An IVCPS confirmation method based on multiple-class scene multi-working conditions

Abstract

The application belongs to the technical field of intelligent vehicle CPS network physical system, and discloses an IVCPS confirmation method based on multiple scenes and working conditions, which constructs a unified demand ontology model of stakeholders' demand, application scene demand and function demand, realizes demand element deconstruction and collaborative mapping of multi-scale function, multi-level logic and multi-agent performance based on a knowledge graph, and introduces an adaptive weight model driven by working condition characteristics and a key working condition recognition mechanism enhanced by a large language model, so as to realize systematic quantitative evaluation and overall confirmation of the fusion of perception ability, collaborative decision-making ability and collaborative control ability of IVCPS under multiple scenes and working conditions. The application can solve the problems of the existing IVCPS verification method, such as multi-agent demand heterogeneity, huge scale of multi-scene and multi-working condition combination, and difficulty in unified confirmation of multi-scale and multi-level closed-loop logic, and is suitable for application scenes such as integrated verification of vehicle-road-cloud collaborative system, intelligent transportation infrastructure evaluation and engineering deployment of automatic driving system.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicle cyber-physical system technology, specifically involving an IVCPS system-level verification method for multiple operating scenarios and traffic conditions. This method achieves systematic quantitative evaluation and overall verification of the functions, logic, and performance of typical IVCPS reference system prototypes by constructing a multi-dimensional requirement ontology model, a cross-domain knowledge graph, and a multi-scale-multi-level-multi-agent collaborative mapping mechanism. It is applicable to application scenarios such as vehicle-road-cloud collaborative systems, intelligent transportation infrastructure, autonomous driving system integration verification, and intelligent transportation operation optimization. Background Technology

[0002] Existing intelligent connected vehicle cyber-physical systems (IVCPS) primarily focus on the vehicle, road, and cloud, and their integrated and collaborative operation. However, the following problems still exist in the actual system design and verification process:

[0003] (1) The sources of demand from multiple subjects and multiple dimensions are heterogeneous, and there is a lack of a unified confirmation modeling and mapping mechanism;

[0004] The requirements for IVCPS systems originate from multiple stakeholders, including direct participants, traffic management entities, technology providers, and derivative service providers. They cover various application scenarios such as urban roads, highways, and industrial park roads, and encompass multi-functional systems integrating perception, collaborative decision-making, and collaborative control. These requirements exhibit characteristics of multi-stakeholder, multi-dimensional, multi-level, and highly coupled nature. Existing methods often remain at the level of requirement listing or functional decomposition, lacking a systematic verification mechanism to unify stakeholder requirements, scenario requirements, and functional requirements into a computable requirement ontology, and further map them to multi-scale functions, multi-level logic, and multi-stakeholder performance. This makes it difficult to support the overall verification of typical IVCPS reference system prototypes.

[0005] (2) The combination of multiple scenarios and multiple working conditions is huge, and traditional verification methods are difficult to support system-level coverage confirmation;

[0006] In actual operation, IVCPS is affected by multiple factors such as road structure, traffic flow status, traffic control strategies, communication conditions, environmental disturbances, and traffic events. The system's functionality and performance vary significantly under different traffic scenarios and operating conditions, with the scale of operating condition combinations growing exponentially. Existing testing and verification methods mostly focus on single-vehicle functional testing, single-module simulation, or verification in a few typical scenarios. This makes it difficult to achieve systematic, quantitative, and comparable verification of the IVCPS's fusion perception performance, collaborative decision-making performance, and collaborative control performance under multiple scenarios and operating conditions. Furthermore, the lack of a differentiated modeling mechanism for operating condition weights leads to insufficient coverage and reliability of system verification results. Closed-loop logic verification is also challenging.

[0007] (3) The closed-loop logic of multi-scale, multi-level, and multi-subject is complex, making it difficult to confirm the overall system and evaluate its operational effectiveness;

[0008] Under vehicle-road-cloud collaborative operation conditions, the "perception-prediction-decision-control-coordination" closed-loop logic of IVCPS spans the road network scale, road segment scale, and vehicle-to-vehicle scale, involving collaborative mechanisms at the vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels, and is related to the functions and performance of multiple entities at the cloud, roadside, and vehicle levels. Existing methods struggle to construct a unified system-level conformity assessment model among multi-scale functions, multi-level logic, and multi-entity performance, resulting in a disconnect between IVCPS system-level validation and operational effectiveness evaluation. This makes it difficult to support the overall validation requirements of typical reference system prototypes under at least 16 intelligent vehicle design and operation domains (ODDs).

[0009] Therefore, there is an urgent need to propose a comprehensive verification method for IVCPS systems under multiple scenarios and operating conditions. This method would involve constructing a multi-dimensional verification requirement system, a requirement deconstruction and mapping mechanism based on knowledge graphs, a multi-condition collaborative decision-making and performance evaluation method enhanced by LLM, and a multi-scale, multi-level, and multi-agent operational performance conformity quantitative evaluation model. This would enable the systematic, verifiable, and quantifiable comprehensive verification of typical IVCPS reference system prototypes at the functional, logical, and performance levels, providing a methodological foundation and technical support for the engineering deployment and reliable operation of IVCPS. Summary of the Invention

[0010] In view of this, the purpose of this invention is to provide a holistic verification method for intelligent connected vehicle cyber-physical systems (IVCPS) based on multi-dimensional requirements, multi-scenario and multi-operating condition collaboration. By constructing a unified deconstruction mechanism of stakeholder requirements, application scenario requirements and functional requirements, a knowledge graph-driven requirement ontology and multi-scale functional / multi-level logic / multi-agent performance mapping model, and an LLM-enhanced multi-operating condition functional compliance quantitative assessment and weight adaptive mechanism, this method achieves systematic, verifiable and quantifiable holistic verification of typical IVCPS reference system prototypes under no less than 16 design operating domains (ODDs). This solves the problems of heterogeneous multi-agent requirement expression, huge scale of multi-scenario and multi-operating condition combinations, and difficulty in unified verification of multi-scale, multi-level, and multi-agent closed-loop logic in existing technologies.

[0011] The present invention solves the above problems by adopting the following technical solution:

[0012] An IVCPS validation method based on multiple scenarios and operating conditions includes the following steps:

[0013] S1. Construct a multi-dimensional requirement set for IVCPS based on stakeholder needs, application scenario needs, and functional needs, and form a requirement ontology model based on domain ontology and cross-domain knowledge graph;

[0014] S2. Based on the aforementioned demand ontology model, construct a mapping model between demand elements and multi-scale functional architecture, multi-level collaborative logical architecture, and multi-subject cyber-physical architecture;

[0015] S3. An adaptive weight model is constructed based on traffic condition characteristics to dynamically adjust the importance of different demand elements and corresponding functions, logic and performance indicators under different conditions; a large language model is introduced to identify key conditions and enhance test cases in multi-scenario and multi-condition test spaces.

[0016] S4. Based on the mapping model and adaptive weight model, conduct multi-scale functional compliance assessment, multi-level logical compliance assessment and multi-entity functional and performance compliance assessment on the prototype of the typical IVCPS reference system.

[0017] S5. Integrate the evaluation results to form an overall system compliance index, and determine whether the system passes the overall confirmation based on a preset confirmation threshold.

[0018] Furthermore, step S1 includes the following sub-steps:

[0019] S1.1 Starting from the needs of direct participants, governance stakeholders, technology suppliers, and derivative service providers, and combining the needs of urban road scenarios, highway scenarios, and other typical operational scenarios, as well as the needs of integrated perception functions, collaborative decision-making functions, and collaborative control functions, we construct an IVCPS original requirement set covering multiple subjects, multiple scenarios, and multiple functions.

[0020] S1.2 Deconstruction of requirements and semantic modeling based on domain ontology and AI-enhanced cross-domain knowledge graph;

[0021] By employing domain ontology knowledge and AI-enhanced cross-domain knowledge graph methods, the original set of requirements is semantically parsed, structurally expressed, and relationally modeled to form an IVCPS requirement ontology, requirement elements, and their relationship network, thereby achieving standardized representation of requirement elements and cross-domain semantic unification.

[0022] S1.3 Construct the IVCPS requirements ontology model;

[0023] Based on the results of the demand deconstruction, an IVCPS demand ontology model is formed, which includes demand categories, demand attributes, constraints and semantic relationships, providing a unified semantic foundation and formal expression carrier for the correlation reasoning and system-level confirmation between demand elements.

[0024] Furthermore, the specific content of step S2 is as follows:

[0025] I. Construct a linear mapping model from demand elements to functional indicators at the road network, road segment, and vehicle scales across multiple scale dimensions.

[0026] The demand set R is transformed into functional vectors at different scales, and their mapping relationship is expressed as follows:

[0027] F=∑w i ·f i (R)

[0028] In the formula, F is the function vector under multiple scale dimensions; f i (R) represents the i-th type of functional feature item extracted from the demand elements, w i The importance coefficients for corresponding functional characteristics are used to characterize the degree of contribution of different demand elements to the system's functional performance.

[0029] II. Construct a mapping model from requirements to the system's interactive logic structure at multiple levels;

[0030] The semantics of demand are mapped to a set of collaborative logic at the vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels. The mapping relationship is expressed as follows:

[0031] L=Φ(L vc ,L vr ,L vv )

[0032] In the formula, L represents the set of collaborative logic across multiple dimensions; L vc L vr L vv These represent the sets of interaction logic at the vehicle-cloud, vehicle-road, and vehicle-to-vehicle levels, respectively; Φ(·) is a cross-level logic fusion function used to describe the constraints of requirements on the system's operational logic under multi-level collaborative relationships;

[0033] III. Construct a subject performance aggregation model across multiple subject dimensions;

[0034] The demand elements are mapped to vectors of functional and performance indicators of the cloud, roadside, and vehicle-side entities. The mapping relationship is expressed as follows:

[0035] P=A·[P c ,P r ,P v ] T

[0036] In the formula, P represents a vector of functional and performance indicators under multiple subject dimensions; P c P r P v These represent the performance characteristic vectors of the cloud, roadside, and vehicle-side entities, respectively; A is the entity collaboration influence matrix, used to characterize the performance coupling relationship between different entities in the process of information interaction and physical collaboration.

[0037] Furthermore, the adaptive weight model in step S3 is based on a set of working condition features consisting of road structure elements, environmental elements, and traffic flow elements, and generates a dynamic weight vector through the eigenvalue method, entropy weight method, or a combination thereof.

[0038] Furthermore, the specific content of step S3 is as follows:

[0039] S3.1 Multi-scale scenario complexity assessment based on cyber-physical multi-dimensional weight adaptive matching;

[0040] I. Construction of the matter-element extension model;

[0041] The operating environment of the transportation system is divided into three categories: physical elements, information elements, and cyber-physical coupling elements, and corresponding matter-element extension models are constructed for each category. Let the set of matter-element features under scenario C be represented by X(C), and the matter-element modeling process be expressed as:

[0042] X(C) = {x1(C), x2(C), ..., x} n (C)}

[0043] In the formula, x i (C) represents the quantifiable attribute value of the i-th physical or information element in scenario C, which is used to form the basic feature space for subsequent complexity assessment and weight calculation.

[0044] II. Scene element feature extraction;

[0045] Based on the matter-element model, computable indicators are extracted from three dimensions: structural features, operational features, and interaction features, forming a multi-dimensional scene feature vector; let the comprehensive feature vector of scene C be F(C), then its expression is:

[0046] F(C)=[f1(C),f2(C),…,f m (C)]

[0047] In the formula, f i (C) represents the index value of scenario C on the i-th feature dimension;

[0048] III. Hierarchical entropy weight calculation;

[0049] First, a judgment matrix M(C) is constructed based on the eigenvectors, where the matrix elements represent the importance comparison relationship between different features under the current working condition. Then, the weight vector W is solved using the eigenvalue method, and its core calculation relationship is as follows:

[0050] M(C)·W=λmax·W

[0051] In the formula, λmax is the largest eigenvalue of matrix M(C), and W is the corresponding eigenvector. After normalization, it is used as the adaptive weight vector of scene features to reflect the relative importance of different elements under the current traffic conditions.

[0052] IV. Multi-scale scene complexity calculation;

[0053] After obtaining the weight vector W, corresponding feature vectors F1(C), F2(C), and F3(C) are constructed for the road network scale, road segment scale, and vehicle scale, respectively. These feature vectors are then weighted and aggregated to form a scene complexity index for each scale. The unified expression for this index is:

[0054] S k (C)=F k (C)·W

[0055] In the formula, k∈{1,2,3}, F k (C) represents the set of eigenvectors at scale k; S k (C) represents the scene complexity evaluation result at scale k;

[0056] V. Aggregation of multi-scale complexity;

[0057] The complexity results obtained at the road network scale, road segment scale, and vehicle scale are unified and fused to form an overall operational complexity description vector; let the multi-scale complexity fusion result be S(C), then its expression is:

[0058] S(C)=∑a k· S k (C)

[0059] In the formula, a k Let be the importance coefficient corresponding to scale k, and satisfy ∑ k α k =1; S(C) is the final comprehensive scenario complexity index, used to characterize the overall operational challenges faced by IVCPS under the current traffic conditions;

[0060] S3.2 Key Operating Condition Identification and Adaptive Weight Generation Based on LLM;

[0061] I. Construction of the evaluation indicator matrix;

[0062] M={m ij ∣r i ∈R,k j ∈K}

[0063] In the formula, M is the evaluation index matrix, used to quantify the correlation between demand elements and system capability indicators; r i k represents the demand factor. j This represents the capability indicator item, m.ij Represents demand element r i With ability indicator item k j The strength of the association or the weight of the influence between them;

[0064] II. Identification of critical operating conditions;

[0065] Use LLM intelligence to select the scenarios or use cases that best test the system and identify the most critical test conditions;

[0066] T ∗ =LLM(M,Ω(C j ))

[0067] In the formula, T ∗ Represents the set of critical test cases; Ω(C j ) indicates the current operating condition C j Multiscale complexity;

[0068] III. Adaptive weight generation;

[0069] Based on the difficulty of the current traffic scenario, the demand weights are dynamically adjusted to make the system evaluation closer to the actual working conditions.

[0070] W(C j )=Norm(W0⊙Γ(Ω(C j )))

[0071] In the formula, W(C) j ) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; Γ(Ω(C) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; Γ(Ω(C) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; W0 ... j )) represents the operating condition impact factor; ⊙ is the Hadamard product, which means that each weight is multiplied by the operating condition impact factor; Norm(·) represents the normalization operation, which makes all weights sum to 1;

[0072] IV. Iterative optimization;

[0073] Through iterative optimization, weights are determined and the system's evaluation under critical operating conditions is improved.

[0074] W(t+1)=W(t)+η·∇wL(W,T ∗ )

[0075] In the formula, W(t) represents the weight vector of the t-th iteration; η is the learning rate, which controls the adjustment magnitude in each iteration; L(W,T) ∗ ) represents the loss or bias function based on key test cases, which measures the difference between the system evaluation result under the current weight and the expected coverage of key operating conditions; ∇wL represents the gradient with respect to the weight, which is used to guide the optimization direction;

[0076] S3.3 Assessment of Integrated Perception and Collaborative Decision-Making Capabilities;

[0077] Based on the aforementioned adaptive weighting results, the system's fusion perception capability and collaborative decision-making capability under different traffic conditions are quantitatively evaluated. Let the current traffic condition be C, and the fusion perception capability score is obtained by weighted summation of each perception index, expressed as:

[0078] P(C) = ∑w i (C)·s i (C)

[0079] In the formula, P(C) represents the fusion perception capability score under operating condition C; s i (C) represents the test result of the i-th fusion sensing performance index under operating condition C; w i (C) represents the corresponding indicator s i (C) adaptive weights;

[0080] The collaborative decision-making capability score is obtained by weighted aggregation of various decision performance indicators, and its expression is:

[0081] D(C)=∑w k (C)·d k (C)

[0082] In the formula, D(C) represents the collaborative decision-making capability score under operating condition C; d k (C) represents the test result of the k-th collaborative decision-making performance index under operating condition C; w k (C) represents the corresponding indicator d. k (C) Adaptive weights.

[0083] Furthermore, the specific content of step S4 is as follows:

[0084] Multi-scale functional compliance assessment: This includes calculating the compliance between target values ​​and measured values ​​of functional indicators at the road network scale, road segment scale, and vehicle scale, and then performing weighted fusion.

[0085] Multi-level logical compliance assessment: This includes calculating the compliance between the expected execution results of logical rules and the actual execution results of the system at the vehicle-road-cloud level, the vehicle-road level, and the vehicle-to-vehicle level, and then performing weighted fusion.

[0086] Multi-entity functional and performance compliance assessment: This includes calculating the compliance between the target values ​​and measured values ​​of performance indicators for the cloud-based entity, the roadside entity, and the vehicle-side entity, and then performing weighted fusion.

[0087] Furthermore, the specific content of step S5 is as follows:

[0088] By unifying and integrating multi-scale functional compliance, multi-level logical compliance, and multi-entity performance compliance, an overall system performance confirmation index is formed:

[0089] S=α·F 总 +β·L 总 +γ·P 总

[0090] In the formula, S represents the overall conformity evaluation result of the IVCPS typical reference system prototype; F 总 Indicates the multi-scale functional compliance results; L 总 Indicates the result of multi-level logical compliance; P 总 This represents the result of multi-subject functional and performance compliance; α, β, and γ represent the fusion weights of the three types of evaluation results in the overall confirmation, and satisfy α+β+γ=1;

[0091] When S is greater than the preset confirmation threshold S0, it is determined that the IVCPS typical reference system prototype has passed the overall operation effect confirmation under the current multi-condition conditions; otherwise, it is fed back to step S3 to trigger a new round of parameter optimization and supplementary condition testing to achieve iterative convergence.

[0092] Furthermore, the method performs weighted fusion and iterative optimization on the overall system confirmation results based on no fewer than 16 types of design and operation domain scenarios.

[0093] Beneficial effects:

[0094] This invention provides an IVCPS verification method based on multiple scenarios and operating conditions. This method constructs a unified requirement ontology model encompassing stakeholder needs, application scenario needs, and functional requirements. Based on a knowledge graph, it achieves requirement element deconstruction and collaborative mapping of multi-scale functions, multi-level logic, and multi-agent performance. Furthermore, it introduces an adaptive weight model driven by operating condition features and a key operating condition identification mechanism enhanced by a large language model. This enables a systematic and quantitative evaluation and overall verification of the IVCPS's fusion perception capabilities, collaborative decision-making capabilities, and collaborative control capabilities under multiple scenarios and operating conditions. This invention solves the problems of heterogeneous multi-agent requirements, massive scale of multi-scenario and multi-operating condition combinations, and difficulty in uniformly verifying multi-scale and multi-level closed-loop logic in existing IVCPS verification methods. It is applicable to application scenarios such as vehicle-road-cloud collaborative system integration verification, intelligent transportation infrastructure evaluation, and engineering deployment of autonomous driving systems.

[0095] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0096] Figure 1 A flowchart for the overall validation methodology of IVCPS;

[0097] Figure 2 A flowchart for overall validation of a typical IVCPS reference system prototype;

[0098] Figure 3 A comprehensive framework diagram for confirming the IVCPS basic technologies was created.

[0099] Figure 4 Flowchart for confirming IVCPS basic technologies;

[0100] Figure 5 A method for overall validation of typical IVCPS reference system prototypes based on multiple operating conditions;

[0101] Figure 6 This document outlines the process for confirming the fundamental technologies of IVCPS across various scenarios. Detailed Implementation

[0102] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of this application.

[0103] like Figure 1 As shown, this invention provides an IVCPS verification method based on multiple scenarios and operating conditions, including the following steps:

[0104] S1. Confirm requirements based on multi-dimensional IVCPS;

[0105] S1.1 Overall validation of the typical reference system prototype for IVCPS, see Appendix for details. Figure 2 ;

[0106] S1.1.1 Construct an IVCPS requirement set based on the needs of stakeholders from multiple perspectives, application scenario requirements, and functional requirements;

[0107] Starting from the needs of stakeholders from multiple perspectives, including direct participants, governance stakeholders, technology suppliers, and derivative service providers, and combining the needs of urban road scenarios, highway scenarios, and other typical operational scenarios, as well as the needs of integrated perception functions, collaborative decision-making functions, and collaborative control functions, we construct an IVCPS original requirement set covering multiple subjects, multiple scenarios, and multiple functions.

[0108] S1.1.2 Determining requirements and semantic modeling based on domain ontology and AI-enhanced cross-domain knowledge graph;

[0109] By employing domain ontology knowledge and AI-enhanced cross-domain knowledge graph methods, the original set of requirements is semantically parsed, structurally expressed, and relationally modeled to form an IVCPS requirement ontology, requirement elements, and their relationship network, thereby achieving standardized representation of requirement elements and cross-domain semantic unification.

[0110] S1.1.3 Construct the IVCPS requirements ontology model;

[0111] Based on the results of the demand deconstruction, an IVCPS demand ontology model is formed, which includes demand categories, demand attributes, constraints and semantic relationships, providing a unified semantic foundation and formal expression carrier for the correlation reasoning and system-level confirmation between demand elements.

[0112] S1.1.4 Construct a multi-scale, multi-level, and multi-agent mapping model;

[0113] Based on the IVCPS demand ontology model, an interactive mapping model is constructed between demand elements and the IVCPS multi-scale functional architecture, multi-level logical architecture, and multi-agent cyber-physical architecture, realizing a computable mapping from demand elements to system functions, logic, and performance indicators.

[0114] S1.1.5 Conduct multi-scale functional compliance assessment, multi-level logical compliance assessment, and multi-entity functional and performance compliance assessment;

[0115] The system-in-the-loop testing method is used to obtain the operating data of the typical IVCPS reference system prototype under multiple operating conditions. The conformity quantitative assessment of its multi-scale functions, multi-level logic, and multi-entity functions and performance is carried out, and multi-scale functional conformity results, multi-level logic conformity results, and multi-entity functional and performance conformity results are generated respectively.

[0116] S1.1.6 Achieve overall validation of the typical IVCPS reference system prototype based on compliance results;

[0117] By integrating and analyzing the multi-scale functional compliance results, multi-level logical compliance results, and multi-subject functional and performance compliance results, the extent to which the IVCPS typical reference system prototype satisfies the IVCPS requirement ontology, requirement elements, and their relationships is determined, thereby achieving the overall confirmation of the IVCPS typical reference system prototype.

[0118] S1.2 Comprehensive confirmation of the basic technologies for building a typical IVCPS reference system prototype;

[0119] To address the issue that relying solely on single-technology verification in complex traffic environments makes it difficult to fully reflect the system-level applicability of IVCPS, this invention, based on a typical IVCPS reference system prototype, conducts systematic, multi-scenario, and multi-dimensional comprehensive verification of fundamental technologies such as architecture design, system modeling, system integration, and testing verification, in order to form a fundamental technology credibility support system for multiple design operation domains (ODDs).

[0120] According to the appendix Figure 3 With appendix Figure 4 The process shown involves conducting compliance assessments of operational performance under no fewer than 16 typical traffic scenarios. This process unifies the applicability, stability, and scalability of different basic technologies under multiple operating conditions, thereby transforming basic technologies from single-point verification to system-level validation.

[0121] By considering the different hierarchical roles of various foundational technologies within the system, an interactive mapping mechanism is constructed, oriented towards dimensions such as functionality, logic, and performance. This mechanism provides differentiated verification support for architecture, modeling, integration, and testing technologies, such as... Figure 6 As shown, this further supports the design and operation domain characterization of IVCPS under different road conditions, traffic environments, and cloud service conditions.

[0122] In summary, this invention achieves multi-scale, multi-level, and multi-subject comprehensive verification of IVCPS basic technologies through multi-dimensional conformity assessment of typical reference system prototypes and no less than 16 types of scenarios, and provides a unified technical framework for subsequent comprehensive verification schemes (S3) based on multiple scenarios of basic technologies.

[0123] S2. IVCPS overall verification method based on multiple operating conditions;

[0124] In response to the characteristics of IVCPS in actual operation, such as diverse traffic participants, complex operating condition combinations, highly nonlinear system states, and the emergence of random factors, system test results under a single operating condition or static scenario are insufficient to accurately reflect its overall capability level in a real traffic system. Therefore, this invention proposes a multi-operating condition-based overall IVCPS verification method. By constructing a multi-operating condition coverage mechanism and an adaptive weight mapping model, the system requirement compliance verification is expanded from "point-based testing" to a "systematic verification process involving multiple scenarios, dimensions, levels, and stakeholders," achieving reliable verification of the overall operational effectiveness of IVCPS. This method includes the following steps:

[0125] S2.1 Deconstruction and multidimensional mapping modeling of demand elements based on knowledge graphs;

[0126] This step addresses the challenges of diverse demand sources, complex semantic expressions, and strong coupling among multiple stakeholders in IVCPS systems. It proposes a knowledge graph-based method for demand element deconstruction and multi-dimensional mapping modeling. By integrating stakeholder goals, operational scenario constraints, and system functional capabilities through cross-perspective modeling, implicit demand semantics are transformed into structured demand elements and their interconnected networks. This achieves unified expression and standardized modeling of demand information, providing fundamental support for indicator construction and compliance assessment during subsequent system validation.

[0127] Building upon this, a multi-dimensional mapping mechanism is introduced to map the set of demand elements into multi-scale functional indicators, multi-level interaction logic indicators, and multi-subject functional and performance indicators. First, in the multi-scale dimension, a linear mapping model from demand to functional indicators is constructed to transform the demand set R into functional vectors at different scales. The mapping relationship is expressed as follows:

[0128] F=∑w i ·f i (R)

[0129] Among them, f i (R) represents the i-th type of functional feature item extracted from the demand elements, w i The importance coefficient corresponds to the functional characteristics and is used to characterize the degree of contribution of different demand elements to the system's functional performance.

[0130] At multiple levels, a mapping model is constructed from requirements to the system's interactive logic structure. This mapping transforms requirement semantics into a set of collaborative logic at the vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels. The mapping relationship is expressed as follows:

[0131] L=Φ(L vc ,L vr ,L vv )

[0132] Among them, L vc L vr L vv These represent the sets of vehicle-cloud, vehicle-road, and vehicle-to-vehicle interaction logic, respectively. Φ(·) is a cross-level logic fusion function used to describe the constraints of requirements on the system's operational logic under multi-level collaborative relationships.

[0133] In a multi-entity dimension, by constructing an entity performance aggregation model, demand elements are mapped to functional and performance indicator vectors of entities in the cloud, roadside, and vehicle. The mapping relationship is expressed as follows:

[0134] P=A·[P c ,P r ,P v ] T

[0135] Among them, Pc P r P v These represent the performance characteristic vectors of the cloud, roadside, and vehicle-side entities, respectively. A is the entity collaboration influence matrix, used to characterize the performance coupling relationship between different entities in the process of information interaction and physical collaboration.

[0136] Furthermore, to support the dynamic construction of evaluation weights under multiple operating conditions, the traffic operation environment is abstracted into a set of operating condition characteristics consisting of road elements, environmental elements, and traffic flow elements:

[0137] C={c r ,c e ,c t}

[0138] Based on this set of working condition features, a dynamic judgment matrix M(C) is constructed, and the weight vector W under the current working condition is solved using the eigenvalue method. The core relationship is expressed as follows:

[0139] M(C)·W=λmax·W

[0140] Where λmax is the largest eigenvalue of matrix M(C), and W is the corresponding eigenvector, which is normalized and used as the evaluation weight benchmark under the current working condition.

[0141] Furthermore, a working condition correction mechanism is introduced to dynamically adjust the initial weights, the expression of which is:

[0142] W(C)=Norm(W0⊙G(C))

[0143] Where W0 is the initial weight vector determined based on the demand element structure, G(C) is the working condition influence factor function, used to characterize the correlation strength between working condition characteristics and demand structure, ⊙ represents the element-wise product operator, and Norm(·) is the normalization function.

[0144] Through the above-mentioned knowledge graph-based demand element deconstruction and multi-dimensional mapping modeling method, the unified mapping of IVCPS requirements from semantic space to multi-scale functional space, multi-level logical space and multi-subject performance space was realized. An adaptive evaluation weight foundation for multi-condition operation was also constructed, providing a computable, scalable and evolvable indicator input interface for subsequent collaborative decision-making modeling and overall confirmation of system operation effect.

[0145] S2.2 A multi-dimensional, multi-condition collaborative decision-making mechanism based on LLM enhancement;

[0146] To address the challenges of multi-source heterogeneous information fusion, rapid changes in operating conditions, and highly coupled collaborative decision-making logic faced by IVCPS in complex traffic environments, this invention proposes a multi-dimensional, multi-condition collaborative decision-making mechanism based on LLM enhancement. This mechanism constructs a unified cyber-physical modeling framework to characterize the complexity of the system's operating environment under different traffic conditions. Combined with LLM-driven key condition generation and adaptive weight adjustment, it achieves dynamic evaluation of the IVCPS's fusion perception performance and collaborative decision-making performance. This method includes two parallel algorithm branches:

[0147] (I) Multi-scale scene complexity evaluation branch based on cyber-physical multi-dimensional weight adaptive matching

[0148] (1) Construction of the matter-element extension model;

[0149] First, the traffic system operating environment is divided into three categories: physical elements, information elements, and cyber-physical coupling elements. Corresponding matter-element extension models are then constructed for each category to uniformly express key attributes such as road structure, traffic flow state, communication quality, and collaborative behavior. Let X(C) represent the set of matter-element features in scenario C. Then, the matter-element modeling process can be uniformly represented as:

[0150] X(C) = {x1(C), x2(C), ..., x} n (C)}

[0151] Where, x i (C) represents the quantifiable attribute value of the i-th physical or information element in scenario C, which is used to form the basic feature space for subsequent complexity assessment and weight calculation.

[0152] (2) Scene element feature extraction;

[0153] Based on the matter-element model, computable indicators are extracted from three dimensions: structural features, operational features, and interaction features, forming a multi-dimensional scene feature vector to characterize the differences in system operation under different traffic conditions. Let F(C) be the comprehensive feature vector of scene C, then its expression is:

[0154] F(C)=[f1(C),f2(C),…,f m (C)]

[0155] Among them, f i (C) represents the index value of scenario C on the i-th feature dimension, which is used to reflect information such as environmental structural complexity, traffic operation uncertainty and the strength of system cooperative behavior.

[0156] (3) Hierarchical entropy weight calculation;

[0157] To avoid subjective bias caused by human-imposed weights, a hierarchical entropy weighting method is introduced to adaptively calculate the importance of each feature dimension. First, a judgment matrix M(C) is constructed based on the feature vectors, where the matrix elements represent the importance comparison relationship between different features under the current working condition. Then, the weight vector W is solved using the eigenvalue method, with the core calculation relationship being:

[0158] M(C)·W=λmax·W

[0159] Where M(C) represents the dynamic judgment matrix composed of scene features, λmax is its largest eigenvalue, and W is the corresponding eigenvector. After normalization, it serves as the adaptive weight vector of scene features, which is used to reflect the relative importance of different elements under the current traffic conditions.

[0160] (4) Multi-scale scene complexity calculation;

[0161] After obtaining the weight vector W, corresponding feature vectors F1(C), F2(C), and F3(C) are constructed for the road network scale, road segment scale, and vehicle scale, respectively. These feature vectors are then weighted and aggregated to form a scene complexity index for each scale. The unified expression for this index is:

[0162] S k (C)=F k (C)·W

[0163] Where k∈{1,2,3}, F k (C) represents the set of feature vectors at scale k, W is the weight vector, and S k (C) indicates the scene complexity assessment result at this scale.

[0164] (5) Aggregate multi-scale complexity

[0165] Finally, the complexity results obtained at the road network, road segment, and vehicle scales are unified and fused to form an overall operational complexity description vector, which supports subsequent collaborative decision-making weight adjustment and test case enhancement. Let the multi-scale complexity fusion result be S(C), then its expression is:

[0166] S(C)=∑a k· S k (C)

[0167] Among them, S k (C) represents the complexity result at the k-th scale, a k Let be the importance coefficient for the corresponding scale, and satisfy ∑ k α k =1, S(C) is the final comprehensive scenario complexity index, used to characterize the overall operational challenges faced by IVCPS under the current traffic conditions.

[0168] (ii) Key operating condition identification and weighted adaptive branch generation based on LLM;

[0169] (1) Construction of the evaluation index matrix:

[0170] M={m ij ∣r i ∈R,k j ∈K}

[0171] Where, r i k represents the demand factor. j This represents the capability indicator item, m. ij M represents the strength of the association or the weight of the influence. It is a matrix used to quantify the strength of the relationship between each requirement element and the system capability index, providing basic data for the subsequent LLM screening of key test cases.

[0172] (2) Identification of critical operating conditions;

[0173] Use LLM to intelligently select the scenarios or use cases that best test the system and identify the most critical test conditions.

[0174] T ∗ =LLM(M,Ω(C j ))

[0175] T ∗ Ω(C) represents the set of critical test cases. j ) indicates the current operating condition C j Multi-scale complexity, the LLM function is based on the evaluation index matrix M and the complexity Ω(C j It can automatically identify high-risk working conditions, extreme scenarios, or boundary conditions.

[0176] (3) Adaptive weight generation;

[0177] Based on the difficulty of the current traffic scenario, the demand weights are dynamically adjusted to make the system evaluation more closely reflect actual working conditions.

[0178] W(C j )=Norm(W0⊙Γ(Ω(C j )))

[0179] W0 represents the initial weight vector, derived from the initial importance of the demand elements, Γ(Ω(C j )) represents the operating condition influence factor, which describes the degree to which the current scenario complexity adjusts the weights. ⊙ is the Hadamard product (element-wise multiplication), which means that each weight is multiplied by the operating condition influence factor. Norm(·) represents the normalization operation, which makes all weights sum to 1.

[0180] (4) Iterative optimization;

[0181] Through iterative optimization and determination of weights, the system's assessment under critical operating conditions becomes more accurate and comprehensive.

[0182] W(t+1)=W(t)+η·∇wL(W,T ∗ )

[0183] Where W(t) represents the weight vector in the t-th iteration; η is the learning rate, controlling the adjustment magnitude in each iteration, and L(W,T) ∗ ) represents the loss or bias function based on key test cases, which measures the difference between the system evaluation result under the current weight and the expected coverage of key operating conditions. ∇wL represents the gradient with respect to the weight, which is used to guide the optimization direction.

[0184] This branch enhances test cases for high-risk operating conditions, extreme scenarios, and boundary conditions through LLM, enabling adaptive operating condition weights and metric optimization.

[0185] (III) Assessment of Integrated Perception and Collaborative Decision-Making Capabilities

[0186] Finally, based on the aforementioned adaptive weighting results, the system's fusion perception capability and collaborative decision-making capability under different traffic conditions are quantitatively evaluated. Let the current traffic condition be C, and the fusion perception capability score is obtained by weighted summation of each perception index, expressed as:

[0187] P(C) = ∑w i (C)·s i (C)

[0188] Where P(C) represents the fusion perception capability score, s i (C) represents the test result of the i-th fusion sensing performance index under operating condition C, w i (C) represents the adaptive weight of the corresponding indicator, which is used to reflect the importance of the indicator under the current working conditions.

[0189] The collaborative decision-making capability score is obtained by weighted aggregation of various decision performance indicators, and its expression is:

[0190] D(C)=∑w k (C)·d k (C)

[0191] Where, d k (C) represents the test result of the k-th collaborative decision-making performance index under operating condition C, w k (C) represents the adaptive weight of the corresponding indicator under this working condition.

[0192] Through the above mechanism, the system can achieve condition perception modeling, enhanced coverage of key test cases, and adaptive quantitative evaluation of weights for IVCPS to integrate perception and collaborative decision-making capabilities under multiple operating conditions, providing capability benchmarks and evaluation scales for confirming the overall operating effect of the S2.3 system.

[0193] S2.3 Multi-dimensional and multi-condition IVCPS typical reference system prototype operational performance compliance assessment;

[0194] Through multi-condition, multi-scale functional compliance assessment, multi-level logical compliance assessment, and multi-entity functional and performance compliance assessment, the overall validation of the IVCPS typical reference system prototype is ultimately achieved. Specifically, multi-scale functional validation includes road network, road segment, and peri-vehicle scales, corresponding to road network-scale functional compliance assessment, road segment-scale functional compliance assessment, and peri-vehicle-scale functional compliance assessment under multiple traffic conditions, respectively. Multi-level logical validation includes vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels, corresponding to vehicle-road-cloud, vehicle-road-level, and vehicle-to-vehicle level logical compliance assessments under multiple traffic conditions, respectively. Multi-entity functional and performance validation includes cloud-based functions and performance, road segment functions and performance, and vehicle-side functions and performance, corresponding to cloud-based functions and performance compliance assessment, road segment functions and performance compliance assessment, and vehicle-side functions and performance compliance assessment under multiple traffic conditions, respectively. See appendix for details. Figure 5 .

[0195] This step, based on the requirement element mapping model constructed in S2.1 and the working condition adaptive weights generated in S2.2, performs unified quantitative modeling and compliance judgment on the system's functional implementation capability, logical coordination behavior and main operating performance under real operating conditions, realizing a progressive closed-loop verification of the IVCPS typical reference system prototype from capability assessment to overall confirmation.

[0196] (1) Multi-scale functional compliance assessment method;

[0197] First, under different traffic conditions, the conformity assessment of the system's functional operation results is conducted at the road network scale, road segment scale, and surrounding vehicle scale. Let the target value of the i-th functional indicator at a certain scale be T. i The measured value is A i The single indicator compliance degree is defined as:

[0198] F i =T i / A i

[0199] Among them, F i A represents the degree of compliance with the i-th functional indicator. i T represents the actual operating result of the system.i This indicates the design target value corresponding to this indicator.

[0200] Furthermore, multiple functional indicators at the same scale are weighted and fused to obtain the overall functional compliance at that scale:

[0201] F=∑w i ·F i

[0202] Among them, w i Let F represent the importance weight of the i-th functional indicator, and let F represent the overall degree of compliance of the system functions under this scale.

[0203] Finally, the functional compliance at the road network scale, road segment scale, and surrounding vehicle scale are aggregated to obtain the multi-scale functional compliance evaluation results:

[0204] F 总 =(F net +F road +F veh ) / 3

[0205] Among them, F net F road and F veh These represent the functional compliance results at the road network scale, road segment scale, and vehicle perimeter scale, respectively.

[0206] (2) Multi-level logical compliance assessment

[0207] Based on functional compliance, the consistency of logical execution of the system during vehicle-road-cloud collaborative operation is evaluated and confirmed at three levels: vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle.

[0208] Let E be the expected execution result of the j-th logical rule at a certain level. j The actual execution result of the system is R. j Then its logical conformity is defined as:

[0209] L j =E j / R j

[0210] Among them, L j R represents the degree of compliance of the j-th logical rule. j E represents the actual execution result of the system. j This indicates the expected result of the design of the logical rule.

[0211] A comprehensive evaluation of multiple logical rules within the same level:

[0212] L=∑v j· L j

[0213] Among them, v j L represents the importance weight of the j-th logical rule, and L represents the overall logical compliance of this level.

[0214] Further integration of the logical consistency results of vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels:

[0215] L 总 =(L v2x +L v2r +L v2v ) / 3

[0216] Among them, L v2x、 L v2r and L v2v These represent the logical conformity of the vehicle-road-cloud level, the vehicle-road level, and the vehicle-to-vehicle level, respectively.

[0217] (3) Multi-entity functional and performance compliance assessment;

[0218] Further, a comprehensive evaluation of the system's operational functions and performance will be conducted from three aspects: cloud-based main body, roadside main body, and vehicle-side main body.

[0219] Let the target value of the k-th performance index under a certain subject be T. k The measured value is A k Then its performance compliance is defined as:

[0220] P k =A k / T k

[0221] Among them, P k A represents the degree of compliance with the k-th performance indicator. k T represents the measured performance result of the system. k This represents the corresponding design target value. Multiple performance indicators under the same subject are weighted and fused:

[0222] P=∑u k ·P k

[0223] Among them, u k represents the importance weight of the k-th performance indicator, and P represents the overall functional and performance compliance of the subject.

[0224] Further integrate the evaluation results from the cloud, road sections, and vehicle terminals:

[0225] P 总 =(P cloud +P road +P veh ) / 3

[0226] Among them, P cloud P road and P veh These respectively represent the functional and performance compliance of the cloud-based main body, the road segment main body, and the vehicle-side main body.

[0227] (4) Overall Convergence Confirmation and Judgment

[0228] By unifying and integrating multi-scale functional compliance, multi-level logical compliance, and multi-entity performance compliance, an overall system performance confirmation index is formed:

[0229] S=α·F 总 +β·L 总 +γ·P 总

[0230] Where S represents the overall conformity evaluation result of the IVCPS typical reference system prototype; F 总 Indicates the multi-scale functional compliance results; L 总 Indicates the result of multi-level logical compliance; P 总 This represents the result of multi-subject functional and performance compliance; α, β, and γ represent the fusion weights of the three types of evaluation results in the overall confirmation, and satisfy α+β+γ=1.

[0231] When S is greater than the preset confirmation threshold S0, it is determined that the IVCPS typical reference system prototype has passed the overall operation effect confirmation under the current multi-condition conditions; otherwise, it is fed back to the preceding evaluation and weight adjustment module to trigger a new round of parameter optimization and supplementary condition testing to achieve iterative convergence.

[0232] S3 Comprehensive Verification Solution for IVCPS Basic Technologies Based on Multiple Scenarios;

[0233] (a) Construct a reference system and acquire operational data for multiple scenarios;

[0234] Based on the IVCPS typical reference system prototype, the system operates under no fewer than 16 typical operating scenarios, collecting multi-source operating data such as perception, communication, decision-making, and control from the vehicle, road, and cloud ends, forming a scenario operating sample set covering multiple subjects, multiple levels, and multiple scales.

[0235] (ii) Establish a dynamic weight allocation model based on scenario criticality

[0236] Using a large language model, we analyze the semantics, operating conditions, and system challenge characteristics of each scenario. We then quantitatively evaluate each scenario based on dimensions such as failure severity, frequency of occurrence, system challenge, and scenario complexity, obtaining a scenario criticality score: S. i =aF i +bO i +cCi +dD i

[0237] Where Si is the critical score for the i-th scene, and F i O i C i D i These represent failure severity, occurrence frequency, system challenge, and scenario complexity, respectively, with a, b, c, da, b, c, da, b, c, and d being weighting coefficients.

[0238] All scenarios are normalized to obtain scenario weights, which are used for subsequent weighted fusion of results from multiple scenarios:

[0239] W i =S i / ∑S j

[0240] (iii) Construct a multi-dimensional compliance assessment model and verify the basic technologies;

[0241] Based on the above operational data, multi-dimensional compliance assessment models were constructed for the four fundamental technologies of IVCPS: architecture technology, modeling technology, refactoring and integration technology, and testing technology. These models were used to quantitatively confirm compliance from the perspectives of functional consistency, logical consistency, performance effectiveness, and coverage adequacy.

[0242] I. Architectural Technology Validation: Evaluate the consistency and effectiveness of multi-scale functional structures, multi-level logical structures, and multi-agent cyber-physical structures in multiple scenarios.

[0243] II. Modeling Technology Validation: Evaluate the logical consistency and accuracy of the modeling results of multi-level fusion perception and collaborative decision-making of vehicles, vehicles-roads, and vehicles-road-cloud in multiple scenarios.

[0244] III. Confirmation of Reconstruction and Integration Technology: Evaluate the compliance level of functional and performance co-integration after reconstruction of cloud, roadside, and vehicle-side systems in multiple scenarios.

[0245] IV. Testing Technology Validation: The effectiveness of testing methods in covering the functions and performance of multi-agent, multi-scale systems is evaluated across multiple scenarios. During this validation process, a unified compliance scoring function is constructed based on the characteristics of different underlying technologies, employing a hierarchical entropy weight-extension evaluation model, a collaborative relationship neural network model, and a hierarchical analysis model. Its general expression is as follows:

[0246] C i =f(X i )

[0247] Among them, C i X is the compliance score for a certain basic technology in the i-th scenario. iThis is the set of runtime data features corresponding to this scenario.

[0248] (iv) Comprehensive confirmation and judgment based on multi-scenario weight fusion

[0249] For a certain basic technology to be confirmed, its compliance score C in various scenarios is determined. i W by scenario weight i By performing weighted fusion, the overall confirmation value of this technology is obtained:

[0250] T=∑W i ·C i

[0251] When T≥θ (where θ is a preset confirmation threshold), the basic technology is determined to have passed confirmation; otherwise, it is determined to have failed confirmation.

[0252] (v) Develop basic technology verification capabilities for the design and operation domain

[0253] Through the aforementioned multi-scenario weighted compliance assessment mechanism, this invention can achieve unified, quantifiable, and comparable verification of IVCPS architecture design, system modeling, refactoring, integration, and testing verification technologies under no fewer than 16 typical design and operating domain conditions, providing a reliable basis for system deployment, security verification, and engineering applications.

[0254] It is hereby declared that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for IVCPS verification based on multiple scenarios and working conditions, characterized in that, Includes the following steps: S1. Construct a multi-dimensional requirement set for IVCPS based on stakeholder needs, application scenario needs, and functional needs, and form a requirement ontology model based on domain ontology and cross-domain knowledge graph; S2. Based on the aforementioned demand ontology model, construct a mapping model between demand elements and multi-scale functional architecture, multi-level collaborative logical architecture, and multi-subject cyber-physical architecture; S3. An adaptive weight model is constructed based on traffic condition characteristics to dynamically adjust the importance of different demand elements and corresponding functions, logic and performance indicators under different conditions; a large language model is introduced to identify key conditions and enhance test cases in multi-scenario and multi-condition test spaces. S4. Based on the mapping model and adaptive weight model, conduct multi-scale functional compliance assessment, multi-level logical compliance assessment and multi-entity functional and performance compliance assessment on the prototype of the typical IVCPS reference system. S5. Integrate the evaluation results to form an overall system compliance index, and determine whether the system passes the overall confirmation based on a preset confirmation threshold.

2. The IVCPS verification method based on multiple scenarios and working conditions according to claim 1, characterized in that, Step S1 includes the following sub-steps: S1.1 Starting from the needs of direct participants, governance stakeholders, technology suppliers, and derivative service providers, and combining the needs of urban road scenarios, highway scenarios, and other typical operational scenarios, as well as the needs of integrated perception functions, collaborative decision-making functions, and collaborative control functions, we construct an IVCPS original requirement set covering multiple subjects, multiple scenarios, and multiple functions. S1.2 Deconstruction of requirements and semantic modeling based on domain ontology and AI-enhanced cross-domain knowledge graph; By employing domain ontology knowledge and AI-enhanced cross-domain knowledge graph methods, the original set of requirements is semantically parsed, structurally expressed, and relationally modeled to form an IVCPS requirement ontology, requirement elements, and their relationship network, thereby achieving standardized representation of requirement elements and cross-domain semantic unification. S1.3 Construct the IVCPS requirements ontology model; Based on the results of the demand deconstruction, an IVCPS demand ontology model is formed, which includes demand categories, demand attributes, constraints and semantic relationships, providing a unified semantic foundation and formal expression carrier for the correlation reasoning and system-level confirmation between demand elements.

3. The IVCPS verification method based on multiple scenarios and working conditions according to claim 2, characterized in that: The specific content of step S2 is as follows: I. Construct a linear mapping model from demand elements to functional indicators at the road network, road segment, and vehicle scales across multiple scale dimensions. The demand set R is transformed into functional vectors at different scales, and their mapping relationship is expressed as follows: F=∑w i ·f i (R) In the formula, F is the function vector under multiple scale dimensions; f i (R) represents the i-th type of functional feature item extracted from the demand elements, w i The importance coefficients for corresponding functional characteristics are used to characterize the degree of contribution of different demand elements to the system's functional performance. II. Construct a mapping model from requirements to the system's interactive logic structure at multiple levels; The semantics of demand are mapped to a set of collaborative logic at the vehicle-road-cloud, vehicle-road, and vehicle-to-vehicle levels. The mapping relationship is expressed as follows: L=Φ(L vc ,L vr ,L vv ) In the formula, L represents the set of collaborative logic across multiple dimensions; L vc L vr L vv These represent the sets of interaction logic at the vehicle-cloud, vehicle-road, and vehicle-to-vehicle levels, respectively; Φ(·) is a cross-level logic fusion function used to describe the constraints of requirements on the system's operational logic under multi-level collaborative relationships; III. Construct a subject performance aggregation model across multiple subject dimensions; The demand elements are mapped to vectors of functional and performance indicators of the cloud, roadside, and vehicle-side entities. The mapping relationship is expressed as follows: P=A·[P c ,P r ,P v ] T In the formula, P represents a vector of functional and performance indicators under multiple subject dimensions; P c P r P v These represent the performance characteristic vectors of the cloud, roadside, and vehicle-side entities, respectively; A is the entity collaboration influence matrix, used to characterize the performance coupling relationship between different entities in the process of information interaction and physical collaboration.

4. The IVCPS verification method based on multiple scenarios and working conditions according to claim 3, characterized in that: The adaptive weight model in step S3 is based on a set of working condition features consisting of road structure elements, environmental elements, and traffic flow elements, and generates a dynamic weight vector through the eigenvalue method, entropy weight method, or a combination thereof.

5. The IVCPS verification method based on multiple scenarios and working conditions according to claim 4, characterized in that, The specific content of step S3 is as follows: S3.1 Multi-scale scenario complexity assessment based on cyber-physical multi-dimensional weight adaptive matching; I. Construction of the matter-element extension model; The operating environment of the transportation system is divided into three categories: physical elements, information elements, and cyber-physical coupling elements, and corresponding matter-element extension models are constructed for each category. Let the set of matter-element features under scenario C be represented by X(C), and the matter-element modeling process be expressed as: X(C)={x1(C),x2(C),…,x n (C)} In the formula, x i (C) represents the quantifiable attribute value of the i-th physical or information element in scenario C, which is used to form the basic feature space for subsequent complexity assessment and weight calculation. II. Scene element feature extraction; Based on the matter-element model, computable indicators are extracted from three dimensions: structural features, operational features, and interaction features, forming a multi-dimensional scene feature vector; let the comprehensive feature vector of scene C be F(C), then its expression is: F(C)=[f1(C),f2(C),…,f m (C)] In the formula, f i (C) represents the index value of scenario C on the i-th feature dimension; III. Hierarchical entropy weight calculation; First, a judgment matrix M(C) is constructed based on the eigenvectors, where the matrix elements represent the importance comparison relationship between different features under the current working condition. Then, the weight vector W is solved using the eigenvalue method, and its core calculation relationship is as follows: M(C)·W=λmax·W In the formula, λmax is the largest eigenvalue of matrix M(C), and W is the corresponding eigenvector. After normalization, it is used as the adaptive weight vector of scene features to reflect the relative importance of different elements under the current traffic conditions. IV. Multi-scale scene complexity calculation; After obtaining the weight vector W, corresponding feature vectors F1(C), F2(C), and F3(C) are constructed for the road network scale, road segment scale, and vehicle scale, respectively. These feature vectors are then weighted and aggregated to form a scene complexity index for each scale. The unified expression for this index is: S k (C)=F k (C)·W In the formula, k∈{1,2,3}, F k (C) represents the set of eigenvectors at scale k; S k (C) represents the scene complexity evaluation result at scale k; V. Aggregation of multi-scale complexity; The complexity results obtained at the road network scale, road segment scale, and vehicle scale are unified and fused to form an overall operational complexity description vector; let the multi-scale complexity fusion result be S(C), then its expression is: S(C)=∑a k· S k (C) In the formula, a k Let be the importance coefficient corresponding to scale k, and satisfy ∑ k α k =1; S(C) is the final comprehensive scenario complexity index, used to characterize the overall operational challenges faced by IVCPS under the current traffic conditions; S3.2 Key Operating Condition Identification and Adaptive Weight Generation Based on LLM; I. Construction of the evaluation indicator matrix; M={m ij ∣r i ∈R,k j ∈K} In the formula, M is the evaluation index matrix, used to quantify the correlation between demand elements and system capability indicators; r i k represents the demand factor. j This represents the capability indicator item, m. ij Represents demand element r i With ability indicator item k j The strength of the association or the weight of the influence between them; II. Identification of critical operating conditions; Use LLM intelligence to select the scenarios or use cases that best test the system and identify the most critical test conditions; T ∗ =LLM(M,Ω(C j )) In the formula, T ∗ Represents the set of critical test cases; Ω(C j ) indicates the current operating condition C j Multiscale complexity; III. Adaptive weight generation; Based on the difficulty of the current traffic scenario, the demand weights are dynamically adjusted to make the system evaluation closer to the actual working conditions. W(C j )=Norm(W0⊙Γ(Ω(C j ))) In the formula, W(C) j ) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; Γ(Ω(C) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; Γ(Ω(C) represents the adaptive weight vector after operating condition correction; W0 represents the initial weight vector; W0 ... j )) represents the operating condition impact factor; ⊙ is the Hadamard product, which means that each weight is multiplied by the operating condition impact factor; Norm(·) represents the normalization operation, which makes all weights sum to 1; IV. Iterative optimization; Through iterative optimization, weights are determined and the system's evaluation under critical operating conditions is improved. W(t+1)=W(t)+η·∇wL(W,T ∗ ) In the formula, W(t) represents the weight vector of the t-th iteration; η is the learning rate, which controls the adjustment magnitude in each iteration; L(W,T) ∗ ) represents the loss or bias function based on key test cases, which measures the difference between the system evaluation result under the current weight and the expected coverage of key operating conditions; ∇wL represents the gradient with respect to the weight, which is used to guide the optimization direction; S3.3 Assessment of Integrated Perception and Collaborative Decision-Making Capabilities; Based on the aforementioned adaptive weighting results, the system's fusion perception capability and collaborative decision-making capability under different traffic conditions are quantitatively evaluated. Let the current traffic condition be C, and the fusion perception capability score is obtained by weighted summation of each perception index, expressed as: P(C)=∑w i (C)·s i (C) In the formula, P(C) represents the fusion perception capability score under operating condition C; s i (C) represents the test result of the i-th fusion sensing performance index under operating condition C; w i (C) represents the corresponding indicator s i (C) adaptive weights; The collaborative decision-making capability score is obtained by weighted aggregation of various decision performance indicators, and its expression is: D(C)=∑w k (C)·d k (C) In the formula, D(C) represents the collaborative decision-making capability score under operating condition C; d k (C) represents the test result of the k-th collaborative decision-making performance index under operating condition C; w k (C) represents the corresponding indicator d. k (C) Adaptive weights.

6. The IVCPS verification method based on multiple scenarios and working conditions according to claim 5, characterized in that, The specific content of step S4 is as follows: Multi-scale functional compliance assessment: This includes calculating the compliance between target values ​​and measured values ​​of functional indicators at the road network scale, road segment scale, and vehicle scale, and then performing weighted fusion. Multi-level logical compliance assessment: This includes calculating the compliance between the expected execution results of logical rules and the actual execution results of the system at the vehicle-road-cloud level, the vehicle-road level, and the vehicle-to-vehicle level, and then performing weighted fusion. Multi-entity functional and performance compliance assessment: This includes calculating the compliance between the target values ​​and measured values ​​of performance indicators for the cloud-based entity, the roadside entity, and the vehicle-side entity, and then performing weighted fusion.

7. The IVCPS verification method based on multiple scenarios and working conditions according to claim 6, characterized in that, The specific content of step S5 is as follows: By unifying and integrating multi-scale functional compliance, multi-level logical compliance, and multi-entity performance compliance, an overall system performance confirmation index is formed: S=α·F 总 +β·L 总 +γ·P 总 In the formula, S represents the overall conformity evaluation result of the IVCPS typical reference system prototype; F 总 Indicates the multi-scale functional compliance results; L 总 Indicates the result of multi-level logical compliance; P 总 This indicates the results of multi-entity functional and performance compliance. α, β, and γ represent the fusion weights of the three evaluation results in the overall confirmation, and satisfy α+β+γ=1; When S is greater than the preset confirmation threshold S0, it is determined that the IVCPS typical reference system prototype has passed the overall operation effect confirmation under the current multi-condition conditions; otherwise, it is fed back to step S3 to trigger a new round of parameter optimization and supplementary condition testing to achieve iterative convergence.

8. The IVCPS verification method based on multiple scenarios and working conditions according to claim 7, characterized in that: The method performs weighted fusion and iterative optimization on the overall system confirmation results based on no fewer than 16 types of design and operation domain scenarios.

Images