave protocol data packet for feedforward calibration.

[0012] In an implementation form of the present application, the method further comprises: obtaining traffic data based on the roadside perception device, and predicting the carbon emission of each lane-level fine-grained region within a preset time in the future using a spatio-temporal graph neural network model; when it is predicted that the carbon emission of the target region will exceed a preset threshold after a preset time, triggering a carbon right reverse auction for a subset of vehicles entering the region.

[0013] In an implementation form of the present application, based on the received bidding package, a carbon right reverse auction is initiated, a heuristic search algorithm is used to solve the combinatorial optimization problem, and a winning strategy set is determined, specifically comprising: in the objective function of the optimization problem, the expected carbon emission of each vehicle is combined with a reputation weight coefficient to reduce the winning probability; the selection of each vehicle for the candidate driving strategy is encoded as a gene, and the selection of all vehicles forms a chromosome; the total expected carbon emission of all vehicles in the region is used as the fitness function; by iteratively performing selection, crossover and mutation operations, a chromosome with the minimum fitness function value is evolved, and the solution represented by the chromosome is the winning strategy set.

[0014] In an implementation form of the present application, after obtaining the traffic data based on the roadside perception device, the method further comprises: calculating a carbon emission fairness index in the region in combination with the actual carbon emission, the fairness index being used to measure the difference in unit mileage carbon emission between different vehicles; according to the reputation score, a double-objective weight for the reverse auction is allocated, wherein the first weight corresponds to the minimum total expected carbon emission in the region, and the second weight corresponds to the optimization target of the carbon emission fairness index; the double-objective weight is integrated into the objective function to construct a double-objective optimization model; when generating the candidate driving strategy through the vehicle end node, the unit mileage expected carbon emission corresponding to each candidate driving strategy is calculated in parallel and packaged into the bidding package; after the edge computing node solves the double-objective optimization model, the solution that meets the total carbon emission and the unit mileage emission difference within a preset threshold is selected.

[0015] In a second aspect, the embodiments of the present application also provide a carbon flow optimization device for edge computing under vehicle-road cooperation, comprising at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of any one of the carbon flow optimization methods for edge computing under vehicle-road cooperation.

[0016] In a third aspect, the embodiments of the present application also provide a nonvolatile computer storage medium for carbon flow optimization of edge computing under vehicle-road cooperation, which stores computer executable instructions configured to perform the steps of any one of the carbon flow optimization methods for edge computing under vehicle-road cooperation.

[0017] The carbon flow optimization method, device and medium for edge computing under vehicle-road cooperation provided by the embodiments of the present application can realize the generation and cost prediction of differentiated candidate driving strategies through a high-fidelity digital twin model at the vehicle end, realize global optimization through a reverse auction mechanism with the introduction of a reputation weight coefficient at a roadside edge node, and dynamically update the reputation based on actual carbon emission audit data after execution, so as to directly link the incentive mechanism and the physical emission reduction target, effectively suppress information distortion, and realize direct and real-time optimization control of regional total carbon emissions. Through model self-calibration and environment adaptive strategy generation, the accuracy of carbon emission prediction and the applicability of strategies under different working conditions are significantly improved, providing reliable input for optimization decisions; through predictive triggering based on a spatio-temporal graph neural network and carbon wave protocol feedforward across nodes, the optimization range is expanded from isolated single points to road network level cooperation, realizing forward-looking dredging of carbon emission congestion; through deep fusion of reputation points and optimization objective functions, and the innovative introduction of a double-objective optimization model considering fairness and efficiency, not only is the malicious behavior suppressed at the mathematical level, but also the fairness of all participating vehicles and the long-term stability of the system are improved, and finally, sustainable regional carbon flow fine management is realized. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which are included to provide a further understanding of the present application, form a part of the present application and illustrate the illustrative embodiments of the present application and together with the description, serve to explain the present application. In the drawings:

[0019] Figure 1 A flowchart of a carbon flow optimization method for edge computing under vehicle-road cooperation provided by the embodiments of the present application;

[0020] Figure 2 An internal structure schematic diagram of a carbon flow optimization device for edge computing under vehicle-road cooperation provided by the embodiments of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] This application provides a carbon flow optimization method, device, and medium for edge computing under vehicle-road cooperation, which solves the problem that existing vehicle-road cooperation technologies do not have a directly quantifiable and real-time optimized control target for the total physical carbon emissions of a region, thus making it impossible to perform direct, real-time, reliable, and refined cooperative optimization of the instantaneous physical carbon emissions of vehicles in the region.

[0023] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0024] Figure 1 This document presents a flowchart illustrating a carbon flow optimization method for edge computing under vehicle-road cooperative architecture, as provided in an embodiment of this application. Figure 1 As shown in the embodiment of this application, a carbon flow optimization method for edge computing under vehicle-road cooperation is provided, which specifically includes the following steps:

[0025] Step 10: Obtain the real-time state parameter sequence and environmental information based on the vehicle through the vehicle end node, and generate at least two candidate driving strategies through the built-in carbon emission digital twin model.

[0026] Firstly, it can be understood that the vehicle end node is configured on the vehicle, for example, integrated in the on-board computing unit, the edge computing node is configured in the roadside infrastructure, integrated in the roadside unit, and the edge computing node is used for wireless communication with at least one vehicle end node in the communication coverage. The vehicle end node can include a high-fidelity carbon emission digital twin model and a bid package generation module, the high-fidelity carbon emission digital twin is based on real-time state information of the vehicle obtained from the vehicle bus or the like, generates no less than two candidate driving strategies, and at the same time, the high-fidelity carbon emission digital twin also predicts the corresponding expected carbon emission cost of each candidate driving strategy by using the built-in carbon emission model. The bid package generation module is connected with the high-fidelity carbon emission digital twin, encapsulates the candidate driving strategy and the corresponding expected carbon emission cost into a bid package data structure for subsequent sending. The edge computing node includes a carbon right auction house and a dynamic reputation and market access engine. The carbon right auction house is used for receiving the bid package sent by one or more vehicle end nodes, and initiating a carbon right reverse auction according to the received bid package. In the auction process, the carbon right auction house determines a winning strategy set from the candidate driving strategies contained in all received bid packages by solving a preset optimization problem. Finally, the carbon right auction house generates a cooperative driving strategy according to the winning strategy set and sends the cooperative driving strategy to the corresponding vehicle end node.

[0027] As an optional embodiment, the real-time state parameter sequence of the vehicle and the environmental information are obtained by the vehicle end node, and at least two candidate driving strategies are generated by the built-in carbon emission digital twin model, which can specifically include:

[0028] Step 101: obtaining real-time environmental information through the vehicle end node, the environmental information including rainfall, visibility, road wetness coefficient and environmental temperature.

[0029] In this step, a set of real-time state parameters of the vehicle are obtained in real time from the internal data bus of the vehicle or other vehicle-mounted sensors, which are used to represent the physical running state of the vehicle at the current time. The real-time state parameters can include instantaneous vehicle speed, instantaneous acceleration, engine speed, engine torque, current gear, vehicle total mass, and road slope of the vehicle driving section. Among them, the vehicle total mass can be determined according to the vehicle factory calibration parameters and the load sensor data. The road slope information can be provided by the vehicle-mounted inclination sensor, or by matching the real-time geographic position of the vehicle with the pre-stored high-precision map data.

[0030] Step 102: increasing the predicted safety distance according to the rainfall, visibility and road wetness coefficient, and planning a deceleration curve according to the safety distance; step 103: calling the emission correction factor according to the environmental temperature to calibrate the output result of the carbon emission digital twin model.

[0031] In this step, the high-fidelity carbon emission digital twin also includes a model self-calibration unit, which inputs the prediction error As input, an online machine learning algorithm is applied to calculate the adjustment amount of the microcosmic carbon emission model parameters; during the vehicle driving process, the parameters of the microcosmic carbon emission model stored in the model parameter storage unit are iteratively updated periodically or continuously to improve the prediction accuracy of the expected carbon emission cost.

[0032] In this step, the rainfall, visibility, road wet slip coefficient, and environmental temperature are collected; the rainfall is classified as dry, light rain, moderate rain, and heavy rain; the visibility is classified as good, light haze, moderate haze, and heavy haze; the road wet slip coefficient is classified as ice and snow, rainy day, and dry road; and the environmental temperature covers the full working interval from severe cold to high temperature. Based on the environmental parameters, the safety distance is dynamically adjusted, and the deceleration curve conforming to the road grip limit is re-planned to balance safety and driving stability. The safety distance is composed of reaction distance and braking distance; the reaction distance is positively correlated with the current vehicle speed and negatively correlated with the visibility; the braking distance is positively correlated with the square of the current vehicle speed and negatively correlated with the road wet slip coefficient; the deceleration curve rule is based on uniform deceleration, and the maximum deceleration does not exceed the physical limit corresponding to the road wet slip coefficient to ensure that the braking process does not slip. The environmental temperature affects the engine combustion efficiency, and through the temperature-related correction factor, the expected emission output of the carbon emission digital twin model is calibrated to improve the prediction accuracy. In the temperature interval where the engine combustion efficiency is optimal, the correction factor takes the reference value; when the temperature deviates from this interval, the correction factor increases with the degree of deviation; in low-temperature and high-temperature environments, the correction factor is further increased to match the actual emission characteristics; the expected carbon emission after calibration is obtained by multiplying the correction factor and the original output of the model.

[0033] Further, the state acquisition unit not only acquires real-time state parameters for representing the vehicle driving behavior, but also acquires the actual carbon emission amount of the vehicle in the previous time period from the vehicle-mounted sensors At the same time, the model self-calibration unit inputs the actual driving behavior parameters occurring in the previous time period into the expected carbon emission prediction unit, so that it calculates a predicted carbon emission amount corresponding to the actual driving behavior based on the current model parameters Subsequently, the model self-calibration unit receives and two values and compares them to calculate a prediction error, for example .

[0034] As an optional embodiment, the method can further include: step 104: generating at least two candidate driving strategies based on the carbon emission digital twin model, the two candidate driving strategies including a first candidate driving strategy and a second candidate driving strategy;

[0035] In this step, after receiving the real-time state parameters of the vehicle provided by the state acquisition unit, at least two mutually distinguished candidate driving strategies are generated for a future preset time domain or space domain, each candidate driving strategy being a data set consisting of a series of control instructions or target states, such as a speed and acceleration sequence, which can be executed by the vehicle.

[0036] Step 105: The first candidate driving strategy is optimized to minimize the impact during the acceleration and deceleration of the vehicle, and a smooth speed curve is obtained in combination with the deceleration curve.

[0037] In this step, for example, the first candidate driving strategy can be defined as a smooth and economic strategy, which is optimized to minimize the impact during the acceleration and deceleration of the vehicle or to minimize the output of a benchmark carbon emission model. In generating this strategy, the candidate driving strategy generation unit plans a speed curve with small absolute values of acceleration and deceleration, and the speed change process is continuous and smooth.

[0038] Step 106: The second candidate driving strategy is optimized to minimize the travel time through the front road section, and a minimum time speed curve is obtained.

[0039] In this step, the second candidate driving strategy can be defined as an efficient passing strategy, which is optimized to minimize the travel time through a specific road section, such as an intersection or a ramp merging area. In generating this strategy, the candidate driving strategy generation unit plans a speed curve with relatively large acceleration values to enable the vehicle to quickly reach the target speed or pass through the target area, while complying with the legal speed limit and vehicle physical performance constraints.

[0040] Step 20: For the candidate driving strategy, the corresponding expected carbon emissions are calculated by querying the emission factor mapping table of the vehicle specific power and integrating.

[0041] In this step, each candidate driving strategy k of the candidate driving strategy generation unit is constructed into a standardized data structure, such as a speed sequence indexed by time or an acceleration sequence , where t is within a preset time interval These data structures are then transmitted to the internal processing unit of the high-fidelity carbon emission digital twin for subsequent prediction of the expected carbon emission cost. The expected carbon emission prediction unit receives each candidate driving strategy and calculates a quantitative expected carbon emission cost for each strategy, which is based on a micro carbon emission model stored in the model parameter storage unit.

[0042] In one specific implementation, the microscopic carbon emission model is a model based on vehicle specific power or a similar physical quantity. The carbon emission prediction unit is expected to first target a given candidate driving strategy, for example, based on a speed sequence. and acceleration sequence Define and combine the real-time vehicle state parameters provided by the state acquisition unit to calculate the strategy at each moment during execution. instantaneous power demand Subsequently, the carbon emission prediction unit will calculate the instantaneous power demand. As input, an instantaneous carbon emission rate is queried or calculated using a microscopic carbon emission model. The model maps different power demand ranges to different emission rates. Finally, it maps the entire policy execution time interval. Instantaneous carbon emission rate By performing integration, the candidate driving strategy is obtained. Total expected carbon emission cost This calculation process can be performed by... limited.

[0043] For each candidate driving strategy, the expected carbon emission prediction unit executes the above calculation process once, generating a corresponding expected carbon emission cost value for each strategy. These data pairs, consisting of candidate driving strategies and their corresponding expected carbon emission costs, are sent to the bidding package generation module. For example, an optimization method based on gradient descent is used to fine-tune the model parameters according to the magnitude and direction of the prediction error, so that the error of subsequent predictions tends to decrease. The updated model parameters are written back to the model parameter storage unit for subsequent prediction of expected carbon emission costs for candidate driving strategies. Through the continuous execution of this process, the high-fidelity carbon emission digital twin model can adapt to changes in vehicle status, such as engine aging and tire wear, thereby maintaining its high prediction accuracy.

[0044] Step 30: Package the candidate driving strategies and their corresponding expected carbon emissions into a bidding package and send it to the edge computing node on the roadside.

[0045] In this step, all candidate driving strategies and their corresponding expected carbon emission costs are received. After receiving this data, it is organized and packaged into a standardized data structure, which is the bidding package. This package is used to submit a complete bid to the edge computing node, containing multiple options. A unique vehicle identifier is used by the edge computing node to identify the source vehicle of the bid. A bid set contains at least two elements, each being a data pair represented by data from a candidate driving strategy, such as a speed sequence. ) and its corresponding expected carbon emission cost value composition.

[0046] As an optional embodiment, the candidate driving strategy and the corresponding expected carbon emission are packaged as a bidding package and sent to the edge computing node of the roadside, which can specifically include: step 301: based on the edge computing node including upstream edge computing node and downstream edge computing node; step 302: after the upstream edge computing node completes the candidate driving strategy, the carbon emission inertia index and the average reputation score are obtained; step 303: the carbon emission inertia index and the average reputation score are packaged into a carbon wave protocol data packet and sent to the downstream edge computing node; step 304: the downstream edge computing node uses the carbon wave protocol data packet for feedforward calibration.

[0047] Step 40: based on the received bidding package, initiate a carbon right reverse auction, use a heuristic search algorithm to solve the combinatorial optimization problem, and determine the winning strategy set.

[0048] As an optional embodiment, based on the received bidding package, initiate a carbon right reverse auction, use a heuristic search algorithm to solve the combinatorial optimization problem, and determine the winning strategy set, which can specifically include:

[0049] Step 401: in the objective function of the optimization problem, combine the reputation weight coefficient with the expected carbon emission of each vehicle to reduce the winning probability;

[0050] In this step, when the carbon right auction house determines the winning strategy set, the objective function of the optimization problem to be solved can be defined by the following formula:

[0051] ;

[0052] Wherein: is the index of the vehicle, is the index of the candidate driving strategy, is the vehicle set in the region, is the total number of candidate driving strategies of vehicle ; is a binary decision variable, indicating whether to select the strategy of vehicle ; ; is the expected carbon emission cost of vehicle when executing the strategy ; is the dynamic reputation score of vehicle ; is a dynamic reputation-based reputation risk function, whose function value is a monotonically decreasing function of the dynamic reputation score, used to weight and adjust the expected carbon emission cost. This function makes the expected carbon emission cost of vehicles with lower dynamic reputation scores be given higher weights in the optimization solution.

[0053] Step 402: encode the selection of each vehicle for the candidate driving strategy as a gene, and the selection of all vehicles constitutes a chromosome; Step 403: take the total expected carbon emission of all vehicles in the region as the fitness function; Step 404: evolve the chromosome with the minimum fitness function value by iteratively performing selection, crossover and mutation operations, and the solution represented by the chromosome is the winning strategy set.

[0054] In this step, the objective function can be defined by the following formula:

[0055] ;

[0056] The optimization problem also needs to follow the constraint condition to ensure that each vehicle is finally assigned only one driving strategy.

[0057] is the unique index of the vehicle participating in this auction, is the set of all vehicles currently participating in the auction, is the index of a certain candidate driving strategy submitted by the vehicle , is the total number of candidate driving strategies submitted by the vehicle in its bid package, is the expected carbon emission cost submitted by the vehicle for its candidate driving strategy in its bid package, is the current dynamic reputation score of the vehicle , which is provided by the dynamic reputation and market access engine 22, is a binary decision variable. When its value is 1, it means that the candidate driving strategy of the vehicle wins in this auction; when its value is 0, it means that it does not win, is a reputation risk function used to risk-adjust the bid of the vehicle according to its dynamic reputation score.

[0058] The reputation risk function is a monotonically decreasing function of the dynamic reputation score . For example, the function can be specifically set as , where is a value normalized to the interval , ​It is a system parameter greater than zero, called the risk sensitivity coefficient. This function makes the dynamic credit score... The lower the vehicle's rating, the higher its corresponding credit risk function value. The larger the value, the greater the proportion of the expected carbon emission costs of low-credit vehicles in the objective function during the optimization process, thus reducing the likelihood of their inaccurate bids winning the bid.

[0059] Step 50: Generate a cooperative driving strategy based on the winning strategy set and distribute it to the corresponding vehicle-side nodes.

[0060] In this step, the carbon auction house 21 uses a built-in or external optimization solver to solve the integer programming problem described above, obtaining a set of optimal decision variable solutions that minimize the objective function value (i.e., all...). (The value of ). After the solution is completed, all The strategies with a value of 1 collectively constitute the "winning strategy set" for this auction. Based on this winning strategy set, the carbon credit auction house 21 generates corresponding cooperative driving strategies and distributes the winning driving strategies to the corresponding vehicle-end nodes 10 via the communication unit for execution. For vehicles that did not win a bid, a default or no-intervention driving command can be issued.

[0061] As an optional embodiment, after generating a cooperative driving strategy based on the winning strategy set and distributing it to the corresponding vehicle end node, the method may further include: Step 60: After the vehicle completes the cooperative driving strategy, obtain the actual carbon emissions of the vehicle; Step 70: Compare the actual carbon emissions with the expected carbon emissions and calculate the prediction error; Step 80: Based on the prediction error, update the vehicle's reputation score using a preset update rule, wherein the reputation score and reputation weight coefficient are subject to a preset fixed mapping rule.

[0062] In this step, the dynamic credit and market access engine audits the actual carbon emissions generated by the vehicle after it has executed the cooperative driving strategy issued by the carbon auction house. The engine compares the obtained actual carbon emission figures with the expected carbon emission costs corresponding to the vehicle's winning bid strategy and updates the vehicle's dynamic credit score based on the degree of consistency between the two. The updated dynamic credit score will be used for risk adjustment when the vehicle participates in the next carbon auction.

[0063] Specifically, the dynamic reputation and market access engine can update the dynamic reputation score using an exponential moving average method, the update rules of which are defined by the following formula:

[0064] ;

[0065] in: and vehicles Dynamic reputation score before and after the update; This is an updated weight for historical reputation, used to adjust the weighting of historical reputation and current performance score. This refers to the actual carbon emissions of the vehicle; Bidding strategy for vehicles The corresponding expected carbon emission costs; is an evaluation function whose function value is a monotonically decreasing function of the input variable (i.e., the normalized error between actual carbon emissions and expected carbon emission costs), used to map the prediction error of this driving behavior into a score of instantaneous performance.

[0066] As an optional embodiment, the method may further include: acquiring traffic flow data based on roadside sensing devices, using a spatiotemporal graph neural network model to predict the carbon emissions of each lane-level fine-grained area within a preset time period; and triggering a reverse carbon credit auction for a subset of vehicles entering the area when it is predicted that the carbon emissions of the target area will exceed a preset threshold after a preset time period.

[0067] In this step, the edge computing node, in addition to the carbon auction house and the dynamic reputation and market access engine, may also include a spatiotemporal carbon fluid model. This model aggregates driving intention information reported by all or some vehicle end nodes within its jurisdiction, including data such as vehicle destinations and navigation system-planned routes. Based on the aggregated driving intention information, the spatiotemporal carbon fluid model divides its managed road network into multiple spatiotemporal grids and predicts the traffic conditions of each spatiotemporal grid within a preset future time period. Subsequently, based on the predicted traffic conditions, the model uses a macroscopic traffic carbon emission model to estimate the carbon emission intensity per unit time that will be generated in each spatiotemporal grid. This intensity is defined as the carbon potential of that spatiotemporal grid. In this way, the model generates a spatiotemporal carbon potential distribution prediction map covering the jurisdiction and the preset future time.

[0068] Furthermore, a trigger condition for a carbon credit auction house to initiate a reverse carbon credit auction is determined by the prediction results of a spatiotemporal carbon fluid model. This model periodically performs its prediction function. When the prediction shows that the carbon potential at a certain road location at a future time will exceed a pre-set threshold, the system determines that there is a potential risk of carbon emission congestion at that spatiotemporal point, requiring proactive intervention. When this trigger condition is met, the spatiotemporal carbon fluid model sends a trigger signal to the carbon credit auction house to initiate a reverse carbon credit auction process, thereby proactively guiding and optimizing the emerging carbon emission hotspots. If the carbon potential at all spatiotemporal points within the predicted future period is below the threshold, no auction is triggered.

[0069] Furthermore, the edge computing node includes a carbon auction house connected to the dynamic reputation and market access engine, and is used to receive bidding packages sent by the communication units of one or more vehicle end nodes. Upon receiving a bidding package or a trigger signal sent by the spatiotemporal carbon fluid model, the carbon auction house initiates a reverse carbon auction with the aim of selecting a combination of all candidate driving strategies submitted by all participating vehicles that minimizes the total predicted carbon emission cost within the region while satisfying the basic constraints of traffic system operation.

[0070] As an optional embodiment, after acquiring traffic flow data based on roadside sensing devices, the method may further include: calculating a carbon emission fairness index within the region based on actual carbon emissions, whereby the fairness index measures the degree of difference in carbon emissions per unit mileage among different vehicles; allocating dual-objective weights for the reverse auction based on credit scores, wherein the first weight corresponds to the objective of minimizing the total expected carbon emissions in the region, and the second weight corresponds to the objective of optimizing the carbon emission fairness index; in the combinatorial optimization problem, incorporating the dual-objective weights into the objective function to construct a dual-objective optimization model; when generating candidate driving strategies through vehicle-end nodes, calculating the expected carbon emissions per unit mileage for each candidate driving strategy in parallel and encapsulating it into a bidding package for use by edge computing nodes when calculating the fairness index; after solving the dual-objective optimization model through edge computing nodes, when determining the winning strategy set, prioritizing solutions that meet the total carbon emission standard and whose emission differences per unit mileage are within a preset threshold; if multiple solutions exist, selecting the optimal solution based on vehicle credit scores.

[0071] In this step, the edge computing node also includes a dynamic reputation and market access engine, which is connected to the carbon credit auction house. After an auction clearing and strategy execution cycle ends, it audits the accuracy of the bids of participating vehicles and updates the dynamic reputation score of the vehicle based on the audit results, so as to influence the subsequent auction process.

[0072] The dynamic reputation and market access engine may include an audit data acquisition unit, a reputation assessment and update unit, and a market access control unit. The audit data acquisition unit obtains the vehicle's actual carbon emissions during the execution of its target cooperative driving strategy. The actual carbon emissions can be obtained through direct measurement using roadside remote sensing monitoring equipment, or by receiving reliable emissions data reported by vehicle-end nodes after completing the strategy execution. Simultaneously, the unit obtains the vehicle's winning bid strategy from the carbon credit auction house. The corresponding expected carbon emission cost .

[0073] The credit assessment and update unit is connected to the audit data acquisition unit to receive... and Two values, and based on the degree of consistency between the two, the vehicle... The dynamic credit score is updated, and this update process can employ an exponential moving average method, the update rule of which is defined by the following formula:

[0074] ;

[0075] in, It is a vehicle The updated dynamic credit score It is a vehicle Prior to this update, the dynamic reputation score... It is a historical reputation update weight with a value between 0 and 1, used to adjust the proportion of historical reputation and current performance scores in the synthesis of a new reputation score. It is a function that reflects instantaneous performance, and its purpose is to map the accuracy of a single quote to a standardized score. This function is monotonically decreasing, meaning that the smaller the prediction error, the higher the score.

[0076] Instantaneous behavior of the fractional function It can be specifically set as an exponential decay function:

[0077] ,in For normalization error , Let be the base of the natural logarithm. It is a decay coefficient greater than zero, used to control the sensitivity of the score to error.

[0078] The market access control unit is connected to the credit assessment and update unit, and receives the updated dynamic credit score. The market access eligibility of vehicles can be adjusted based on this score. For example, a market access credit threshold can be preset. .

[0079] When the market access control unit detects a vehicle's dynamic credit score Below this threshold At this time, a control command can be generated to temporarily suspend the vehicle's eligibility to participate in one or more subsequent auction cycles, and update the dynamic credit score. It will also be provided to carbon credit auction houses for risk adjustment of the vehicle's bid in the next auction.

[0080] Edge computing nodes may also include an inter-node collaboration protocol unit, which connects to the carbon auction house and the spatiotemporal carbon fluid model. The function of the inter-node collaboration protocol unit is to generate and send a carbon wave protocol data packet to the upstream node after the upstream node completes a round of carbon auctions for a traffic flow about to leave its jurisdiction and enter the jurisdiction of a downstream node of an adjacent edge computing node. This carbon wave protocol data packet contains a set of data describing the overall carbon emission characteristics of the departing traffic flow. This set of data may include: a carbon emission inertia index, characterizing the overall carbon emission level trend of the traffic flow as it leaves the upstream node's area. Specifically, this index can be calculated as the arithmetic mean of the expected carbon emission costs corresponding to the bidding strategies of all winning vehicles in the traffic flow; and an average reputation score, characterizing the overall credibility of the bids in the traffic flow. Specifically, this score can be calculated as the arithmetic mean of the dynamic reputation scores of all vehicles in the traffic flow. After receiving the carbon wave protocol data packet, the downstream node uses the data contained within it as input to its own spatiotemporal carbon fluid model. The spatiotemporal carbon fluid model of the downstream node uses the carbon emission inertia index and average credit score in the data packet to perform feedforward calibration on the prediction model to adjust the carbon potential prediction for the batch of traffic flows that are about to enter its jurisdiction.

[0081] This inter-node collaboration protocol enables information feedforward between adjacent edge computing nodes, allowing downstream nodes to anticipate the carbon emission characteristics of upcoming traffic flows. This improves the accuracy of their spatiotemporal carbon fluid model predictions and provides data support for more timely auction triggering and optimization decisions.

[0082] As another optional embodiment, this application can be applied to scenarios such as urban core areas and densely populated commercial areas, where multiple vehicle types mix, short-distance frequent start-stop, and carbon emissions are concentrated in time and space. First, the edge computing node classifies vehicles within its jurisdiction by fusing real-time onboard OBD data and roadside camera perception information: High-emission group: mainly including traditional fuel vehicles and heavy-duty pickup trucks, this group is allocated a relatively lower carbon budget base per unit mileage and a stricter credit risk sensitivity coefficient is set; Medium-emission group: mainly including ordinary fuel vehicles and plug-in hybrid vehicles, this group is allocated a medium carbon budget base and a standard credit risk adjustment strategy is adopted; Low-emission group: mainly including pure electric vehicles and non-motorized vehicles, this group is allocated a higher carbon budget base or an exemption mechanism is adopted. In other words, the carbon budget base is not a fixed value, but is dynamically adjusted according to real-time traffic density, road saturation, and historical carbon emission data for the same period. During morning and evening peak hours, the carbon budget for each group will be tightened overall to cope with the expected carbon emission peak.

[0083] Furthermore, to address the frequent starts and stops of vehicles caused by narrow roads and dense intersections, the vehicle-side nodes strengthen smoothness constraints when generating candidate driving strategies. In the candidate driving strategy generation algorithm, the penalty weight for the rate of change of acceleration is increased, prioritizing the generation of speed curves with smooth acceleration and deceleration to reduce instantaneous high emissions caused by sudden acceleration and sudden stops. For vehicle groups approaching traffic light intersections, the edge computing nodes consider coupling the strategies of multiple vehicles during auction optimization to generate a coordinated start-stop sequence for the convoy. This guides vehicles ahead to decelerate smoothly while suggesting that vehicles behind enter an idling coasting state in advance, avoiding a chain reaction of emissions surges caused by wave-like braking. The vehicle-side nodes combine the traffic light phase and timing information obtained from the edge nodes at the intersection ahead to determine whether to proceed in a green wave or predict stops, thereby reducing unnecessary stops and starts.

[0084] Furthermore, by utilizing a spatiotemporal graph neural network model and combining historical and real-time data, the system predicts the future carbon emission potential of each micro-road segment. When an emission hotspot is predicted to form in a certain area, the system will trigger the following coordinated mitigation measures in advance: push alternative route suggestions to vehicles about to enter the hotspot area via the navigation app; for vehicles confirmed to be entering the hotspot area, use a carbon credit auction mechanism to smooth out emission peaks by controlling the number of vehicles entering and their speed curves; and in the objective function of the carbon credit auction, set higher weights or green channels for buses and non-motorized vehicles to guide low-carbon travel modes to pass through first.

[0085] Furthermore, the auction optimization model not only minimizes total regional emissions but also introduces the Gini coefficient or intra-group emission variance per unit mile as auxiliary optimization objectives to limit excessive differentiation of carbon emission opportunities among different vehicles within the same group. For high-emission vehicles, if they consistently implement the winning bid strategy and make accurate predictions over the long term, they can be gradually rewarded with additional carbon budget flexibility or participate in carbon credit trading to exchange virtual quotas with low-emission vehicles, thus forming an internal incentive mechanism.

[0086] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a carbon flow optimization device for edge computing under vehicle-road cooperation, the structure of which is as follows: Figure 2 As shown.

[0087] Figure 2 This is a schematic diagram of the internal structure of a carbon flow optimization device for edge computing under vehicle-road cooperation, provided as an embodiment of this application. Figure 2 As shown, the device includes:

[0088] At least one processor 201;

[0089] And a memory 202 that is communicatively connected to at least one processor;

[0090] The memory 202 stores instructions that can be executed by at least one processor, which are executed by at least one processor 201 to enable at least one processor 201 to: perform any step of a carbon flow optimization method for edge computing under vehicle-road cooperation.

[0091] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium for carbon flow optimization in edge computing under vehicle-road cooperation is provided, which stores computer-executable instructions and is configured to execute any one of the steps of a carbon flow optimization method for edge computing under vehicle-road cooperation.

[0092] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0093] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0094] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0096] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0097] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0098] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0099] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0100] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0101] It should also be noted that 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 a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0102] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A carbon flow optimization method under vehicle-road cooperation edge computing, characterized in that, The method comprises: obtaining a real-time state parameter sequence of a vehicle and environmental information through a vehicle end node, generating at least two candidate driving strategies through a built-in carbon emission digital twin model; for the candidate driving strategies, calculating the corresponding expected carbon emissions by querying the emission factor mapping table of the specific power of the vehicle and integrating; encapsulating the candidate driving strategies and the corresponding expected carbon emissions into a bidding package, and sending it to an edge computing node on the roadside, specifically including: the edge computing node includes an upstream edge computing node and a downstream edge computing node; after the upstream edge computing node completes the candidate driving strategies, the carbon emission inertia index and the average credit score are obtained, wherein the carbon emission inertia index represents the carbon emission level trend of the vehicle flow when leaving the upstream node area, and the average credit score represents the credibility of the bid of the vehicle flow; encapsulating the carbon emission inertia index and the average credit score into a carbon wave protocol data packet and sending it to the downstream edge computing node; the downstream edge computing node uses the carbon wave protocol data packet for feedforward calibration; based on the received bidding package, initiating a carbon right reverse auction, using a heuristic search algorithm to solve the combinatorial optimization problem, and determining a winning strategy set, specifically including: In the objective function of the optimization problem, a reputation weight coefficient is combined with the expected carbon emission of each vehicle to reduce the winning probability, where the objective function is ; the constraint condition is ; wherein is the index of the vehicle, is the index of the candidate driving strategy, is the vehicle set in the region, is the total number of candidate driving strategies of the vehicle ; is a binary decision variable, indicating whether the strategy of the vehicle is selected ; is the expected carbon emission cost of the vehicle when executing the strategy ; is the dynamic reputation score of the vehicle ; is a reputation risk function based on the dynamic reputation score, and the function value is a monotone decreasing function of the dynamic reputation score, used to weight and adjust the expected carbon emission cost; encoding the selection of each vehicle on the candidate driving strategies as a gene, and the selection of all vehicles as a chromosome; taking the total expected carbon emissions of all vehicles in the region as the fitness function; evolve the chromosome with the minimum fitness function value by iteratively performing selection, crossover and mutation operations, and the solution represented by the chromosome is the winning strategy set; generating a cooperative driving strategy according to the winning strategy set and distributing it to the corresponding vehicle end node; after generating a cooperative driving strategy according to the winning strategy set and distributing it to the corresponding vehicle end node, the method further comprises: after the vehicle executes the cooperative driving strategy, obtaining the actual carbon emissions of the vehicle; comparing the actual carbon emissions with the expected carbon emissions to calculate the prediction error; based on the prediction error, updating the credit score of the vehicle using a preset update rule, and the credit score and the credit weight coefficient are pre-set fixed mapping rules; the method further comprises: obtaining vehicle flow data based on roadside perception devices, and predicting the carbon emissions of each lane-level fine-grained region within a preset time in the future using a spatio-temporal graph neural network model; when it is predicted that the carbon emissions of the target region will exceed the preset threshold after a preset time, triggering the carbon right reverse auction for a subset of vehicles entering the region.

2. The carbon flow optimization method of vehicle-road cooperation edge computing according to claim 1, wherein, obtaining a real-time state parameter sequence of a vehicle and environmental information through a vehicle end node, generating at least two candidate driving strategies through a built-in carbon emission digital twin model, specifically including: obtaining real-time environmental information through a vehicle end node, the environmental information including rainfall, visibility, road wetness coefficient and environmental temperature; increasing the predicted safety distance according to the rainfall, visibility and road wetness coefficient, and planning a deceleration curve according to the safety distance; According to the ambient temperature, a discharge correction factor is called to calibrate the output result of the carbon emission digital twin model.

3. The carbon flow optimization method of vehicle infrastructure integration edge computing according to claim 2, wherein, The method further comprises: generating at least two candidate driving strategies based on the carbon emission digital twin model, the two candidate driving strategies including a first candidate driving strategy and a second candidate driving strategy; The first candidate driving strategy optimizes the minimization of impact degree in the vehicle acceleration and deceleration process, and combines the deceleration curve to obtain a smooth speed curve. The second candidate driving strategy optimizes the minimization of driving time through the front road section to obtain a minimum time speed curve.

4. The carbon flow optimization method of vehicle-road cooperation edge computing according to claim 1, characterized in that, After the vehicle flow data is obtained based on the roadside sensing device, the method further comprises: combining the actual carbon emission amount to calculate a carbon emission fairness index in the region, the carbon emission fairness index being used to measure the difference degree of unit mileage carbon emission amount between different vehicles; According to the reputation score, the double-target weight of the reverse auction is allocated, wherein the first weight corresponds to the minimum total expected carbon emission amount of the region, and the second weight corresponds to the optimization target of the carbon emission fairness index; The double-target weight is integrated into a target function to construct a double-target optimization model; When the candidate driving strategies are generated through the vehicle end node, the unit mileage expected carbon emission amount corresponding to each candidate driving strategy is calculated in parallel, and is packaged into the bidding package; After the double-target optimization model is solved through the edge computing node, a solution is selected in which the total carbon emission amount meets the standard and the unit mileage emission difference is within a preset threshold.

5. A carbon flow optimization device under vehicle-road cooperation edge computing, characterized in that, The device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the carbon flow optimization method under vehicle-road cooperation edge computing according to any one of claims 1-4.

6. A non-volatile computer storage medium storing computer-executable instructions of carbon flow optimization of vehicle infrastructure integration edge computing, characterized in that, The computer executable instructions are configured to perform the steps of the carbon flow optimization method under vehicle-road cooperation edge computing according to any one of claims 1-4.

Images