Unmanned vehicle reliability analysis method and device considering computing resource load influence

Abstract

This invention discloses a reliability analysis method and system for unmanned vehicles (UAVs) that considers the impact of computational resource load. The method establishes a shared computational resource load model for the perception, communication, computing, decision-making and control, and execution modules of an UAV system. It defines the total actual computational resource load, available computational resource capacity, and load rate of the system, and determines system failure criteria by combining resource overload criteria and partial load lifetime mapping criteria. The UAV system affected by computational resource load is treated as a two-state repairable system, establishing equivalent failure rate, equivalent repair rate, and state transition relationships. A GO method operator 2ML is designed for dual-module functional units, establishing its state description and instantaneous quantitative calculation formula. A GO method quantitative analysis algorithm considering computational load is constructed to recursively calculate reliability indicators such as system success probability, failure probability, and availability. This invention can be used for the quantitative analysis and evaluation of the operational reliability of UAV systems.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle system reliability analysis technology, specifically to an unmanned vehicle reliability analysis method and apparatus that considers the impact of computing resource load. Background Technology

[0002] Autonomous vehicle (RV) systems typically consist of a perception module, a communication module, a computing module, a decision-making and control module, and an execution module. The computing module is responsible for key tasks such as environmental perception information processing, target recognition, path planning, task decision-making, control calculation, and communication scheduling, and is a crucial part supporting the normal operation of the RV system. During task execution, RVs continuously call various algorithms to process external environmental information and internal system state information; therefore, their computing resources are always dynamically occupied. Common computing resources include processor processing power, storage resources, data transmission capacity, and task scheduling capabilities. The normal operation of a single functional module only reflects the partial operating state of the RV system and cannot fully characterize the reliability level of the entire RV system under complex task conditions. Specifically, when computing resources are sufficient, each functional module can complete its corresponding tasks within a specified time, and the system can maintain normal operation; however, when the computing resource load continuously increases, phenomena such as task processing delays, scheduling blockages, response lags, or partial functional degradation are likely to occur, thus affecting the task execution capability of the RV system and even leading to system failure.

[0003] When conducting reliability analysis on autonomous vehicle systems, modeling solely from the perspective of physical component failures such as sensors, actuators, communication hardware, or power units often fails to accurately reflect the actual operating state of the system. For example, considering only physical hardware component failures while ignoring changes in computational resource load, even without significant hardware damage, the system may still experience functional degradation under high computational load conditions due to the inability to complete critical tasks on time, thus rendering the analysis ineffective. Therefore, relying solely on traditional hardware failure analysis methods is insufficient to accurately describe the evolution of an autonomous vehicle system under complex task conditions, from computational resource constraints and functional degradation to task failure.

[0004] In existing technologies, reliability analysis under varying system load conditions generally employs load-sharing reliability modeling methods to analyze the failure processes of each component in the system. For example, for mechanical characteristic loads [Yang Jing, Zhao Kelun, Li Hui. Reliability Analysis of Redundant Systems with Load Sharing [J]. Reliability and Environmental Testing of Electronic Products, 2024, 42(05): 41-44.], the system reliability under mechanical loads is approximated using the Weibull distribution. However, such methods mainly focus on physical load scenarios in mechanical components, electrical elements, or general redundant systems. The loads studied are mostly structural loads, electrical stresses, or component loads in a general sense. They do not consider the dynamic changes in computational resource loads, local task allocation, resource overload failures, and maintenance recovery characteristics in unmanned vehicle systems. Therefore, it is difficult to accurately characterize the operating state and reliability change patterns of unmanned vehicle systems under computational resource constraints.

[0005] Therefore, it is necessary to propose a reliability analysis method for unmanned vehicles that considers the impact of computing resource load. This method incorporates factors such as changes in computing resource load, resource overload failures, maintenance recovery, and system state transitions into a unified analysis framework, thereby enabling effective analysis and evaluation of the operational reliability of unmanned vehicle systems. Summary of the Invention

[0006] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a method and apparatus for reliability analysis of unmanned vehicles that considers the impact of computing resource load, providing methodological support for reliability modeling and quantitative analysis of unmanned vehicle systems under complex task conditions.

[0007] Technical Solution: To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0008] A reliability analysis method for autonomous vehicles that considers the impact of computational resource load includes the following steps:

[0009] The structure and principle of the autonomous vehicle system are analyzed, and the failure criteria of the autonomous vehicle system under the influence of computing resource load are determined. The failure criteria include the resource overload failure criterion and the load-related performance degradation criterion.

[0010] The unmanned vehicle system under the influence of computing resource load is regarded as a repairable system. The operating state and recovery characteristics of the unmanned vehicle system under the influence of computing resource load are described, and the system state transition relationship is established.

[0011] Introducing computational resource load constraints into the traditional GO method, a new GO method operator 2ML is designed. The GO method operator 2ML has state output logic defined by an operation rule table. The operation rule table takes different combinations of working states of two functional modules with computational load transfer relationships as input to determine the functional output state of the dual-module load-sharing unit. Based on this, combined with the failure rate and maintenance rate of the functional modules, a quantitative calculation formula for the operator is established through state transition equations.

[0012] A quantitative analysis algorithm based on the GO method that takes into account computational load is constructed, and the reliability index of the autonomous vehicle system is calculated.

[0013] Furthermore, the resource overload failure criteria include: determining the total load on real-time computing resources. With available computing resource capacity The relationship, when satisfied When the system fails, it is determined that the autonomous vehicle system has failed due to overload of computing resources; these computing resources include processor computing power, storage resources, data transmission capabilities, and task scheduling capabilities.

[0014] The load-related performance degradation criteria include: when the system is not overloaded, the failure probability of each functional module dynamically changes with its computational load rate; wherein, the operating state of the system before reaching the overload failure criterion is described by a partial load lifetime model, and the modules... Partial load life model Defined as , For module Failure rate under partial load of computing resources Regarding load factor Monotonically increasing mapping function, load factor Based on the predefined computational load allocation ratio and the total system load rate, the real-time failure rate of each functional module under computational resource overload conditions is obtained. .

[0015] Furthermore, the autonomous vehicle system under the influence of computational resource load is considered a repairable system. The operating state and recovery characteristics of the autonomous vehicle system under the influence of computational resource load are described, and system state transition relationships are established, specifically including:

[0016] Describe the "failure-repair-recovery" operational characteristics of an autonomous vehicle system under the influence of computing resource load. The failure of functional modules due to excessive computing resource load, task processing delay, or computing performance degradation, and the recovery process through restart, reconstruction, or repair, are regarded as repairable system behaviors.

[0017] The impact of computing resource load on the autonomous vehicle system at any time The equivalent failure rate is The equivalent maintenance rate is Establish the system within a tiny time interval The probabilistic relationship between the transition between normal and failure states, where if the system at time... If in a normal state, then after After a certain period of time, the probability of the system failing is: The probability that the system will remain in a normal state is If the system is at time ; If it is in a failed state, then after The probability that the system will be repaired and restored to a normal state after a certain period of time is... The probability that the system remains in a failed state is ,in, Indicates in A higher-order infinitesimal that undergoes two or more state changes within a time period.

[0018] Furthermore, assuming the dual-module load sharing unit includes module 1 and module 2, which are functional modules that calculate load transfer relationships, the operation rules of the GO method operator 2ML include:

[0019] Rule 0: When both Module 1 and Module 2 are operating normally under partial load, the output load sharing unit operates normally.

[0020] Rule 1: When module 1 fails or is under maintenance and module 2 continues to operate normally after switching to overload mode, the output load sharing unit will operate normally.

[0021] Rule 2: When module 2 fails or is under maintenance and module 1 continues to operate normally after switching to overload mode, the output load sharing unit is normal;

[0022] Rule 3: When module 1 is under maintenance and module 2 is in a waiting-for-maintenance state, the output load sharing unit fails;

[0023] Rule 4: When both Module 1 and Module 2 are in maintenance status, the output load sharing unit fails.

[0024] Furthermore, the process of establishing the quantitative calculation formula for the GO method operator 2ML includes:

[0025] Based on the state description of the GO operator 2ML, For module 1 in the case of computing resource overload The efficiency of momentary failure, To calculate the failure rate of module 1 under partial resource load conditions, For module 1 in Repair rate at any time Module 2 in case of computing resource overload The efficiency of momentary failure, To calculate the failure rate of module 2 under partial resource load conditions, For module 2 in Based on the repair rate at any given time, obtain the state transition probability matrix for each functional module:

[0026] ;

[0027] Among them, intermediate variables ;

[0028] Modules are established using the law of total probability. The system of state probability differential equations at time t:

[0029] ;

[0030] By solving the system of differential equations using the Laplace transform, the instantaneous quantitative calculation formula for the operator 2ML is obtained:

[0031] ;

[0032] In the formula, , and These represent the success probability, failure rate, and repair rate of the operator 2ML, respectively.

[0033] Furthermore, a quantitative analysis algorithm based on the GO method considering computational load is constructed, and reliability indices of the autonomous vehicle system are calculated, specifically including:

[0034] Based on the functional structure and logical relationships between modules of a single autonomous vehicle system, a GO graph model of the system is established; for dual-module units affected by computational resource load, the 2ML operator is used for modeling, while the remaining parts are modeled using the traditional GO method operator; the dual-module unit refers to the combination of two functional modules that have a local distribution and transfer relationship of computational load during the execution of computational tasks;

[0035] The success probability of the 2ML operator is determined based on the computational resource load status, where the system load rate is... Calculated according to the instantaneous quantitative formula, when The success probability of the output is determined to be zero. The success probability of each node is calculated step by step according to the signal flow direction of the GO diagram. Finally, the success probability and failure probability of the top-level output node of the system are obtained as the system reliability index.

[0036] Furthermore, according to claim 6, the method is characterized in that the system load rate Defined as time Total actual computing resource load of the system With available computing resource capacity The ratio of .

[0037] An unmanned vehicle reliability analysis device that considers the impact of computational resource load includes:

[0038] The criterion determination module is used to analyze the structure and principle of the autonomous vehicle system and determine the failure criteria of the autonomous vehicle system under the influence of computing resource load. The failure criteria include the resource overload failure criterion and the load-related performance degradation criterion.

[0039] The state transition modeling module is used to treat the unmanned vehicle system under the influence of computing resource load as a repairable system, describe the operating state and recovery characteristics of the unmanned vehicle system under the influence of computing resource load, and establish the system state transition relationship.

[0040] An operator construction module is used to introduce computational resource load constraints into the traditional GO method and design a new GO method operator 2ML. The GO method operator 2ML has state output logic defined by an operation rule table. The operation rule table takes different combinations of working states of two functional modules with computational load transfer relationships as input to determine the functional output state of the dual-module load-sharing unit. Based on this, combined with the failure rate and maintenance rate of the functional modules, a quantitative calculation formula for the operator is established through state transition equations.

[0041] The GO method analysis module is used to construct a quantitative analysis algorithm for the GO method that takes into account computational load, and to calculate the reliability index of the autonomous vehicle system.

[0042] The present invention also provides an electronic device, characterized in that it includes: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the unmanned vehicle reliability analysis method considering the impact of computational resource load as described above.

[0043] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the unmanned vehicle reliability analysis method considering the impact of computing resource load as described above.

[0044] The present invention also provides a computer program product, including a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the unmanned vehicle reliability analysis method considering the impact of computing resource load as described above.

[0045] Beneficial effects:

[0046] (1) This invention introduces the computing resource load factor into the reliability analysis of unmanned vehicle system, establishes a computing resource load sharing model and system failure criteria, and can characterize the impact of dynamic changes in computing resources, local allocation of tasks and resource overload on the functional status and task completion capability of unmanned vehicle system. It overcomes the shortcomings of existing methods that mainly focus on physical component failures and are difficult to reflect the role of computing resource constraints.

[0047] (2) This invention regards the unmanned vehicle system under the influence of computing resource load as a repairable system, and establishes the system equivalent failure rate, equivalent maintenance rate and state transition relationship. It can describe the dynamic evolution law of the system in the process of "normal-failure-maintenance-recovery", thereby more realistically reflecting the reliability change characteristics of the unmanned vehicle system in the actual operation process.

[0048] (3) This invention introduces computational resource load constraints into the traditional GO method, designs a new GO method operator 2ML, and constructs a quantitative analysis algorithm for the GO method that considers computational load. It can realize the quantitative evaluation of reliability indicators such as success probability, failure probability and availability of a single unmanned vehicle system, and improves the applicability and effectiveness of the GO method in the reliability analysis of unmanned vehicle systems under computational resource load conditions. Attached Figure Description

[0049] Figure 1 This is a schematic diagram illustrating the steps of the unmanned vehicle reliability analysis method that takes into account the impact of computational resource load according to the present invention.

[0050] Figure 2 This is a diagram of the GO model of the autonomous vehicle system constructed by the present invention, taking into account the impact of computational resource load.

[0051] Figure 3 This is a comparison chart of the output results of the method of the present invention and the method of 10,000 Monte Carlo simulations.

[0052] Figure 4 This is a comparison chart of the output results of the method of the present invention and the Monte Carlo simulation method of 100,000 times.

[0053] Figure 5 This is a comparison chart of the output results of the method of the present invention and the one million Monte Carlo simulation method. Detailed Implementation

[0054] To provide a clearer understanding of the features and advantages of the technical solution of the present invention, the composition and implementation of the specific solution are described below in conjunction with the accompanying drawings.

[0055] This invention provides a reliability analysis method for unmanned vehicles (UAVs) that considers the impact of computational resource load. Addressing issues such as functional degradation, resource overload failure, and maintenance recovery caused by dynamic changes in computational resource load during task execution, this method establishes a computational resource load sharing model, constructs system failure criteria and repairable system state transition relationships, designs new Go-of-Time (GO) operators, and extends the corresponding GO quantitative analysis algorithm. This enables the analysis and evaluation of reliability indicators such as success probability, failure probability, and availability for individual UAV systems, providing methodological support for reliability modeling and quantitative analysis of UAV systems under complex task conditions.

[0056] Reference Figure 1 A reliability analysis method for autonomous vehicles that considers the impact of computational resource load includes the following steps:

[0057] S1 analyzes the reliability characteristics of the unmanned vehicle system considering the impact of computing resource load and determines its failure criteria.

[0058] Specifically, it includes:

[0059] S11, firstly, analyze the structure and principle of the unmanned vehicle system under the influence of computational resource load.

[0060] Autonomous vehicle systems typically consist of a perception module, a communication module, a computing module, a decision-making and control module, and an execution module. Among these, the computing module is the core of the autonomous vehicle's environmental perception, information fusion, path planning, task decision-making, and control calculations. Computing resources mainly include processor processing power, storage resources, data transmission capabilities, and task scheduling capabilities. During task execution, the autonomous vehicle needs to continuously invoke various algorithms to complete state estimation, target recognition, cooperative control, and task response; therefore, computing resources are always dynamically occupied. The computing requirements differ across different task phases; during phases such as cruising, searching, recognition, and cooperative operations, the computing load typically increases significantly.

[0061] When computing resources are sufficient, the perception, decision-making, control, and communication modules can complete their respective computing tasks within a specified time, and the system can maintain normal operation. However, when the computing resource load continues to increase, task processing delays, control output lags, slower communication responses, or degradation of some functions may occur, leading to a decrease in the autonomous vehicle's task execution capability. Therefore, changes in computing resource load directly affect the functional state of the autonomous vehicle system and further impact task completion.

[0062] Therefore, in the reliability analysis of autonomous vehicle systems, it is not enough to only consider physical failures of components such as sensors, actuators, or communication hardware; the impact of computational load variations on the system's ability to maintain functionality must also be considered. Especially under complex and collaborative task conditions, even without significant hardware failure, computational resource overload can lead to functional degradation or even task failure, thus affecting the system reliability assessment results. Therefore, it is necessary to consider computational resource load as a crucial influencing factor in the reliability analysis of autonomous vehicle systems to more accurately characterize the operational status and failure processes of autonomous vehicle systems during actual tasks.

[0063] S12. Based on the above analysis, the failure criteria of the unmanned vehicle system under the influence of computing resource load are determined.

[0064] The amount of computational tasks undertaken by autonomous vehicles varies at different stages of the mission, and the system is also affected by factors such as environmental changes, changes in cooperative relationships, and local functional abnormalities during operation, causing the level of computational resource occupancy to change dynamically over time. Therefore, this invention uses a variable load form to describe the computational resource occupancy of the autonomous vehicle system.

[0065] Suppose a single autonomous vehicle system at time... The actual total computing resource load is Available computing resource capacity is Then define the total system load rate. for:

[0066]

[0067] When a computing module in an autonomous vehicle system experiences performance degradation or malfunction, some of its computing tasks will be transferred to other normally functioning modules, resulting in a redistribution of the system's internal computing load. Considering the clear functional relationships and task couplings among modules such as perception, decision-making, control, and communication, the computing tasks of the malfunctioning module are usually preferentially transferred to the module with the closest functional similarity, strongest coupling, or most intimate connection to it.

[0068] In this invention, a local allocation method is used to describe the transfer process of computational tasks within the autonomous vehicle system. Let module 1 and module 2 at time... The allocation ratios of the total system computing load are as follows: and ,in Module 1 and Module 2 at time The respective load percentages are:

[0069]

[0070] (1) Resource overload criterion:

[0071] Determine the total load of real-time computing resources With available computing resource capacity The relationship. When satisfied. When the system load exceeds the capacity threshold, it is determined that the autonomous vehicle system has failed due to computing resource overload. This indicates an overload failure state, where the system load exceeds the capacity threshold and cannot maintain normal operation through localized internal task allocation.

[0072] (2) Dynamic reliability mapping criterion:

[0073] The system's operating state before reaching the aforementioned failure criteria is described by a partial load life model. The life distribution functions of the two modules under partial load conditions are defined. for:

[0074]

[0075] in, For module Failure rate under partial load of computing resources Regarding load factor A monotonically increasing mapping function is used to characterize the real-time impact of different task load levels on the failure rate of functional modules. The failure rates of each module under computational resource overload are as follows: , .

[0076] Therefore, this invention uses the failure criterion of unmanned vehicle system under the influence of computing resource load as the failure of computing resource load to exceed the available computing resource capacity of the system after local allocation, or the failure of key functions to meet task requirements.

[0077] S2 treats the unmanned vehicle system under the influence of computing resource load as a repairable system and establishes the system state transition relationship.

[0078] Specifically, it includes:

[0079] S21 describes the operating status and recovery characteristics of the unmanned vehicle system under the influence of computing resource load.

[0080] In this embodiment, the unmanned vehicle system consists of multiple functional modules constrained by computing resources. Each module has only two operating states: a normal state and a failed state. When the system starts operating, all modules are in the normal state. During task execution, each module may fail due to excessive computing resource load, task processing delays, scheduling blockages, response lags, or decreased computing performance. After a module fails, its normal operating capability can be restored through module restart, functional reconstruction, software recovery, or hardware repair, allowing it to rejoin system operation. Therefore, the unmanned vehicle system under the influence of computing resource load exhibits a "failure-repair-recovery" operating characteristic and can be considered a repairable system.

[0081] S22, Establish the state transition relationship of the unmanned vehicle system under the influence of computing resource load.

[0082] The impact of computing resource load on the autonomous vehicle system at any time The equivalent failure rate is The equivalent maintenance rate is In any infinitesimal time interval Within, if the system at time If in a normal state, then after After a certain period of time, the probability of the system failing is: The probability that the system will remain in a normal state is If the system is at time... If it is in a failed state, then after The probability that the system will be repaired and restored to a normal state after a certain period of time is... The probability that the system remains in a failed state is... .in, Indicates in A higher-order infinitesimal that undergoes two or more state changes within a time interval. and The system equivalent parameters are given in their specific form by the quantitative calculation of the operator 2ML in the following text.

[0083] S3 introduces computational resource load constraints into the traditional GO method, designs a new GO method operator 2ML (2-Module with Load-transfer), and establishes a quantitative calculation formula for this operator.

[0084] Specifically, it includes:

[0085] S31, Design the state description of the GO operator 2ML.

[0086] First, we define the concept of a dual-module load-sharing unit. In an autonomous vehicle system, a dual-module load-sharing unit is defined as two functional modules that have computational load transfer due to redundant design or functional coupling. Specifically, this includes homogeneous redundant module pairs such as dual computing chips, dual sensing channels, and dual communication links, as well as heterogeneous module pairs with coupled computational load requirements, such as the perception module and the decision-making and control module. When perception degrades, the decision-making module performs compensatory calculations, increasing its load. Without loss of generality, we assume that the dual-module load-sharing unit contains module 1 and module 2.

[0087] The operator 2ML takes the combined operating status of module 1 and module 2 as input and outputs whether the load-sharing unit is functioning correctly. A value of 1 indicates that the load-sharing unit is functioning normally (i.e., capable of completing the assigned computing tasks), while 0 indicates that the unit is malfunctioning. The module's operating status includes: partial load normal, overload normal, under maintenance, and awaiting maintenance. Assume the maintenance resource is 1, and only one module can be repaired at a time. The operation rules are shown in Table 1.

[0088] Table 1. Type 2 ML Operator Operation Rules

[0089] rule Module 1 status Module 2 status Unit output G(t) meaning 0 Partial load normal Partial load normal 1 The two modules are sharing the load normally, and the unit is functioning normally. 1 Under repair Overload normal 1 Module 1 is faulty and under repair; Module 2 is overloaded but can still complete the calculation tasks of this unit; the unit is normal. 2 Overload normal Under repair 1 Module 2 is under repair due to a fault. Module 1 is overloaded but can still complete the calculation tasks of this unit. The unit is functioning normally. 3 Under repair Waiting for repair 0 Both modules failed, unit inoperability. 4 Waiting for repair Under repair 0 Both modules failed, unit inoperability.

[0090] To facilitate the subsequent establishment of the state transition differential equation, we define... and Modules 1 and 2 under partial load conditions, respectively. Failure rate at any moment; and These are the scenarios for module 1 and module 2 under resource overload conditions. Failure rate at any moment; and Module 1 and Module 2 are respectively in The maintenance rate at any given time.

[0091] S32, Design a quantitative calculation formula for the GO method operator 2ML.

[0092] Based on the state description of the GO operator 2ML, the five states are numbered sequentially as follows: Define the state probability vector

[0093]

[0094] in Indicates at time The unit is in state The probability satisfies The initial conditions are: The rest are .

[0095] Considering the constraints of a maintenance station, the transition events between states and their transition rates are as follows:

[0096] · Module 1 failed under partial load, transfer rate

[0097] · Module 2 failed under partial load, transfer rate

[0098] · Module 1 repair, transfer rate

[0099] · Module 2 repair, transfer rate

[0100] · Module 2 failed due to overload before Module 1 was repaired, resulting in a transfer rate of [missing information].

[0101] · Module 1 failed due to overload before Module 2 was repaired, resulting in a transfer rate of [missing information].

[0102] · Module 1 completes repair and is immediately put into operation (at this time, Module 2 enters maintenance mode), transfer rate

[0103] · Module 2 completes repairs and is immediately put into operation (at this time, Module 1 enters maintenance mode), transfer rate

[0104] The state transition probability matrix for each module can be obtained:

[0105]

[0106] in, .

[0107] According to the law of total probability, the module in The state probability expression at time t is:

[0108]

[0109] remember for The derivative, when Then, from the above equation, the state differential equations of the module can be obtained as follows:

[0110]

[0111] The solver module is in probability at time step Taking the Laplace transform of the above equation, we get:

[0112]

[0113] in, , , The complex conjugate is , For the Laplace operator.

[0114] Therefore, the instantaneous quantitative calculation formula for the operator 2ML is as follows:

[0115]

[0116] In the formula, , and These represent the success probability, failure rate, and repair rate of the operator 2ML, respectively.

[0117] S4. Construct a quantitative analysis algorithm for the GO method that takes into account computational load, and calculate the reliability index of the autonomous vehicle system.

[0118] Specifically, it includes:

[0119] S41, Establish a GO graph model of the autonomous vehicle system that takes into account the computational load;

[0120] In this embodiment, based on the state description and quantitative calculation formula of the GO method operator 2ML, a quantitative analysis algorithm for the GO method considering the impact of computational resource load is constructed. First, a GO graph model of the system is established according to the functional structure and logical relationships between modules of a single autonomous vehicle system. Assume the system GO graph includes... The nth operator node, the nth The success probability of the output of each operator node is denoted as . For ordinary operators in the GO graph, their success probability is calculated according to the traditional GO algorithm rules and is uniformly expressed as:

[0121]

[0122] in, Indicates the first The set of input nodes for each operator. For the cumulative probability of the state, This refers to the signal flow operation function corresponding to the operator. For a dual-module unit affected by computational resource load, the dual-module unit refers to a combination of two functional modules that have a local distribution and transfer relationship of computational load during the execution of the computational task; the operator 2ML is used for modeling, and its output success probability is determined by the instantaneous quantitative calculation formula obtained in S32, denoted as:

[0123]

[0124] in, For the system at time The actual total load on computing resources. For the system at time Available computing resource capacity.

[0125] S42, the system reliability index is obtained by recursively calculating according to the signal flow direction in the GO diagram;

[0126] Based on the computational resource load status, determine the success probability of the 2ML operator's output: when the following conditions are met... When the system is in a computable state, it is calculated according to the instantaneous quantitative formula described above; when the condition is met... When the system enters an overload failure state, the success probability of the corresponding operator output is determined as follows:

[0127]

[0128] Then, following the signal flow direction in the GO diagram, the output probability of each operator node is recursively calculated level by level to obtain the success probability sequence of each node in the system: Let the top-level output node be the... If there are nodes, then a single autonomous vehicle system at time... The probability of success is:

[0129]

[0130] Therefore, by using the GO method quantitative analysis algorithm that takes into account computational load, the operational reliability of the autonomous vehicle system can be quantitatively assessed under dynamic changes in computational resource load.

[0131] To verify the effectiveness of the reliability analysis method for unmanned vehicle systems that considers the impact of computational resource load proposed in this invention, MATLAB 2019b is used as the simulation software in this embodiment. At the same time, the method of this invention is compared and verified by 10,000 Monte Carlo simulations, 100,000 Monte Carlo simulations, and 1 million Monte Carlo simulations.

[0132] In this embodiment, MATLAB 2019b is used as the simulation software. It is assumed that an autonomous vehicle includes the following modules: a perception module, a communication module, a computing module, a decision-making and control module, and an execution module. Figure 2 As shown, a GO graph model of a single autonomous vehicle system is established. In the figure, 5-1 is type 5 operator number 1, representing the startup unit; 1-2, 1-3, 1-4, 1-5, and 1-6 are type 1 operators numbered 2 to 6, corresponding to the perception module, communication module, computing module, decision control module, and execution module, respectively; and operators 10-11 are AND logic gates. 2ML-7, 2ML-8, 2ML-9, and 2ML-10 are the GO method operator 2ML designed in this invention. Their output results are summarized by the output node numbered 11 to obtain the success probability of the top-level output of a single autonomous vehicle system.

[0133] In this embodiment, the success probability of the start operator 5-1 is 0.99994. The failure rates of the modules corresponding to the type 1 operators numbered 2 to 6 are respectively:

[0134] .

[0135] The corresponding repair rates are respectively:

[0136] .

[0137] For the operators 2ML-7, 2ML-8, 2ML-9, and 2ML-10, the actual total computational resource load is taken as follows:

[0138] .

[0139] The available computing resource capacity is uniformly set as: .

[0140] The load distribution ratios are respectively set as follows: .

[0141] Therefore, the load factor for each operator is: .

[0142] Right now: .

[0143] In this embodiment, the load influence function is taken as: According to the instantaneous quantitative calculation formula of the aforementioned 2ML operator, the output success probability of each node is calculated, and then recursively extrapolated along the signal flow direction of the GO diagram to obtain the success probability sequence of each node in the system: Let node number 11 be the top-level output node. Then the success probability of a single autonomous vehicle system at time t is: The system failure probability is: .

[0144] Figure 3 The comparison between the method of this invention and the results of 10,000 Monte Carlo simulations is shown. The solid black line represents the availability change curve of a single autonomous vehicle system within a given task period, calculated using the method of this invention; the gray dotted square line represents the availability change curve of a single autonomous vehicle system within a given task period, obtained using 10,000 Monte Carlo simulations. Figure 3 As can be seen, during the process of increasing the task execution time from 0 hours to 200 hours, the overall system reliability shows a gradual decreasing trend and eventually tends to stabilize. The black solid line changes relatively smoothly and can well characterize the decay law of system reliability over time; although the gray dotted line generally changes in the same trend as the black solid line, it still exhibits a certain degree of random fluctuation. The inset in the figure further shows the comparison results of the computation time of the two methods, where the computation time of the method of this invention is approximately 5 seconds, and the computation time of 10,000 Monte Carlo simulations is approximately 10 seconds.

[0145] Figure 4 The diagram shows a comparison between the method of this invention and the results of 100,000 Monte Carlo simulations. The solid black line represents the availability change curve of a single autonomous vehicle system within a given task period, calculated using the method of this invention; the gray dotted square line represents the availability change curve of a single autonomous vehicle system within a given task period, obtained using 100,000 Monte Carlo simulations. Figure 4 It can be seen that as the number of Monte Carlo simulations increases to 100,000, the random fluctuations in the gray square dots and lines are relatively... Figure 3 The value decreases significantly, and its overall trend is closer to the black solid line, indicating that as the number of samplings increases, the Monte Carlo simulation results gradually stabilize and converge towards the results obtained by the method of this invention. The inset in the figure further illustrates the comparison of computation time between the two methods, where the computation time of the method of this invention is still approximately 5 seconds, while the computation time for 100,000 Monte Carlo simulations is approximately 100 seconds.

[0146] Figure 5 The diagram shows a comparison between the method of this invention and the results of 1 million Monte Carlo simulations. The solid black line represents the availability change curve of a single autonomous vehicle system within a given task period, calculated using the method of this invention; the gray dotted square line represents the availability change curve of a single autonomous vehicle system within a given task period, obtained using 1 million Monte Carlo simulations. Figure 5As can be seen, when the number of Monte Carlo simulations is further increased to 1 million, the random fluctuations in the gray square dotted line are basically eliminated, and its overall trend basically coincides with the black solid line, indicating that the method of the present invention has high evaluation accuracy and can accurately characterize the evolution of system reliability over time. The inset in the figure further shows the comparison of the computation time of the two methods, where the computation time of the method of the present invention is about 5 seconds, while the computation time of 1 million Monte Carlo simulations is about 1000 seconds.

[0147] Depend on Figures 3 to 5 As can be seen, as the number of Monte Carlo simulations gradually increases from 10,000 to 100,000 and then to 1 million, the random fluctuations in the Monte Carlo simulation results (represented by the gray dotted line) gradually weaken, and the reliability change curve gradually stabilizes, eventually becoming essentially consistent with the results of the method of this invention (represented by the black solid line). This indicates that the reliability analysis method for unmanned vehicle systems proposed in this invention, which considers the impact of computational resource load, has good accuracy and effectiveness. Furthermore, compared to high-order Monte Carlo simulation methods, the method of this invention significantly reduces computation time and cost while maintaining evaluation accuracy, making it more suitable for rapid quantitative analysis of the reliability and availability of unmanned vehicle systems in complex task scenarios.

[0148] The simulation results further demonstrate that the reliability analysis method for unmanned vehicle systems proposed in this invention, which considers the impact of computational resource load, can effectively characterize the evolution of system availability over task execution time. Its evaluation results are largely consistent with those of high-order Monte Carlo simulations, thus verifying the accuracy and effectiveness of the method. Simultaneously, the method exhibits significant advantages in computational efficiency, enabling quantitative analysis of system reliability and availability at a relatively low computational cost.

[0149] In summary, the unmanned vehicle system reliability analysis method considering the impact of computing resource load provided by this invention can achieve rapid and effective quantitative analysis of the reliability and availability of unmanned vehicle systems in complex task scenarios. It has the advantages of high evaluation accuracy, low computing cost, and strong engineering applicability, and can provide effective analytical basis for unmanned vehicle system reliability design, task planning, and operation and maintenance.

[0150] Based on the same inventive concept, the present invention provides an unmanned vehicle reliability analysis device that considers the impact of computational resource load, comprising:

[0151] The criterion determination module is used to analyze the structure and principle of the autonomous vehicle system and determine the failure criteria of the autonomous vehicle system under the influence of computing resource load. The failure criteria include the resource overload failure criterion and the load-related performance degradation criterion.

[0152] The state transition modeling module is used to treat the unmanned vehicle system under the influence of computing resource load as a repairable system, describe the operating state and recovery characteristics of the unmanned vehicle system under the influence of computing resource load, and establish the system state transition relationship.

[0153] An operator construction module is used to introduce computational resource load constraints into the traditional GO method and design a new GO method operator 2ML. The GO method operator 2ML has state output logic defined by an operation rule table. The operation rule table takes different combinations of working states of two functional modules with computational load transfer relationships as input to determine the functional output state of the dual-module load-sharing unit. Based on this, combined with the failure rate and maintenance rate of the functional modules, a quantitative calculation formula for the operator is established through state transition equations.

[0154] The GO method analysis module is used to construct a quantitative analysis algorithm for the GO method that takes into account computational load, and to calculate the reliability index of the autonomous vehicle system.

[0155] It should be understood that the unmanned vehicle reliability analysis device considering the impact of computing resource load in the embodiments of the present invention can realize all the technical solutions in the above method embodiments. The functions of each functional module can be specifically implemented according to the methods in the above method embodiments. The specific implementation process can be referred to the relevant descriptions in the above embodiments, which will not be repeated here.

[0156] The present invention also provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the unmanned vehicle reliability analysis method considering the impact of computational resource load as described above.

[0157] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the unmanned vehicle reliability analysis method considering the impact of computing resource load as described above.

Claims

1. A reliability analysis method for unmanned vehicles considering the impact of computational resource load, characterized in that, Includes the following steps: The structure and principle of the autonomous vehicle system are analyzed, and the failure criteria of the autonomous vehicle system under the influence of computing resource load are determined. The failure criteria include the resource overload failure criterion and the load-related performance degradation criterion. The unmanned vehicle system under the influence of computing resource load is regarded as a repairable system. The operating state and recovery characteristics of the unmanned vehicle system under the influence of computing resource load are described, and the system state transition relationship is established. Introducing computational resource load constraints into the traditional GO method, a new GO method operator 2ML is designed. The GO method operator 2ML has state output logic defined by an operation rule table. The operation rule table takes different combinations of working states of two functional modules with computational load transfer relationships as input to determine the functional output state of the dual-module load-sharing unit. Based on this, combined with the failure rate, maintenance rate, and load rate of the functional modules, a quantitative calculation formula for the operator is established through state transition equations. A quantitative analysis algorithm based on the GO method that takes into account computational load is constructed, and the reliability index of the autonomous vehicle system is calculated.

2. The method according to claim 1, characterized in that, Resource overload failure criteria include: determining the total load on real-time computing resources. With available computing resource capacity The relationship, when satisfied When the system fails, it is determined that the autonomous vehicle system has failed due to overload of computing resources; these computing resources include processor computing power, storage resources, data transmission capabilities, and task scheduling capabilities. The load-related performance degradation criteria include: when the system is not overloaded, the failure probability of each functional module dynamically changes with its computational load rate; wherein, the operating state of the system before reaching the overload failure criterion is described by a partial load lifetime model, and the modules... Partial load life model Defined as , For module Failure rate under partial load of computing resources Regarding load factor Monotonically increasing mapping function, load factor Based on the predefined computational load allocation ratio and the total system load rate, the real-time failure rate of each functional module under computational resource overload conditions is obtained. .

3. The method according to claim 1, characterized in that, The autonomous vehicle system under the influence of computational resource load is treated as a repairable system. The operating state and recovery characteristics of the autonomous vehicle system under the influence of computational resource load are described, and the system state transition relationship is established, specifically including: Describe the "failure-repair-recovery" operation characteristics of an autonomous vehicle system under the influence of computing resource load. The failure of functional modules due to excessive computing resource load, task processing delay or computing performance degradation and the recovery process through restart, reconstruction or repair are regarded as repairable system behaviors. The impact of computing resource load on the autonomous vehicle system at any time The equivalent failure rate is The equivalent maintenance rate is Establish the system within a tiny time interval The probabilistic relationship between the transition between normal and failure states, where if the system at time... If in a normal state, then after After a certain period of time, the probability of the system failing is: The probability that the system will remain in a normal state is If the system is at time ; If it is in a failed state, then after The probability that the system will be repaired and restored to a normal state after a certain period of time is... The probability that the system remains in a failed state is ,in, Indicates in A higher-order infinitesimal that undergoes two or more state changes within a time period.

4. The method according to claim 1, characterized in that, Assuming the dual-module load sharing unit includes module 1 and module 2, which are functional modules that calculate load transfer relationships, the operation rules of the GO method operator 2ML include: Rule 0: When both Module 1 and Module 2 are operating normally under partial load, the output load sharing unit operates normally. Rule 1: When module 1 fails or is under maintenance and module 2 continues to operate normally after switching to overload mode, the output load sharing unit will operate normally. Rule 2: When module 2 fails or is under maintenance and module 1 continues to operate normally after switching to overload mode, the output load sharing unit is normal; Rule 3: When module 1 is under maintenance and module 2 is in a waiting-for-maintenance state, the output load sharing unit fails; Rule 4: When both Module 1 and Module 2 are in maintenance status, the output load sharing unit fails.

5. The method according to claim 4, characterized in that, The process of establishing the quantitative calculation formula for the GO method operator 2ML includes: Based on the state description of the GO operator 2ML, For module 1 in the case of computing resource overload The efficiency of momentary failure, To calculate the failure rate of module 1 under partial resource load conditions, For module 1 in Repair rate at any time Module 2 in case of computing resource overload The efficiency of momentary failure, To calculate the failure rate of module 2 under partial resource load conditions, For module 2 in Based on the repair rate at any given time, obtain the state transition probability matrix for each functional module: ； Among them, intermediate variables ; Modules are established using the law of total probability. The system of state probability differential equations at time t: ； By solving the system of differential equations using the Laplace transform, the instantaneous quantitative calculation formula for the operator 2ML is obtained: ； In the formula, , and These represent the success probability, failure rate, and repair rate of the operator 2ML, respectively.

6. The method according to claim 1, characterized in that, A quantitative analysis algorithm based on the Go method, considering computational load, is constructed, and reliability indices of the autonomous vehicle system are calculated, specifically including: Based on the functional structure and logical relationships between modules of a single autonomous vehicle system, a GO graph model of the system is established; for dual-module units affected by computational resource load, the 2ML operator is used for modeling, while the remaining parts are modeled using the traditional GO method operator; the dual-module unit refers to the combination of two functional modules that have a local distribution and transfer relationship of computational load during the execution of computational tasks; The success probability of the 2ML operator is determined based on the computational resource load status, where the system load rate is... Calculated according to the instantaneous quantitative formula, when The success probability of the output is determined to be zero. The success probability of each node is calculated step by step according to the signal flow direction of the GO diagram. Finally, the success probability and failure probability of the top-level output node of the system are obtained as the system reliability index.

7. The method according to claim 6, characterized in that, System load rate Defined as time Total actual computing resource load of the system With available computing resource capacity The ratio of .

8. A reliability analysis system for unmanned vehicles that considers the impact of computational resource load, characterized in that, include: The criterion determination module is used to analyze the structure and principle of the autonomous vehicle system and determine the failure criteria of the autonomous vehicle system under the influence of computing resource load. The failure criteria include the resource overload failure criterion and the load-related performance degradation criterion. The state transition modeling module is used to treat the unmanned vehicle system under the influence of computing resource load as a repairable system, describe the operating state and recovery characteristics of the unmanned vehicle system under the influence of computing resource load, and establish the system state transition relationship. An operator construction module is used to introduce computational resource load constraints into the traditional GO method and design a new GO method operator 2ML. The GO method operator 2ML has state output logic defined by an operation rule table. The operation rule table takes different combinations of working states of two functional modules with computational load transfer relationships as input to determine the functional output state of the dual-module load-sharing unit. Based on this, combined with the failure rate, maintenance rate and load rate of the functional modules, a quantitative calculation formula for the operator is established through state transition equations. The GO method analysis module is used to construct a quantitative analysis algorithm for the GO method that takes into account computational load, and to calculate the reliability index of the autonomous vehicle system.

9. An electronic device, characterized in that, include: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the autonomous vehicle reliability analysis method considering the impact of computational resource load as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the unmanned vehicle reliability analysis method considering the impact of computing resource load as described in any one of claims 1-7.

Images