Analysis method and device for autonomous cognition and decision behavior failure mode of unmanned ship

Abstract

The invention provides an unmanned surface vehicle autonomous cognition and decision behavior failure mode analysis method and device, and the method comprises the steps: obtaining the data of the autonomous cognition and decision behavior failure mode of an unmanned surface vehicle, and the data comprise the potential failure path, the element part, the failure mode of the fault, and the occurrence reason of each failure mode; analyzing coupling information existing among different failure modes from structure information, hierarchy and boundary by adopting a method of first division and then total; analyzing the autonomous cognition and decision behavior failure according to the coupling information, and taking the polymorphic sample space construction of the water surface WR system and the typical failure mechanism and robust representation of the autonomous cognition ability as scientific problems; the problems that the natural scene generalization ability is poor, the failure mechanism is unknown and the evaluation system is missing in an autonomous system can be effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to an analysis method and device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat. Background Art

[0002] The concept of autonomy is relatively broad, and there are descriptions of autonomy and autonomy in many fields. However, due to different objects, the descriptions of autonomy and autonomy in different fields are different. In the process of studying unmanned systems, people found that the key reason why unmanned systems can operate without the participation of operators is that they can manage themselves, that is, they have a certain degree of autonomy. Therefore, when studying unmanned systems, the first thing to do is to clarify the connotation of autonomy. The definition of autonomy is involved in many related studies.

[0003] Since the U.S. military first proposed the concept of autonomous operations in 2000, the technical connotation of autonomous drones has been continuously enriched and improved abroad. The evolution of its connotation can be summarized into the following categories:

[0004] (1) The Unmanned Systems Roadmap 2007-2032 states that autonomy is reflected in two aspects of all unmanned systems: on the one hand, it must provide minimal manning and bandwidth requirements and enable the tactical range of operation beyond the line of sight; on the other hand, it must be cooperative (or collaborative) coordination among multiple machines.

[0005] (2) The Autonomy Levels for Unmanned Systems (ALFUS) working group of the National Institute of Standards and Technology of the United States pointed out that autonomy is the ability of unmanned systems to perceive, observe, analyze, communicate, plan, make decisions and act, and complete the tasks assigned to them by humans through human-machine interaction. Autonomy can be divided into levels based on factors such as the complexity of the task, the difficulty of the environment and the degree of human-machine interaction to complete the task, thereby indicating the state and quality of the self-management of the unmanned system.

[0006] (3) American scholar Panos J. Antsaklis and others pointed out that autonomy means the ability to govern itself. Autonomous controllers have the ability and authority to control themselves when performing control functions. Autonomous controllers are composed of a series of hardware and software, and can continuously complete the necessary control functions for a period of time without human intervention.

[0007] (4) The report “The Status of Autonomy in Department of Defense Unmanned Systems” released by the U.S. Defense Science Board in July 2012 states that autonomy is a capability or a set of capabilities that enable unmanned systems to automatically complete certain actions or achieve self-management within the scope of program specifications. “There are no completely autonomous unmanned systems in the world. All autonomous unmanned systems are human-machine joint cognitive systems.” Autonomy does not mean that unmanned systems can complete tasks independently, but rather that human-machine collaboration issues must be considered.

[0008] (5) The "Autonomy" research report released by the U.S. Defense Science Board in June 2016 believes that autonomy is the granting of decision-making power to an entity, giving it the right to act within a specified range to produce results, that is, autonomy is the freedom to make decisions without external interference. In essence, autonomy is a test of whether people are willing to give up control.

[0009] Cognitive abilities refer to the ability to acquire and apply knowledge, or the ability to process information. Specifically, it refers to the ability of the human brain to process, store and retrieve information, that is, people's ability to grasp the composition, performance, relationship with other things, development momentum, development direction and basic laws of things.

[0010] Cognitive abilities include sensation, perception, memory, attention, thinking and language.

[0011] Sensation is the human brain's recognition of the individual attributes of things. It provides people with internal and external information, ensures the information balance between the body and the environment, and is the basis of other higher-level and more complex cognitive abilities (such as perception, memory, etc.), and is also the basis of all human psychological phenomena. People can recognize the color, brightness, smell, etc. of external objects through sensation, and thus can understand the various attributes of things.

[0012] Perception is the understanding of the whole thing in the mind through the direct use of the sense organs. According to the characteristics of the thing to be understood, perception can be divided into spatial perception, temporal perception, motion perception, etc. Through perception, people discover the existence of things, distinguish a certain thing or its attributes from other things or their attributes, and use existing knowledge and experience and currently obtained information to determine the object of perception, name it, and put it into a certain category.

[0013] Memory refers to the process by which people store and retrieve past experiences and apply this information to current situations, that is, the process by which the human brain encodes, stores, and retrieves information input from the outside world. According to the length of time that information is retained, memory is divided into (1) sensory memory, which has a storage time of approximately 0.25-2 seconds; (2) short-term memory, which has a storage time of approximately 5 seconds-2 minutes and generally includes two components: one is direct memory, which means that the input information has not been further processed and has a capacity of approximately 7±2 units; the other component is working memory, which means that the input information has been re-encoded to expand its capacity; (3) long-term memory, which is a permanent storage, ranging from more than 1 minute, and some can last a lifetime, with no capacity limit, and it is divided into two categories: one is episodic memory, which refers to people's memory of a certain time based on time and space relationships, that is, events experienced by individuals; the other is semantic memory, which refers to people's memory of general knowledge and rules, that is, general social knowledge.

[0014] Attention is the active processing of limited information from a large amount of existing information through sensation, stored memory and other cognitive processes. In simple terms, it is the concentration of mental activity or consciousness on a certain object. The basic function of attention is to make corresponding selections for the information provided by the outside world. Attention can be divided into (1) Selective attention is when an individual selects certain stimuli to pay attention to and ignores other stimuli when multiple stimuli are presented at the same time; (2) Sustained attention is to maintain attention to certain stimuli for a certain period of time, that is, the stability of attention; (3) Divided attention is when an individual pays attention to two or more stimuli at the same time.

[0015] Thinking refers to the general and indirect understanding of objective things achieved by people with the help of language, images or actions. Thinking is a reorganization of experience and is general and indirect. In the human mind, the information input from the outside world must be analyzed, synthesized, compared, abstracted and summarized using the knowledge and experience stored in long-term memory. According to the nature and content of the task and the solution method, thinking can be divided into (1) practical thinking, also known as intuitive action thinking, in which the task is presented in an intuitive form and the problem is solved by actual action; (2) figurative thinking, which refers to solving problems with images in the human mind; (3) logical thinking, which refers to using existing concepts and theoretical knowledge to solve theoretical problems.

[0016] Language refers to the behavior of people using highly structured sound combinations, or a symbol system composed of written symbols, gestures, etc., and using this symbol system to communicate ideas. Language is creative, structural, meaningful, referential, social and individual. Its greatest function is to enable people to communicate better with each other. According to different characteristics, language is divided into: (1) Dialogue language refers to the language activities when people communicate directly with each other. It is situational, concise, direct and responsive; (2) Monologue language refers to the long and continuous narrative language carried out by an individual alone. It is solitary, open and planned; (3) Written language refers to the use of words by individuals to express their own thoughts or to accept the influence of others through reading. It is casual, open and planned; (4) Inner language refers to the language activities of asking and answering questions by oneself or silently. It is hidden and concise.

[0017] Decision-making refers to the strategy or method of making decisions. It is the process of people coming up with ideas and making decisions for various events. It is a complex thinking operation process, a process of collecting and processing information, and finally making judgments and drawing conclusions.

[0018] The general decision-making process includes:

[0019] (1) Problem identification, that is, recognizing the entire process of the event, determining where the problem lies, and proposing decision-making goals.

[0020] (2) Problem diagnosis, which involves studying general principles, analyzing and formulating various possible action plans, predicting possible problems and proposing countermeasures.

[0021] (3) Action selection, that is, screening out the best option from various options and establishing a corresponding feedback system.

[0022] Therefore, how to provide an analysis method for the failure modes of autonomous cognition and decision-making behavior of unmanned boats has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0023] The purpose of the present invention is to provide an analysis method and device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat.

[0024] According to a first aspect of the present invention, there is provided an analysis of failure modes of autonomous cognition and decision-making behavior of an unmanned surface vehicle, comprising obtaining data of failure modes of autonomous cognition and decision-making behavior of a surface unmanned vehicle, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode;

[0025] Adopt the method of first dividing and then summarizing to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary;

[0026] The failure of the autonomous cognition and decision-making behavior is analyzed based on the coupling information.

[0027] Optionally, the structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

[0028] Optionally, the method of first dividing and then generalizing is adopted to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary, including:

[0029] According to the preset output requirements, select the highest level of the system;

[0030] The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience;

[0031] Perform system maintenance and repair based on lower system levels.

[0032] Optionally, the data of failure modes of autonomous cognition and decision-making behavior of the unmanned surface vehicle is obtained from training data of different scenarios.

[0033] Optionally, the method further comprises:

[0034] The unstable region is determined as the region with dense contour lines. In the unstable region, the absolute value of the gradient of y=f(x) is large, that is, the function value y changes rapidly with x, and a small change in x will have a large impact on the value of y;

[0035] The stable region is determined as the region with sparse contour lines. In the stable region, the function value y changes slowly with x, and a small change in x will not have a big impact on the value of y;

[0036] If the input data falls into the unstable region of the neural network model, then the model is easily fooled by adversarial samples at this input data;

[0037] If the input data falls in the stable region of the neural network model, then the model is not easily fooled by adversarial samples at this input data.

[0038] According to a second aspect of the present invention, there is provided a device for analyzing failure modes of autonomous cognition and decision-making behavior of an unmanned boat, comprising:

[0039] An acquisition module is used to acquire data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode;

[0040] The analysis module is used to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary by taking a method of first dividing and then summarizing;

[0041] A determination module is used to analyze the failure of the autonomous cognition and decision-making behavior according to the coupling information.

[0042] Optionally, the structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

[0043] Optionally, the analysis module is used to:

[0044] According to the preset output requirements, select the highest level of the system;

[0045] The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience;

[0046] Perform system maintenance and repair based on lower system levels.

[0047] In a third aspect, the present application shows an electronic device, which includes: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the method described in any of the above aspects.

[0048] In a fourth aspect, the present application illustrates a non-temporary computer-readable storage medium, which, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to execute a method as described in any of the above aspects.

[0049] In a fifth aspect, the present application illustrates a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the method described in any of the above aspects.

[0050] The beneficial effects brought by the present invention are as follows:

[0051] It can be seen from the above scheme that the embodiment of the present invention provides an analysis method and device for failure modes of autonomous cognition and decision-making behavior of unmanned boats, including: obtaining data on failure modes of autonomous cognition and decision-making behavior of surface unmanned boats, wherein the data include potential failure pathways, failure modes of components and their failures, and causes of each failure mode; adopting a method of first dividing and then generalizing to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary; analyzing the failure of autonomous cognition and decision-making behavior based on the coupling information, and taking the construction of polymorphic sample space of surface WR system and typical failure mechanism and robust representation of autonomous cognitive ability as scientific problems, which can effectively solve the dilemma of poor generalization ability of natural scenes, unclear failure mechanism, and lack of evaluation system faced by autonomous systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic flow chart of a method for analyzing failure modes of autonomous cognition and decision-making behavior of an unmanned boat provided according to an embodiment;

[0053] Figure 2 A schematic flow chart of another method for analyzing failure modes of autonomous cognition and decision-making behavior of an unmanned boat provided according to an embodiment;

[0054] Figure 3 A schematic diagram of a decision boundary of a recognition model provided according to an embodiment;

[0055] Figure 4 This is a structural block diagram of an analysis device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat of the present application.

[0056] Figure 5 It is a block diagram of an electronic device of the present application.

[0057] Figure 6 It is a block diagram of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] In order to solve the dilemma faced by autonomous systems, such as poor generalization ability of natural scenes, unclear failure mechanisms, and lack of evaluation systems, it is proposed to construct a polymorphic sample space of surface WR systems and typical failure mechanisms and robust characterization of autonomous cognitive capabilities as scientific issues. It is planned to test the capabilities of surface WR systems by constructing various extreme conditions that may be encountered in natural working conditions, and to explore their failure mechanisms through a large number of failure cases of other autonomous systems combined with the test results of the surface WR system, and to screen an evaluation index system that is sensitive to the failure mechanism to form a robust characterization capability of the surface WR system. Specific research ideas are as follows: Figure 2 shown.

[0060] Reference Figure 1 , shows a flowchart of the steps of an analysis method of failure modes of autonomous cognition and decision-making behavior of an unmanned boat of the present application, which can be applied to electronic equipment, wherein the method can specifically include the following steps:

[0061] S101. Obtain data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode;

[0062] S102. Adopt the method of first dividing and then summarizing to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary;

[0063] S103. Analyze the failure of autonomous cognition and decision-making behavior based on the coupling information.

[0064] Generally speaking, there are many implementation methods and forms of expression for failure mode and effect analysis (FMEA), which is usually achieved by identifying failure modes, related failure causes, and direct and final effects of failure. The analysis results are presented in a worksheet, the core content of which includes basic information and detailed information of the entire system. The table gives the potential failure paths of the system, components and their failure modes that may cause system failure, and the causes of each failure mode.

[0065] When performing FMEA, the following principles apply: determine the focus of work, the more serious the impact of the algorithm failure mode, and the more novel the algorithm design, the more emphasis needs to be placed on the analysis. The analysis of intelligent systems is a very large and complex task, and it is impossible to cover everything. Therefore, it is necessary to consider the aspects with the greatest impact after failure and the aspects that are most likely to fail. Otherwise, analysts may focus on some failure modes that have no or little impact on the system.

[0066] The more cutting-edge algorithms are, the less reliability and safety analysis they have undergone, so they should be analyzed more carefully. If a relatively mature and stable algorithm is applied to a new scenario, it may be necessary to review the previously completed FMEA. For new working conditions, different environments or working stresses, a completely new FMEA may also need to be established.

[0067] When analyzing, the method of first dividing and then summarizing should be adopted, while paying attention to analyzing the possible coupling between different failure modes. The consequences on the operation, function or status of the system are called the failure effects of the failure mode. When determining the final impact, the impact of the failure on the highest level of the system should be evaluated and defined through all intermediate levels of analysis. A failure effect may be caused by one or more failure modes of a product or multiple products, and the final impact described may be the consequence of multiple failures (for example: a catastrophic final impact caused by the failure of a safety device, which only occurs when the safety device fails and the main function of the safety device exceeds the allowable limit). These final effects caused by multiple failures should be indicated in the worksheet.

[0068] Another embodiment of the present application further supplements the analysis method of the failure mode of autonomous cognition and decision-making behavior of the unmanned boat provided in the above embodiment.

[0069] Optionally, the structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

[0070] Optionally, a method of first dividing and then summarizing is adopted to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary, including:

[0071] According to the preset output requirements, select the highest level of the system;

[0072] The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience;

[0073] Perform system maintenance and repair based on lower system levels.

[0074] For the system analysis in the FMEA process, we can start from the structural information, hierarchy, and boundary.

[0075] The system structure information shall include the following:

[0076] 1) Different elements of the system and their characteristics, performance, role and functions;

[0077] 2) The logical relationship between the elements;

[0078] 3) Redundancy level and its nature;

[0079] 4) The location and importance of the system in the entire equipment (if possible);

[0080] 5) System input and output;

[0081] 6) Changes in system structure under different working modes.

[0082] At all levels of system consideration, including the highest level, information about functions, features, and performance is required so that FMEA can appropriately address failure modes that prevent these functions from being performed. The system boundary constitutes the physical and functional interface between the system and the environment, including other systems with which the system interacts. The system boundaries defined for analysis should correspond to the boundaries defined for system design and maintenance, and this principle should be applied at all levels of the system. Systems and / or components that are outside the boundaries should be identified for exclusion from analysis. It should be ensured that other systems and components outside the system boundaries are not omitted, and their exclusion from detailed study should be clearly specified or explained.

[0083] The following guidelines can be used when conducting hierarchical analysis:

[0084] 1) Select the highest level of the system based on the design concept and specified output requirements;

[0085] 2) The lowest level of information of the system being analyzed is most useful for determining the definition and description of the function. The selection of the appropriate system level is influenced by previous experience. For systems with mature designs and good reliability, maintainability and safety records, less detailed analysis is reasonable; on the contrary, any newly designed system with unknown reliability history requires a more detailed analysis and specifies a lower system level.

[0086] 3) Specified or predetermined maintenance and repair levels are valuable guidance in determining lower system levels.

[0087] In FMEA, the definition of failure modes, failure causes and failure effects depends on the level of analysis and the system failure criteria. As the analysis goes deeper, the failure effects defined at a lower level may be the failure modes at a higher level, and the failure modes at a lower level may be the failure causes at a higher level, and so on.

[0088] Intelligent systems often use artificial intelligence problem-solving models to obtain results. Compared with the problem-solving models used by traditional systems, intelligent systems have three obvious characteristics: their problem-solving algorithms are often non-deterministic or heuristic. Their problem-solving relies heavily on knowledge, and the problems of intelligent systems often have exponential computational complexity.

[0089] It is generally believed that intelligent systems contain four elements, namely data, computing power, algorithms, and scenarios. The intelligence of intelligent systems is contained in the data used for training. Computing power provides basic computing power support for intelligent systems. Algorithms are the fundamental way to realize artificial intelligence and an effective way to mine data intelligence. Data, computing power, and algorithms are inputs, and only when they are output in actual scenarios can their actual value be reflected. There are many ways to classify the levels of intelligent systems. On the one hand, we can refer to the division of computer systems and slightly improve and simplify them into hardware layers, software layers including operating systems and related auxiliary software, algorithm layers that implement intelligent algorithms of intelligent systems, and data layers for training intelligent algorithms. It can also be divided into data collection layers, recognition processing layers, decision layers, execution layers, etc. according to the workflow of intelligent systems.

[0090] According to the analysis of intelligent systems in the previous article, intelligent systems have great uncertainty and are quite complex. Therefore, it is difficult to derive potential failure modes by analyzing the specific structure of intelligent systems. We can only find out the links with higher failure probability based on past research. A two-pronged approach can be adopted to simulate the working environment of the system and test the working conditions in different environments. On the other hand, we can analyze the parts with low reliability separately and give possible failure modes and probabilities by referring to previous research results.

[0091] Optionally, the data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle is obtained from training data of different scenarios.

[0092] Optionally, the method further comprises:

[0093] The unstable region is determined as the region with dense contour lines. In the unstable region, the absolute value of the gradient of y=f(x) is large, that is, the function value y changes rapidly with x, and a small change in x will have a large impact on the value of y;

[0094] The stable region is determined as the region with sparse contour lines. In the stable region, the function value y changes slowly with x, and a small change in x will not have a big impact on the value of y;

[0095] If the input data falls into the unstable region of the neural network model, then the model is easily fooled by adversarial samples at this input data;

[0096] If the input data falls in the stable region of the neural network model, then the model is not easily fooled by adversarial samples at this input data.

[0097] Mechanism 1: The source of knowledge for intelligent systems is derived from training data. Any scenario not included in the training data set is likely to fail. Therefore, it is necessary to organize different scenario elements reasonably and test the model with as many scenarios as possible. In order to accurately describe the scenario, it is necessary to analyze the various elements in the scenario and quantitatively evaluate the comprehensiveness of the test dataset.

[0098] According to the definition above, the failure of intelligent systems can be divided into two categories: identification failure and decision failure. Identification failure means that the sample cannot be classified into the correct category, and decision failure means that the decision is not in line with expectations. For the core intelligent algorithm of the intelligent system, the failure causes can be divided into two categories, namely underfitting and overfitting. "Underfitting" often occurs when the algorithm is relatively simple. At this time, the model cannot learn the "general rules" in the data set due to insufficient learning ability, resulting in weak generalization ability. On the contrary, "overfitting" often occurs when the algorithm has too strong learning ability or too many training rounds. At this time, the model learns too many features from the data, so that it captures the characteristics of the individual samples in the training set and regards them as "general rules". This situation will also lead to a decrease in the generalization ability of the model.

[0099] Mechanism 2: It is generally believed that deep learning can easily form nonlinear decision boundaries. When a point is close to a linear boundary, even a small amount of noise can push it to the other side of the decision boundary. It can also be seen from the figure below that in practice, deep learning classifiers have a very linear response.

[0100] In response to the robustness problem of intelligent algorithms, scholars have developed the field of adversarial attacks. Whether it is logistic regression, softmax regression, SVM, decision tree, nearest neighbor or deep learning model, no ML algorithm is immune to adversarial attacks. The occurrence of adversarial examples is caused by excessive linearity in the system.

[0101] like Figure 3 The left figure shows a linear binary classification problem. The difference between the two categories is large, so it is easy to find the classification boundary between the two categories. However, for practical problems, the distribution of the data to be classified is often unknown. At the same time, the complexity of the actual problem, noise interference and other factors lead to unclear distinctions between different categories of data in the data set to be classified. The following discusses the binary classification problems of several typical data sets.

[0102] Unstable region: An area with dense contour lines. In an unstable region, the absolute value of the gradient of y=f(x) is large, that is, the function value y changes rapidly with x, and a small change in x will have a large impact on the value of y.

[0103] Stable region: An area with sparse contour lines. In a stable region, the function value y changes slowly with x, and a small change in x will not have a big impact on the value of y.

[0104] If the input data falls in the unstable region of the neural network model, the model is easily fooled by adversarial samples at this input data. If the input data falls in the stable region of the neural network model, the model is not easily fooled by adversarial samples at this input data. This explains why in actual neural network models, such as neural networks for image recognition, some input images can cause the model to misclassify after slight changes, while other images can still cause the model to output correct classification results even after major changes.

[0105] In addition, from the definition of the gradient, we can know that the gradient vector is orthogonal to the contour line. The function value changes fastest along the direction of the gradient, while the function value does not change along the direction of the contour line. Therefore, for input data x that falls in the unstable area, its sensitivity to the disturbance Δx depends on the angle between Δx and the gradient vector (or contour line). If Δx is along the gradient direction, then a small ‖Δx‖ will cause a large change in the output y of the model function. If Δx is along the contour line direction, then even if ‖Δx‖ is large, the output y of the function will not change. This explains why in actual neural network models, such as neural networks used for image recognition, some images will only cause classification errors after specific disturbances, but not any disturbance to the image will cause classification errors.

[0106] Mechanism 3: The background noise caused by hardware degradation in special cases makes the intelligent model invalid, and the fragility of model parameters poses a huge challenge to the stability and generalization research of the model. Hardware degradation can be regarded as parameter damage, so field experts proposed a parameter perturbation index to measure the stability of neural network parameters. This index describes the maximum loss change in the non-trivial worst case under parameter perturbation. The gradient-based perturbation focuses on the stability of deep neural networks.

[0107] In this loss function example, the traditional optimizer prefers b with lower loss over A because b has lower loss. But the parameters of B are more easily corrupted.

[0108]

[0109] Gradient-based parameter perturbation can be expressed as follows:

[0110]

[0111] Based on the above analysis of the failure mechanism and the investigation of the reasons for the failure of the intelligent algorithm, it can be seen that the reasons for the failure of the intelligent algorithm can be summarized into the following categories:

[0112] (1) Insufficient or inaccurate data: Intelligent algorithms require a large amount of high-quality data for training and prediction. If the data is insufficient or the data quality is poor, the algorithm will not be able to accurately predict or will produce wrong results.

[0113] (2) Data bias: If there are more or fewer certain types of data in a dataset than other types of data, the algorithm may tend to make predictions based on the larger amount of data and ignore the smaller amount of data, causing the algorithm to fail.

[0114] (3) Concept drift: When the data distribution changes, the originally trained model may not be able to adapt to the new data distribution, resulting in a decrease in prediction accuracy.

[0115] (4) Improper feature selection: Selecting appropriate features is crucial to the performance of the algorithm. If irrelevant or redundant features are selected, the algorithm will fail.

[0116] (5) Overfitting and underfitting: Overfitting means that the model performs well on the training set but performs poorly on the test set or new data; underfitting means that the model cannot capture the complex relationships in the data and has poor predictive ability. Both of these situations will cause the algorithm to fail.

[0117] (6) Improper parameter selection: There are many parameters that need to be set in intelligent algorithms. Improper selection may cause the algorithm to fail. The selection of parameters usually needs to be determined through experiments and tuning.

[0118] (7) Improper algorithm selection: Different problems require different algorithms. Improperly selected algorithms may not be able to effectively solve the problem, resulting in algorithm failure.

[0119] (8) Algorithm design errors: The design of the algorithm may contain logical or implementation errors, which may cause the algorithm to fail to work properly or produce incorrect results.

[0120] According to the above summary and conclusion of the failure mechanism, it can be found that the failure of intelligent algorithms is partly caused by external sample factors and partly by internal parameter factors. Therefore, the concept of JSZNAQ credibility is refined into robustness, reliability and interpretability, which respectively characterize the impact of different external samples on the model, the physical meaning of the model and the model's internal parameter disturbances on the model.

[0121] The severity of the failure effect of the system can be evaluated by severity. The severity classification is highly dependent on the application of FMEA and is given after considering the following factors:

[0122] 1) The properties of the system and the impact of failure on the product users or the environment;

[0123] 2) functional characteristics of the system or process;

[0124] 3) All contractual requirements raised by the customer

[0125] 4) Government or industrial safety requirements

[0126] 5) Requirements contained in the guarantee

[0127] Severity Level:

[0128] Category I - catastrophic, causing death or system damage;

[0129] Category II - fatal, causing serious injury to personnel, significant economic losses, or serious damage to systems leading to mission failure;

[0130] Category III - Critical, minor damage to systems causing delay or degradation of minor damage to personnel;

[0131] Category IV - A minor failure that is not serious enough to cause personal injury, certain economic losses or system damage, but it will lead to unplanned maintenance.

[0132] Severity Analysis:

[0133] Grade A is frequent occurrence, K>20%;

[0134] Grade B occurs occasionally, 10%<K<20%;

[0135] Grade C is accidental, 1%<K<10%;

[0136] Grade D is a rare occurrence, 0.1%<K<1%;

[0137] Grade E is very rare, K < 0.1%

[0138] The embodiment of the present invention provides an analysis method and device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat, including: obtaining data on failure modes of autonomous cognition and decision-making behavior of a surface unmanned boat, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode; adopting a method of first dividing and then generalizing to analyze coupling information between different failure modes from structural information, hierarchy, and boundary; analyzing failures of autonomous cognition and decision-making behavior based on the coupling information, and taking the construction of a polymorphic sample space of a surface WR system and typical failure mechanisms and robust representations of autonomous cognitive capabilities as scientific issues, which can effectively solve the dilemma of poor generalization capabilities of natural scenarios, unclear failure mechanisms, and lack of evaluation systems faced by autonomous systems.

[0139] It should be noted that each implementable method in this embodiment may be implemented separately, or may be implemented in combination in any manner without conflict, and this application is not limited thereto.

[0140] Another embodiment of the present application provides an analysis device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat, which is used to execute the analysis method for failure modes of autonomous cognition and decision-making behavior of an unmanned boat provided in the above embodiment.

[0141] like Figure 4 , which is a schematic diagram of the structure of the analysis device for the failure mode of autonomous cognition and decision-making behavior of an unmanned boat provided in an embodiment of the present application. The analysis device for the failure mode of autonomous cognition and decision-making behavior of an unmanned boat includes:

[0142] The acquisition module 401 is used to obtain data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle, wherein the data includes potential failure paths, failure modes of components and their failures, and causes of each failure mode;

[0143] The analysis module 402 is used to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary by adopting a method of first dividing and then summarizing;

[0144] The determination module 403 is used to analyze the failure of autonomous cognition and decision-making behavior according to the coupling information.

[0145] Regarding the device in this embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0146] Another embodiment of the present application further supplements the analysis device for the failure mode of autonomous cognition and decision-making behavior of the unmanned boat provided in the above embodiment.

[0147] Optionally, the structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

[0148] Optionally, the analysis module is used to:

[0149] According to the preset output requirements, select the highest level of the system;

[0150] The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience;

[0151] Perform system maintenance and repair based on lower system levels.

[0152] The embodiment of the present invention provides an analysis method and device for failure modes of autonomous cognition and decision-making behavior of an unmanned boat, including: obtaining data on failure modes of autonomous cognition and decision-making behavior of a surface unmanned boat, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode; adopting a method of first dividing and then generalizing to analyze coupling information between different failure modes from structural information, hierarchy, and boundary; analyzing the failure of the autonomous cognition and decision-making behavior based on the coupling information, and taking the construction of a polymorphic sample space of a surface WR system and typical failure mechanisms and robust representations of autonomous cognitive capabilities as scientific issues, which can effectively solve the dilemma of poor generalization capabilities of natural scenarios, unclear failure mechanisms, and lack of an evaluation system faced by autonomous systems.

[0153] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0154] Optionally, an embodiment of the present application further provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0155] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0156] Figure 5 800 is a block diagram of an electronic device 800 shown in the present application. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0157] Reference Figure 5 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0158] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0159] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, images, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0160] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0161] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0162] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0163] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0164] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0165] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast operation information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0166] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0167] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by a processor 820 of an electronic device 800 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0168] Figure 6 19 is a block diagram of a computer-readable storage medium 1900 shown in the present application. For example, the computer-readable storage medium 1900 may be provided as a server.

[0169] Reference Figure 6 , the computer-readable storage medium 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions, such as an application, that can be executed by the processing component 1922. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0170] The computer readable storage medium 1900 may also include a power supply component 1926 configured to perform power management of the computer readable storage medium 1900, a wired or wireless network interface 1950 configured to connect the computer readable storage medium 1900 to a network, and an input / output (I / O) interface 1958. The computer readable storage medium 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0171] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0172] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0173] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0174] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0175] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0176] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0177] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0178] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0179] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0180] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0181] The above are preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats, characterized in that: include: Obtaining data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode; Adopt the method of first dividing and then summarizing to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary; The failure of the autonomous cognition and decision-making behavior is analyzed based on the coupling information.

2. The method for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 1 is characterized in that: The structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

3. The method for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 1 is characterized in that: The method of first dividing and then summarizing is adopted to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary, including: According to the preset output requirements, select the highest level of the system; The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience; Perform system maintenance and repair based on lower system levels.

4. The method for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 1 is characterized in that: The data on the failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle are obtained from training data of different scenarios.

5. The method for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 1 is characterized in that: The method further comprises: The unstable region is determined as the region with dense contour lines. In the unstable region, the absolute value of the gradient of y=f(x) is large, that is, the function value y changes rapidly with x, and a small change in x will have a large impact on the value of y; The stable region is determined as the region with sparse contour lines. In the stable region, the function value y changes slowly with x, and a small change in x will not have a big impact on the value of y; If the input data falls into the unstable region of the neural network model, then the model is easily fooled by adversarial samples at this input data; If the input data falls in the stable region of the neural network model, then the model is not easily fooled by adversarial samples at this input data.

6. An analysis device for failure modes of autonomous cognition and decision-making behavior of unmanned boats, characterized in that: include: An acquisition module is used to acquire data on failure modes of autonomous cognition and decision-making behavior of the surface unmanned vehicle, wherein the data includes potential failure pathways, failure modes of components and their failures, and causes of each failure mode; The analysis module is used to analyze the coupling information between different failure modes from the perspective of structural information, hierarchy, and boundary by taking a method of first dividing and then summarizing; A determination module is used to analyze the failure of the autonomous cognition and decision-making behavior according to the coupling information.

7. The device for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 6 is characterized in that: The structural information includes at least: different elements and their characteristics, performance, role and function; logical relationship between elements; redundancy level and its nature; position and importance in the whole equipment; input and output; changes in system structure under different working modes.

8. The device for analyzing failure modes of autonomous cognition and decision-making behavior of unmanned boats according to claim 6 is characterized in that: The analysis module is used to: According to the preset output requirements, select the highest level of the system; The lowest level of information about the system being analyzed is most useful for determining the definition and description of functionality, and the choice of the appropriate system level is influenced by prior experience; Perform system maintenance and repair based on lower system levels.

9. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the method according to any one of claims 1 to 5 when executed by the processor.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images