Safety analysis method based on functional model machine

Abstract

The invention relates to the technical field of security analysis, in particular to a security analysis method based on a function prototype, and provides the following scheme: receiving function demand information and analyzing, and extracting functional elements of perception, decision, control and execution modules; constructing a virtual function prototype, and generating a sub-function module model; and setting operation parameters and parameter transmission constraint conditions. In a virtual simulation environment, a driving scene is constructed, each function module is tested, and potential system failure points and potential safety hazards are identified. And further performing structured processing on the analyzed function requirements, and establishing a dependency graph through graph theory analysis. Finally, the stability and safety of the system are evaluated by using multi-dimensional data analysis and a Monte Carlo simulation method, and a comprehensive safety evaluation report is generated to ensure the reliability of the automatic driving system.

Description

Technical Field

[0001] The present invention relates to the technical field of safety analysis, and in particular to a safety analysis method based on a functional prototype. Background Art

[0002] With the rapid development of autonomous driving technology, the functional requirements of the system are increasing, leading to higher requirements for interaction, data transmission and real-time performance between functional modules. Traditional safety analysis methods are often unable to effectively handle the complex dependencies between multifunctional modules and real-time data synchronization issues. In addition, existing technologies lack systematicity and comprehensiveness in evaluating system safety, and fail to fully consider dynamic changes and potential risks in different scenarios. Therefore, a safety analysis method based on functional prototypes is urgently needed to provide a more comprehensive and effective solution to ensure the safety and reliability of autonomous driving systems.

[0003] For example, the Chinese patent with the authorization announcement number CN102722598B discloses an aircraft engine non-contained failure safety analysis system and method, which belongs to the technical field of aircraft special risk analysis and assessment. The aircraft engine non-contained failure safety analysis system includes a demand information processing module, a parameter setting module, a simulation and result output module. The aircraft engine non-contained failure safety analysis method: determine the position parameters and size parameters of the engine rotor fragments relative to the aircraft digital prototype; within the reachable area of ​​the failure fragments, use the area division and hierarchical bounding box method to detect the aircraft equipment model that may cause failure due to the engine rotor fragments, and identify catastrophic functional hazards and quantitatively give the analysis results of the rotor non-contained failure safety by performing Boolean operations on the simulation result matrix and the minimum cut set. The invention quickly and accurately identifies the potential dangers of rotor non-contained failure during the aircraft design stage, providing technical support for aircraft system safety design and configuration design.

[0004] The above patents all have the problems raised by this background technology: lack of systematic multi-level analysis methods, failure to deeply identify potential failure points and performance bottlenecks, and limiting the overall evaluation of the safety of the autonomous driving system. To solve the above problems, this application designs a safety analysis method based on functional prototypes. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and provide a safety analysis method based on a functional prototype. The method receives and analyzes functional requirement information, extracts functional elements of perception, decision-making, control and execution modules; constructs a virtual functional prototype, generates a sub-functional module model; and sets operating parameters and parameter transfer constraints. In a virtual simulation environment, a driving scenario is constructed and each functional module is tested to identify potential system failure points and safety hazards. The parsed functional requirements are further structured, and a dependency graph is established through graph theory analysis. Finally, multidimensional data analysis and Monte Carlo simulation are used to evaluate the stability and safety of the system, and a comprehensive safety assessment report is generated to ensure the reliability of the autonomous driving system.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A safety analysis method based on a functional prototype, the method comprising:

[0008] Receiving functional requirement information of the autonomous driving system, parsing the functional requirement information, and extracting functional elements of a perception module, a decision module, a control module, and an execution module;

[0009] Construct a virtual functional prototype, perform functional decomposition according to the functional elements, and generate a sub-functional module model;

[0010] Setting the operating parameters of the sub-function module model and setting the parameter transfer constraints between modules;

[0011] In a virtual simulation environment, a driving scenario is constructed according to the virtual functional prototype, and each functional module is tested according to operating parameters and parameter transfer constraints;

[0012] Based on the test results, analyze the safety of the autonomous driving system and identify potential system failure points, performance bottlenecks and safety hazards.

[0013] Parsing the functional requirement information includes:

[0014] Extracting the functional requirement information and converting it into structured information, wherein the structured information includes function definition, input, output and dependency relationship;

[0015] The function definitions in the structured information are classified by cluster analysis technology, and the function definitions are decomposed into the function modules of perception, decision-making, control and execution;

[0016] Define the I / O interface of each functional module according to the input and output, and calculate the relationship mapping matrix according to the relationship between the I / O interfaces of each functional module;

[0017] The dependency relationship is processed by a dependency analysis method of graph theory to establish a dependency graph. The relationship mapping matrix and the dependency graph are subjected to feature extraction by a convolution operation to output functional elements. The calculation formula of the functional elements is:

[0018] F i (t) = σ[∑ j∈N(i) W×A ij ×X j (t-1)+b],

[0019] Among them, F i (t) represents the functional element of the i-th functional module at time step t in the relationship mapping matrix and dependency graph, σ[·] represents the ReLU activation function, N(i) represents the set of related functional modules of functional module i, j represents a single related functional module of functional module i, W represents the weight matrix of the convolution operation, and A ij represents the element corresponding to the jth function of functional module i in the relationship mapping matrix, X j (t-1) represents the input and output elements of the jth functional module at time step t-1, and b represents the bias term.

[0020] The generating sub-function module model comprises:

[0021] Designing the overall architecture of the virtual functional prototype according to the modular design principle and determining the components, wherein the components include virtual representations of a perception module, a decision module, a control module, and an execution module;

[0022] Map the functional elements extracted from the functional requirements analysis to the virtual functional prototype, and define the functional parameters, data format and interface protocol of each module by establishing the functional dictionary of the functional modules;

[0023] According to the functional dictionary, the functional dictionary is decomposed into sub-functional modules through a recursive decomposition algorithm, wherein, in the functional decomposition process, a hierarchical decomposition strategy is adopted to refine each module layer by layer into smaller sub-modules according to input and output conditions until each module can be represented by a clear operation or instruction.

[0024] The generating sub-function module model further includes:

[0025] Generate a virtual interface for each sub-function module model, wherein the virtual interface includes an input terminal and an output terminal for simulating communication between modules;

[0026] According to the recursive mapping optimization algorithm, the execution order and parameter transfer path of each sub-function module model are recursively optimized according to the functional logic of each sub-function module model;

[0027] Construct a dependency graph of the sub-functional module models, and calculate the degree of dependency between the sub-functional module models through the dependency matrix.

[0028] The operating parameters include:

[0029] The sampling rate, resolution, and noise level of the perception module;

[0030] The processing delay and data processing rate of the decision module;

[0031] Control module feedback frequency and system response time;

[0032] The operation delay and execution accuracy of the execution module.

[0033] The parameter transfer constraints include:

[0034] Data synchronization constraints are used to meet the real-time and synchronization of data transmission between functional modules. The time constraints of parameter transmission include the physical distance of modules and communication delay;

[0035] Delay constraints, where the delay is limited by the processing speed of the module and the network transmission rate;

[0036] Bandwidth constraint, where the bandwidth constraint is used to ensure that data transmission between modules does not exceed the maximum bandwidth in the case of concurrent data transmission;

[0037] The calculation formula for parameter transfer constraints is as follows:

[0038]

[0039] Among them, S k,m represents the data synchronization constraint between functional module k and functional module m, L k,m represents the delay constraint between functional module k and functional module m, B k,m represents the bandwidth constraint between functional module k and functional module m, T k,m represents the data transfer time between functional module k and functional module m, T sync Indicates the synchronization time threshold, D k,m represents the data packet size between function module k and function module m, C k,m represents the communication capability between functional module k and functional module m, Δ max Indicates the maximum delay, D k,m represents the bandwidth constraint between functional module k and functional module m, B max represents the maximum bandwidth, λ represents the transmission loss coefficient, d k,m represents the physical distance, τ k,m represents the additional network delay, Q represents the total number of transmissions, q represents a single transmission, and E qrepresents the processing delay of the qth transmission, G k,m (q) represents the available bandwidth between functional module k and functional module m, φ q represents the transfer delay of the qth transmission, t k,m Indicates the maximum data synchronization time.

[0040] The testing of each functional module according to the operating parameters and parameter transfer constraints includes:

[0041] Set the sampling rate, resolution, and noise level of the perception module, simulate the sensor working status in different scenarios, and obtain sensor data under various environmental conditions;

[0042] By inputting dynamic data streams, the data processing rate and processing delay of the decision module are obtained, and decision path planning is simulated in driving scenarios;

[0043] Setting the feedback frequency and system response time of the control module to test the execution capability of the command signal, wherein the execution capability includes the accuracy of the execution action and the hysteresis of the system's response to the control command;

[0044] The execution accuracy and operation delay of the execution module are set to test the execution efficiency of the physical instructions, which includes the execution time of controlling the direction, speed and acceleration of the vehicle.

[0045] The safety analysis of the autonomous driving system includes:

[0046] Performing a multi-level analysis on the test results through multi-dimensional data analysis technology to generate a comprehensive safety assessment report, wherein the multi-level analysis includes signal integrity analysis, failure mode impact analysis, and performance assessment;

[0047] In a virtual simulation environment, variable environmental parameters are set according to different driving scenarios, and a large number of possible system operation trajectories are generated through the Monte Carlo simulation method to evaluate the stability and safety of the system in various driving scenarios.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The present invention adopts multi-dimensional data analysis technology, which can deeply evaluate the stability and safety of the system in different driving scenarios, identify potential failure points and performance bottlenecks, and enhance the overall safety and reliability of the system;

[0050] 2. The present invention systematically analyzes the functional requirements of the autonomous driving system and provides efficient functional decomposition and modular design. It ensures the real-time and synchronization of data between functional modules by setting operating parameters and parameter transfer constraints, thereby improving the system's responsiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0052] Figure 1 This is a flow chart of a safety analysis method based on a functional prototype according to Embodiment 1 of the present invention;

[0053] Figure 2 This is a schematic diagram of the process of converting functional requirement information of embodiment 1 of the present invention;

[0054] Figure 3 This is a schematic diagram of the splitting of the perception module of embodiment 1 of the present invention. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0056] Example 1

[0057] See also Figure 1 The present invention provides an embodiment: a safety analysis method based on a functional prototype, the specific steps of the method are as follows:

[0058] S1: Receive and analyze the functional requirements of the autonomous driving system, and extract the key functional elements of the perception, decision-making, control and execution modules;

[0059] In this step, information is extracted from the functional requirements document to convert unstructured data into structured information, including functional definitions, input and output parameters, and dependencies between modules. Cluster analysis is used to classify functional elements, laying the foundation for subsequent functional decomposition.

[0060] S2: Build a virtual functional prototype, perform functional decomposition based on the extracted functional elements, and generate corresponding sub-functional module models;

[0061] In this step, the overall architecture of the virtual functional prototype is designed to ensure that it contains virtual representations of each module. Through the recursive decomposition algorithm, the functional elements are mapped to the corresponding modules, and a detailed description and functional dictionary of each sub-functional module are generated to clarify the functional parameters, data format and interface protocol of each module.

[0062] S3: Set operating parameters for each sub-functional module and define parameter transfer constraints between modules to ensure effective collaboration among modules;

[0063] In this step, according to the characteristics of the module, the operating parameters such as sampling rate, processing delay and feedback frequency of each sub-function module are set. At the same time, parameter transfer constraints between modules are established, including data synchronization constraints, delay constraints and bandwidth constraints, to ensure real-time performance and effectiveness.

[0064] S4: In a virtual simulation environment, design a variety of driving scenarios based on the constructed virtual functional prototype, and conduct systematic tests on each functional module based on operating parameters and constraints;

[0065] In this step, a variety of driving scenarios are created to simulate the sensor states and decision paths under various environmental conditions. Through dynamic data stream input, the performance indicators of each module, such as processing rate, response time, and execution accuracy, are recorded, and the interaction effect of the modules is evaluated.

[0066] S5: Analyze test results, evaluate the safety of the autonomous driving system, and identify potential system failure points, performance bottlenecks, and safety hazards;

[0067] In this step, multi-dimensional data analysis technology is used to comprehensively analyze the test results, identify key performance indicators and conduct failure mode impact analysis. A safety assessment report is generated, potential failure points and safety hazards are marked, and their impact on system performance is analyzed.

[0068] S6: Based on the analysis results, propose system optimization suggestions to improve the overall safety and reliability of the autonomous driving system;

[0069] In this step, based on the analysis results, an optimization strategy is developed, including adjusting the parameter settings between modules, improving the function implementation method and optimizing the algorithm. The suggestions are implemented and their effectiveness is verified through subsequent tests to ensure the stability and safety of the autonomous driving system in a changing environment.

[0070] The specific steps of S1 are as follows:

[0071] S1.1: extracting the functional requirement information and converting it into structured information, wherein the structured information includes function definition, input, output and dependency relationship;

[0072] See also Figure 2 , a schematic diagram of a process for converting functional requirement information according to an embodiment of the present invention, preprocessing the received functional requirement information to ensure that the format of the input information is consistent, including: removing redundant spaces, using unified terminology standards, and ensuring that each functional module in the requirement description uses the same naming convention;

[0073] Divide the demand information into independent functional description units. Use separators or rules to segment sentences so that each sentence describes an independent functional module or its submodule;

[0074] Identify the functional modules in the demand information through keyword matching. The system identifies the functional definitions in the text through rule matching or keywords based on the predefined functional module set, such as "perception module", "decision-making module", and "control module";

[0075] By matching the input signal type (such as "lidar data", "camera image data") and the output result (such as "obstacle position", "speed"), the input-output relationship of each functional module is extracted;

[0076] Identify the dependencies between functional modules by parsing the dependency keywords in the requirement information, such as "dependency", "input required", "based on", etc.

[0077] S1.2: Classify the function definitions in the structured information through cluster analysis technology and decompose the function definitions into functional modules of perception, decision-making, control and execution;

[0078] The extracted function definitions are classified by clustering algorithm, similar function definitions are classified into the same category, and the clustering results are decomposed into layers to determine the specific responsibilities and functions of each functional module, thus forming a preliminary model of perception, decision-making, control and execution modules;

[0079] The K-means clustering algorithm is used to classify the extracted function definitions. First, each function definition is converted into a vector representation, and the semantic feature vector of the function definition is generated using word embedding technology. Subsequently, the function definitions are classified through K-means clustering, and automatically identified and grouped into four main functional modules: "perception", "decision-making", "control", and "execution". Combining word embedding technology and K-means clustering algorithm, the function definitions are automatically processed and classified, reducing manual intervention and improving the efficiency and accuracy of classification;

[0080] S1.3: defining the I / O interface of each functional module according to the input and output, and calculating the relationship mapping matrix according to the relationship between the I / O interfaces of each functional module;

[0081] For each functional module, define its I / O interface. The input interface includes data type, data format, and signal source, and the output interface includes data type, expected result, and data destination. Establish a standard template for the I / O interface to ensure consistency;

[0082] Calculate the relationship mapping matrix, analyze the I / O interfaces between the functional modules, and establish the mapping relationship between the modules' data interactions. Each element of the relationship mapping matrix represents the data transfer relationship between two functional modules. By defining standardized I / O interfaces and using matrix operations to calculate the relationship mapping between modules, the accuracy and efficiency of data transfer can be ensured.

[0083] S1.4: The dependency relationship is processed by a dependency analysis method of graph theory to establish a dependency graph, and features are extracted from the relationship mapping matrix and the dependency graph by convolution operation to output functional elements. The calculation formula of the functional elements is:

[0084] F i (t) = σ[∑ j∈N(i) W×A ij ×X j (t-1)+b],

[0085] Among them, F i (t) represents the functional element of the i-th functional module at time step t in the relationship mapping matrix and dependency graph, σ[·] represents the ReLU activation function, N(i) represents the set of related functional modules of functional module i, j represents a single related functional module of functional module i, W represents the weight matrix of the convolution operation, and A ij represents the element corresponding to the jth function of functional module i in the relationship mapping matrix, X j (t-1) represents the input and output elements of the jth functional module at time step t-1, and b represents the bias term;

[0086] Based on the identified dependencies, a functional module dependency graph is constructed. Each node represents a functional module, and each edge represents the dependency relationship between modules. Use graph theory algorithms (such as Dijkstra algorithm) to analyze the dependency graph and determine the shortest path and critical path between modules;

[0087] Convolutional neural network (CNN) is used to extract features from the relationship mapping matrix and dependency graph, extract important features through convolution operation, and generate the final functional element representation;

[0088] Combining graph theory and deep learning technology, we can visualize and analyze complex dependencies, use CNN to extract features from high-dimensional data, and improve the expressiveness of functional elements.

[0089] The specific steps of S2 are as follows:

[0090] S2.1: Design the overall architecture of the virtual functional prototype according to the modular design principle and determine the components, wherein the components include virtual representations of a perception module, a decision module, a control module, and an execution module;

[0091] In the design phase, modular design principles are adopted to ensure the scalability and maintainability of the system. First, the framework of the overall architecture is determined, and appropriate modeling tools (such as UML) are selected for system design. The functions of each module are defined in detail to ensure clear interfaces and data flows;

[0092] The perception module is responsible for receiving and processing data from sensors, the decision module performs data analysis and judgment, the control module executes decisions and controls system behavior, and the execution module is responsible for physical execution of actions. The functions, input and output interfaces, and data flows of each module are clearly defined;

[0093] The modular design can achieve a highly cohesive and low-coupling system structure, which is convenient for later functional expansion and system maintenance, forming a clear system architecture that can effectively support the implementation of virtual functional prototypes and subsequent development and testing. For example, the perception module collects data through multiple sensors (such as cameras and lidar) and passes the processed results to the decision module.

[0094] S2.2: Map the functional elements extracted from the functional requirements analysis to the virtual functional prototype, and define the functional parameters, data format and interface protocol of each module by establishing a functional dictionary of the functional modules;

[0095] The extracted functional elements are systematically organized through the functional dictionary. The functional dictionary includes the functional description, input and output parameters, data format (such as JSON, XML, etc.) and interface protocol (such as RESTfulAPI, Socket communication, etc.) of each module. Use visualization tools to convert the functional dictionary into a table or graphic structure for subsequent reference. Through the mapping of functional elements, the consistency and traceability of the functions of each module are achieved, laying the foundation for subsequent system integration;

[0096] S2.3: According to the function dictionary, decompose the function dictionary into sub-function modules by a recursive decomposition algorithm, wherein in the function decomposition process, a hierarchical decomposition strategy is adopted to refine each module layer by layer into smaller sub-modules according to input and output conditions until each module can be represented by a clear operation or instruction;

[0097] The recursive decomposition algorithm is a recursive process based on a tree structure. It first identifies the high-level functional modules in the functional dictionary, and then decomposes them layer by layer. Each layer generates smaller sub-functional modules based on input and output conditions. For example, the perception module can be further decomposed into "lidar processing module", "camera image processing module", etc.

[0098] In the decomposition process, by setting thresholds (such as functional complexity, inter-module dependencies, etc.), the depth and degree of decomposition are judged to ensure that the decomposed modules have clear functions and operability, forming a clear sub-functional module hierarchy, making the functions of each module clear, which is conducive to subsequent development and testing;

[0099] See also Figure 3 , a schematic diagram of the decomposition of the perception module of an embodiment of the present invention, taking the "perception module" as an example, first decompose it into multiple sub-functional modules, including "lidar processing", "image processing" and "sensor fusion".

[0100] For the LiDAR Processing module, it is further broken down into:

[0101] Data preprocessing (denoising, filtering);

[0102] Obstacle detection (clustering algorithm, shape recognition);

[0103] For the "Image Processing" module, it is further decomposed into:

[0104] Image preprocessing:

[0105] Noise removal: Use image processing techniques (such as Gaussian filtering and bilateral filtering) to eliminate noise in the image.

[0106] Edge detection: Apply the Canny edge detection algorithm to extract important edge information in the image.

[0107] Target recognition includes: object detection, using deep learning models (such as YOLO, SSD) to identify and classify targets in images; feature extraction, using convolutional neural networks (CNN) to extract feature information of targets;

[0108] For "Sensor Fusion", it is further broken down into:

[0109] Data fusion algorithm:

[0110] Kalman filter: Fuse lidar and camera data to improve the estimation accuracy of target position.

[0111] Deep learning fusion: Use neural network models to integrate data from different sensors to optimize perception performance.

[0112] State Estimation:

[0113] Target tracking: Realize real-time tracking of dynamic targets and update their status information;

[0114] Each submodule specifies its input and output. For example, the input of the "obstacle detection" module is point cloud data, and the output is the coordinates and type of the detected obstacles.

[0115] S2.4: Generate a virtual interface for each sub-function module model, wherein the virtual interface includes an input terminal and an output terminal for simulating communication between modules;

[0116] Define a virtual interface for each sub-functional module, with the input end receiving data and the output end transmitting data. According to the requirements of the functional module, select a suitable communication protocol (such as HTTP, WebSocket, etc.) to implement the function of the virtual interface. Use simulation tools to generate interface documents, including interface name, request method, parameter description, return value and other information;

[0117] The communication between modules is realized through virtual interfaces, which improves the decoupling of the system and facilitates independent verification and integration of modules;

[0118] S2.5: According to the recursive mapping optimization algorithm, for the functional logic of each sub-functional module model, recursively optimize the execution order and parameter transfer path of the sub-functional module model;

[0119] Use recursive mapping optimization algorithms to analyze dependencies and data flows between modules, and use simulation tools to test the performance of different execution orders and parameter transfer paths. Use heuristic algorithms (such as genetic algorithms, simulated annealing, etc.) to optimize the execution order and select the optimal path to reduce system latency and resource consumption. The recursive mapping optimization algorithm dynamically adjusts the execution order between modules to improve the overall system efficiency, reduce bottlenecks, and achieve efficient module execution order and data transfer paths, thereby improving system performance and response speed.

[0120] For example, in the "perception module", "image processing" is performed first, followed by "obstacle detection" to obtain environmental information in the shortest possible time. In the execution sequence diagram, the order of module calls is optimized to ensure minimal delay. For example, the lidar data is processed first, followed by image data analysis, to improve the overall response speed;

[0121] S2.6: construct a dependency graph of the sub-function module models, and calculate the degree of dependency between the sub-function module models through the dependency matrix;

[0122] According to the input, output and dependency of each sub-function module, a dependency graph is constructed. Each node represents a sub-function module, and each edge represents the dependency between modules. The dependency matrix is ​​a two-dimensional matrix, and each element of the matrix represents the degree of dependency between two modules, represented by 0 (no dependency) and 1 (dependency);

[0123] The construction of dependency graphs and relationship matrices can intuitively display the interactions and dependencies between modules, facilitating system analysis and optimization. Through dependency graphs, the relationships and impacts between modules can be clarified, supporting subsequent system analysis and optimization decisions.

[0124] The processing delay and data processing rate of the decision module. The processing delay refers to the time required for the decision module to analyze the input data and make a decision, usually expressed in milliseconds (ms). The data processing rate refers to the amount of data that the decision module can process per second, usually expressed in packets / second.

[0125] The feedback frequency and system response time of the control module. The feedback frequency refers to the frequency at which the control module receives instructions from the decision module and executes feedback, usually expressed in Hertz (Hz). The system response time refers to the time required from receiving an instruction to executing the instruction, usually expressed in milliseconds (ms);

[0126] The operation delay and execution accuracy of the execution module. The operation delay refers to the delay time for the execution module to convert the control signal into the actual action, usually expressed in milliseconds (ms). The execution accuracy refers to the accuracy of the execution module when executing the operation, usually expressed in centimeters (cm);

[0127] The parameter transfer constraints include:

[0128] Data synchronization constraints are used to meet the real-time and synchronization of data transmission between functional modules. The time constraints of parameter transmission include the physical distance of modules and communication delay;

[0129] Each module adds a timestamp to the data during data collection to ensure that the data received by different modules can be accurately compared and synchronized. Synchronous clocks (such as the Network Time Protocol NTP) are used to ensure that all modules in the system have consistent time, avoid data conflicts caused by time asynchrony, and ensure that the information between the perception, decision-making and control modules is seamlessly connected in real time. For example, the perception module can promptly transmit obstacle information to the decision-making module, and the decision-making module can quickly formulate driving strategies to reduce misjudgments caused by data asynchrony, thereby improving vehicle safety;

[0130] When the amount of data increases or the environment becomes more complex, the real-time requirements become higher, and the system needs to be optimized accordingly during design to meet the stricter real-time requirements;

[0131] Delay constraints, where the delay is limited by the processing speed of the module and the network transmission rate;

[0132] By analyzing the processing speed of each functional module, for example, using high-performance computing chips (such as GPU) to speed up the processing speed of the decision-making module, regularly measuring and optimizing the network transmission rate, and ensuring low latency in high concurrency situations;

[0133] Realize fast response of information processing between modules. For example, the processing delay of obstacle information does not exceed 100ms, so as to quickly generate driving strategies, improve the real-time response capability of the system, reduce the operation delay of the system in complex environments, and ensure that the vehicle can avoid obstacles in time;

[0134] Bandwidth constraint, where the bandwidth constraint is used to ensure that data transmission between modules does not exceed the maximum bandwidth in the case of concurrent data transmission;

[0135] By analyzing the data requirements of each module, dynamically evaluating and allocating bandwidth, giving priority to bandwidth requirements of key functional modules, implementing adaptive flow control algorithms, adjusting data flows to adapt to changes in network bandwidth, and ensuring the stability and effectiveness of data transmission between modules under high concurrency conditions. For example, when multiple sensors transmit data at the same time, the system can effectively manage bandwidth and avoid data packet loss. This improves the overall data transmission efficiency of the system, allowing each functional module to operate stably under high load.

[0136] The calculation formula for parameter transfer constraints is as follows:

[0137]

[0138] Among them, S k,m represents the data synchronization constraint between functional module k and functional module m, L k,m represents the delay constraint between functional module k and functional module m, B k,m represents the bandwidth constraint between functional module k and functional module m, T k,m represents the data transfer time between functional module k and functional module m, T sync Indicates the synchronization time threshold, D k,m represents the data packet size between function module k and function module m, C k,m represents the communication capability between functional module k and functional module m, Δ max Indicates the maximum delay, D k,m represents the bandwidth constraint between functional module k and functional module m, B max represents the maximum bandwidth, λ represents the transmission loss coefficient, d k,m represents the physical distance, τ k,m represents the additional network delay, Q represents the total number of transmissions, q represents a single transmission, and E q represents the processing delay of the qth transmission, G k,m (q) represents the available bandwidth between functional module k and functional module m, φ q represents the transfer delay of the qth transmission, t k,m Indicates the maximum data synchronization time;

[0139] The testing of each functional module according to the operating parameters and parameter transfer constraints includes:

[0140] Set the sampling rate, resolution, and noise level of the perception module, simulate the sensor working status in different scenarios, and obtain sensor data under various environmental conditions;

[0141] By inputting dynamic data streams, the data processing rate and processing delay of the decision module are obtained, and decision path planning is simulated in driving scenarios;

[0142] Setting the feedback frequency and system response time of the control module to test the execution capability of the command signal, wherein the execution capability includes the accuracy of the execution action and the hysteresis of the system's response to the control command;

[0143] The execution accuracy and operation delay of the execution module are set to test the execution efficiency of the physical instructions, which includes the execution time of controlling the direction, speed and acceleration of the vehicle.

[0144] The safety analysis of the autonomous driving system includes:

[0145] Performing a multi-level analysis on the test results through multi-dimensional data analysis technology to generate a comprehensive safety assessment report, wherein the multi-level analysis includes signal integrity analysis, failure mode impact analysis, and performance assessment;

[0146] In a virtual simulation environment, variable environmental parameters are set according to different driving scenarios, and a large number of possible system operation trajectories are generated through the Monte Carlo simulation method to evaluate the stability and safety of the system in various driving scenarios.

[0147] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A safety analysis method based on functional prototype, characterized in that: The method comprises: Receiving functional requirement information of the autonomous driving system, parsing the functional requirement information, and extracting functional elements of a perception module, a decision module, a control module, and an execution module; Construct a virtual functional prototype, perform functional decomposition according to the functional elements, and generate a sub-functional module model; Setting the operating parameters of the sub-function module model and setting the parameter transfer constraints between modules; In a virtual simulation environment, a driving scenario is constructed according to the virtual functional prototype, and each functional module is tested according to operating parameters and parameter transfer constraints; Based on the test results, analyze the safety of the autonomous driving system and identify potential system failure points, performance bottlenecks and safety hazards.

2. A safety analysis method based on functional prototype according to claim 1, characterized in that: Parsing the functional requirement information includes: Extracting the functional requirement information and converting it into structured information, wherein the structured information includes function definition, input, output and dependency relationship; The function definitions in the structured information are classified by cluster analysis technology, and the function definitions are decomposed into the function modules of perception, decision-making, control and execution; Define the I / O interface of each functional module according to the input and output, and calculate the relationship mapping matrix according to the relationship between the I / O interfaces of each functional module; The dependency relationship is processed by a dependency analysis method of graph theory to establish a dependency graph. The relationship mapping matrix and the dependency graph are subjected to feature extraction by a convolution operation to output functional elements. The calculation formula of the functional elements is: F i (t)=σ[∑ j∈N(i) W×A ij ×X j (t-1)+b], Among them, F i (t) represents the functional element of the i-th functional module at time step t in the relationship mapping matrix and dependency graph, σ[·] represents the ReLU activation function, N(i) represents the set of related functional modules of functional module i, j represents a single related functional module of functional module i, W represents the weight matrix of the convolution operation, and A ij represents the element corresponding to the jth function of functional module i in the relationship mapping matrix, X j (t-1) represents the input and output elements of the jth functional module at time step t-1, and b represents the bias term.

3. A safety analysis method based on functional prototype according to claim 2, characterized in that: The generating sub-function module model comprises: Designing the overall architecture of the virtual functional prototype according to the modular design principle and determining the components, wherein the components include virtual representations of a perception module, a decision module, a control module, and an execution module; Map the functional elements extracted from the functional requirements analysis to the virtual functional prototype, and define the functional parameters, data format and interface protocol of each module by establishing the functional dictionary of the functional modules; According to the functional dictionary, the functional dictionary is decomposed into sub-functional modules through a recursive decomposition algorithm, wherein, in the functional decomposition process, a hierarchical decomposition strategy is adopted to refine each module layer by layer into smaller sub-modules according to input and output conditions until each module can be represented by a clear operation or instruction.

4. A safety analysis method based on functional prototype according to claim 3, characterized in that: The generating sub-function module model further includes: Generate a virtual interface for each sub-function module model, wherein the virtual interface includes an input terminal and an output terminal for simulating communication between modules; According to the recursive mapping optimization algorithm, the execution order and parameter transfer path of each sub-function module model are recursively optimized according to the functional logic of each sub-function module model; Construct a dependency graph of the sub-functional module models, and calculate the degree of dependency between the sub-functional module models through the dependency matrix.

5. The safety analysis method based on functional prototype according to claim 1, characterized in that: The operating parameters include: The sampling rate, resolution, and noise level of the perception module; The processing delay and data processing rate of the decision module; Control module feedback frequency and system response time; The operation delay and execution accuracy of the execution module.

6. A safety analysis method based on functional prototype according to claim 5, characterized in that: The parameter transfer constraints include: Data synchronization constraints are used to meet the real-time and synchronization of data transmission between functional modules. The time constraints of parameter transmission include the physical distance of modules and communication delay; Delay constraints, where the delay is limited by the processing speed of the module and the network transmission rate; Bandwidth constraint, where the bandwidth constraint is used to ensure that data transmission between modules does not exceed the maximum bandwidth in the case of concurrent data transmission; The calculation formula for parameter transfer constraints is as follows: Among them, S k,m represents the data synchronization constraint between functional module k and functional module m, L k,m represents the delay constraint between functional module k and functional module m, B k,m represents the bandwidth constraint between functional module k and functional module m, T k,m represents the data transfer time between functional module k and functional module m, T sync Indicates the synchronization time threshold, D k,m represents the data packet size between function module k and function module m, C k,m represents the communication capability between functional module k and functional module m, Δ max Indicates the maximum delay, D k,m represents the bandwidth constraint between functional module k and functional module m, B max represents the maximum bandwidth, λ represents the transmission loss coefficient, d k,m represents the physical distance, τ k,m represents the additional network delay, Q represents the total number of transmissions, q represents a single transmission, and E q represents the processing delay of the qth transmission, G k,m (q) represents the available bandwidth between functional module k and functional module m, φ q represents the transfer delay of the qth transmission, t k,m Indicates the maximum data synchronization time.

7. The safety analysis method based on functional prototype according to claim 1, characterized in that: The testing of each functional module according to the operating parameters and parameter transfer constraints includes: Set the sampling rate, resolution, and noise level of the perception module, simulate the sensor working status in different scenarios, and obtain sensor data under various environmental conditions; By inputting dynamic data streams, the data processing rate and processing delay of the decision module are obtained, and decision path planning is simulated in driving scenarios; Setting the feedback frequency and system response time of the control module to test the execution capability of the command signal, wherein the execution capability includes the accuracy of the execution action and the hysteresis of the system's response to the control command; The execution accuracy and operation delay of the execution module are set to test the execution efficiency of the physical instructions, which includes the execution time of controlling the direction, speed and acceleration of the vehicle.

8. The safety analysis method based on functional prototype according to claim 1, characterized in that: The safety analysis of the autonomous driving system includes: Performing a multi-level analysis on the test results through multi-dimensional data analysis technology to generate a comprehensive safety assessment report, wherein the multi-level analysis includes signal integrity analysis, failure mode impact analysis, and performance assessment; In a virtual simulation environment, variable environmental parameters are set according to different driving scenarios, and a large number of possible system operation trajectories are generated through the Monte Carlo simulation method to evaluate the stability and safety of the system in various driving scenarios.

Images