Fault diagnosis method and device for humanoid robot and storage medium

Abstract

The invention provides a fault diagnosis method and device of a humanoid robot and a storage medium, and relates to the field of humanoid robots, and the fault diagnosis method of the humanoid robot comprises the steps: obtaining the current fault information of the humanoid robot; according to the current fault information, performing multiple times of event search in a pre-constructed fault tree, after each time of event search, predicting the occurrence probability of the searched fault event under the current fault symptom according to the historical fault information, and obtaining the predicted occurrence probability of the searched fault event, the current fault symptom is determined according to the current fault information; according to the predicted occurrence probability of the searched fault event, fault diagnosis information of the humanoid robot is generated, and the fault diagnosis information is used for assisting fault processing of the humanoid robot; wherein the fault tree is constructed according to fault event information related to the humanoid robot. According to the invention, the accuracy of fault diagnosis of the humanoid robot can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of humanoid robots, and in particular to a fault diagnosis method, device and storage medium for a humanoid robot. Background Art

[0002] With the continuous development of humanoid robot technology and the continuous expansion of application scenarios, its structure has become more complex and its operating environment has become more diverse, which has brought severe challenges to the fault diagnosis of humanoid robots.

[0003] In the related art, fault diagnosis of the humanoid robot is performed by judging the sensor threshold of the humanoid robot. When the sensor value of the humanoid robot is greater than the sensor threshold, it is determined that a component related to the sensor value has failed.

[0004] However, the fault diagnosis accuracy of the above method is low. Summary of the invention

[0005] The present application provides a humanoid robot fault diagnosis method, device and storage medium to solve the problem of low accuracy of humanoid robot fault diagnosis.

[0006] In a first aspect, the present application provides a fault diagnosis method for a humanoid robot, comprising: obtaining current fault information of the humanoid robot; based on the current fault information, performing multiple event searches in a pre-constructed fault tree related to the humanoid robot, and after each event search, predicting the probability of occurrence of the searched fault event under the current fault symptom to obtain a predicted probability of occurrence of the searched fault event, wherein the current fault symptom is determined based on the current fault information; based on the predicted probability of occurrence of the searched fault event, generating fault diagnosis information of the humanoid robot, the fault diagnosis information being used to assist fault handling of the humanoid robot; wherein the fault tree is constructed based on fault event information related to the humanoid robot, the fault event information includes multiple fault events and event causal relationships between the multiple fault events, and the distribution of the multiple fault events in the fault tree is determined based on the event impact degrees and event causal relationships corresponding to the multiple fault events.

[0007] In some embodiments, the probability of occurrence of the searched fault event under the current fault symptom is predicted to obtain the predicted probability of occurrence of the searched fault event, including: obtaining reference probability information, the reference probability information including the prior probability of the searched fault event, the probability of occurrence of the current fault symptom when the searched fault event occurs, the prior probabilities corresponding to multiple fault events in the fault tree, and the probability of occurrence of the current fault symptom when multiple fault events occur respectively; based on the reference probability information, the probability of occurrence of the searched fault event under the current fault symptom is predicted by Bayesian theorem to obtain the predicted probability of occurrence.

[0008] In some embodiments, multiple event searches are performed in a pre-built fault tree related to the humanoid robot based on current fault information, including: based on current fault information, multiple event searches are performed in the fault tree through a set search strategy, and the search strategy includes a depth-first search strategy.

[0009] In some embodiments, multiple fault events in a fault tree are distributed according to a hierarchical structure of top events, intermediate events, and sub-events, and the i-th search of multiple event searches includes: when i is equal to 1, based on the current fault information and a depth-first search strategy, the search is started from the top event of the fault tree to obtain the fault event obtained by the first search; when i is greater than 1, based on the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, the starting node of the i-th search is determined, and based on the current fault information, the search is started from the starting node to obtain the fault event corresponding to the i-th search.

[0010] In some embodiments, the starting node of the i-th search is determined based on the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, including: if the predicted probability of occurrence of the fault event searched for the i-1th time is greater than or equal to the probability threshold, then the starting node is determined to be the next layer of intermediate event or the next layer of sub-event connected to the fault event searched for the i-1th time, otherwise the starting node is determined according to the depth-first search strategy.

[0011] In some embodiments, fault diagnosis information of a humanoid robot is generated based on the predicted probability of occurrence of a searched fault event, including: obtaining the probability of occurrence of an associated fault symptom of the searched fault event when the searched fault event occurs; determining the information entropy of the searched fault event based on the probability of occurrence of the associated fault symptoms of the searched fault event and the number of associated fault symptoms of the searched fault event, the information entropy being used to indicate the degree of uncertainty of the searched fault event; and generating fault diagnosis information based on the predicted probability of occurrence of the searched fault event and the information entropy of the searched fault event.

[0012] In some embodiments, the fault tree construction process includes: obtaining fault event information, the fault event information also includes event impact factors corresponding to multiple fault events, the event impact factors corresponding to multiple fault events respectively represent the event impact degrees corresponding to the multiple fault events respectively; determining the top event in the fault tree among multiple fault events according to the event impact factors corresponding to the multiple fault events respectively; determining the intermediate event corresponding to the top event, the logical relationship between the top event and the intermediate event, the sub-event corresponding to the intermediate event, and the logical relationship between the intermediate event and the sub-event among multiple fault events according to the event causal relationship between the multiple fault events; drawing the nodes corresponding to the top event, the nodes corresponding to the intermediate event, the nodes corresponding to the sub-event, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate event, the edges between the nodes corresponding to the intermediate event and the nodes corresponding to the sub-event, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate event, and the logic gate symbols corresponding to the logical relationship between the intermediate event and the sub-event, to construct a fault tree.

[0013] In some embodiments, the fault event information also includes event attribute information corresponding to multiple fault events and / or the mutual influence between multiple fault events. The nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events are drawn. After the fault tree is constructed, it also includes: according to the event attribute information corresponding to the multiple fault events, corresponding event attribute information is added to the multiple nodes in the fault tree, and the event attribute information includes at least one of the following: fault description content, related fault symptoms or related fault detection methods; and / or, according to the mutual influence between the multiple fault events, relationship annotation information is added to the multiple nodes in the fault tree, and the relationship annotation information is used to indicate the mutual influence between the fault events corresponding to the multiple nodes.

[0014] In a second aspect, the present application provides a fault diagnosis device for a humanoid robot, comprising: an acquisition module for acquiring current fault information of the humanoid robot; a probability prediction module for performing multiple event searches in a pre-constructed fault tree related to the humanoid robot based on the current fault information, and after each event search, predicting the probability of occurrence of the searched fault event under the current fault symptom to obtain a predicted probability of occurrence of the searched fault event, wherein the current fault symptom is determined based on the current fault information; a diagnosis information generation module for generating fault diagnosis information of the humanoid robot based on the predicted probability of occurrence of the searched fault event, and the fault diagnosis information is used to assist fault handling of the humanoid robot; wherein the fault tree is constructed based on fault event information related to the humanoid robot, the fault event information includes multiple fault events and event causal relationships between the multiple fault events, and the distribution of the multiple fault events in the fault tree is determined based on the event impact degrees and event causal relationships corresponding to the multiple fault events.

[0015] In a third aspect, the present application provides an electronic device comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the fault diagnosis method for a humanoid robot as described in the first aspect of the present application.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed by a processor, the fault diagnosis method for a humanoid robot as described in the first aspect of the present application is implemented.

[0017] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the fault diagnosis method for a humanoid robot as described in the first aspect of the present application.

[0018] The humanoid robot fault diagnosis method, device and storage medium provided in the present application perform fault event search based on the current fault information of the humanoid robot and the fault tree related to the humanoid robot, thereby improving the accuracy of the fault event search; after each search, the probability of occurrence of the searched fault event under the current fault symptoms is predicted, and fault diagnosis information is generated based on the predicted probability of occurrence of the searched fault event, thereby improving the accuracy of the fault diagnosis information. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0020] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0021] Figure 2 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 1 ;

[0022] Figure 3 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 2 ;

[0023] Figure 4 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 3 ;

[0024] Figure 5 A schematic diagram of a process of constructing a fault tree according to an embodiment of the present application;

[0025] Figure 6 A schematic diagram of the structure of a fault diagnosis device for a humanoid robot provided in an embodiment of the present application;

[0026] Figure 7 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] In the related art, the fault diagnosis of humanoid robots mainly includes the following two methods:

[0029] Method 1: sensor threshold judgment. For example, when the temperature of the motor of the humanoid robot exceeds a preset fixed threshold, it is determined that the motor of the humanoid robot is faulty.

[0030] However, the judgment basis of this method is too single and has the following disadvantages: on the one hand, it is easy to make misjudgments. For example, the motor temperature of the humanoid robot exceeds the fixed threshold due to long-term high-load operation, which may cause a substantial fault inside the motor; on the other hand, for complex fault combinations (for example, the mechanical structure wear of the humanoid robot and the drift of the electrical system parameters coexist), relying on a single sensor threshold cannot make effective judgments.

[0031] Method 2 is empirical troubleshooting, whereby maintenance personnel rely on similar failure situations encountered in the past and relevant theoretical knowledge to gradually troubleshoot the parts of the humanoid robot that may fail.

[0032] However, this method is highly dependent on the personal experience of maintenance personnel and has the following disadvantages: on the one hand, the experience levels of different maintenance personnel vary greatly, resulting in large differences in the accuracy and efficiency of humanoid robot fault repair; on the other hand, with the rapid development of humanoid robot technology, new types of faults continue to emerge, and maintenance personnel's previous experience makes it difficult to deal with unknown fault situations.

[0033] To solve the above problems, the embodiments of the present application provide a method, device and storage medium for fault diagnosis of a humanoid robot. In the method for fault diagnosis of a humanoid robot, a search for fault events of the humanoid robot is performed based on the current fault information of the humanoid robot and the fault tree related to the humanoid robot, thereby improving the efficiency and accuracy of the search for fault events; after each search, the probability of occurrence of the searched fault event under the current fault symptom is predicted, and fault diagnosis information is generated based on the predicted probability of occurrence of the searched fault event, thereby improving the accuracy of the fault diagnosis information. Therefore, the efficiency and accuracy of the fault diagnosis of the humanoid robot are improved. Compared with the sensor threshold judgment and empirical troubleshooting, the fault tree related to the humanoid robot collects a large amount of fault information of the humanoid robot, which can cope with the diagnosis of complex faults and unknown faults of the humanoid robot to a certain extent.

[0034] Among them, the technical means adopted by the devices, equipment and storage media and the technical effects produced can refer to the technical means adopted by the above-mentioned scheme and the technical effects produced, and will not be repeated here.

[0035] Figure 1 This is a schematic diagram of an application scenario provided by an embodiment of the present application. Figure 1 As shown, Figure 1 This is an example diagram of an application scenario applicable to the embodiments of this application. Figure 1 As shown, the application scenario includes a humanoid robot 101 , a fault information collection device 102 and a fault diagnosis device 103 .

[0036] The fault information collection device 102 can collect the current fault information of the humanoid robot 101 and send the current fault information to the fault diagnosis device 103; the fault diagnosis device 103 stores a fault tree related to the humanoid robot, and executes the following method embodiment based on the current fault information and the fault tree to realize fault diagnosis of the humanoid robot 101.

[0037] The fault information collecting device 102 can also collect fault information of other humanoid robots, and send the fault information of multiple humanoid robots to the fault diagnosis device 103. The fault diagnosis device 103 constructs a fault tree based on the fault information of multiple humanoid robots.

[0038] The fault information collection device 102 may be a sensor and / or monitoring system related to the humanoid robot 101. When the fault information collection device 102 is a monitoring system, the fault information collection device 102 may be a terminal or a server. Figure 1 For example, the fault information collection device 102 is a server. The fault information collection device 102 can be deployed on the humanoid robot 101 , or can be independent of the humanoid robot 101 and connected to the humanoid robot 101 .

[0039] The fault diagnosis device 103 may be an electronic device, including a server or a terminal, and the terminal may be a personal digital assistant (PDA), a handheld device with wireless communication function (such as a smart phone, a tablet computer), a computing device (such as a personal computer (PC)), a wearable device (such as a smart watch, a smart bracelet), and a smart home device (such as a smart speaker, a smart display device), etc. The server may be an independent server or a server cluster, and may be a local server or a cloud server. Figure 1 Take the fault diagnosis device 103 as an example, which is a server.

[0040] It should be noted that Figure 1 This is only a schematic diagram of an application scenario provided by the embodiment of the present application. Figure 1 The equipment included is limited.

[0041] Next, a fault diagnosis method for a humanoid robot is introduced through a specific embodiment.

[0042] Figure 2 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 1 .like Figure 2 As shown, the method of the embodiment of the present application includes:

[0043] S201, obtaining current fault information of the humanoid robot.

[0044] The current fault information of the humanoid robot may be collected by a sensor device and / or a monitoring system, and the current fault information refers to the fault information at the current time when the humanoid robot fails.

[0045] In this embodiment, when a humanoid robot fails, fault information of the humanoid robot can be collected to obtain current fault information. Alternatively, when a humanoid robot fails, the current fault information of the humanoid robot can be obtained from a real-time database of the humanoid robot, wherein the real-time database contains real-time data collected by a sensor device and / or a monitoring system; or, when a humanoid robot fails, the current fault information sent by a collection device (such as a sensor device and / or a monitoring system) of the humanoid robot is received.

[0046] Optionally, the current fault information of the humanoid robot is obtained by collecting data on relevant parts of the humanoid robot according to the current fault condition of the humanoid robot, thereby collecting fault information in a targeted manner and improving the accuracy of information collection.

[0047] In one example, the current fault condition of the humanoid robot is that the humanoid robot completely loses the ability to move, and the current fault information may include at least one of the following items of the humanoid robot: voltage data of the power system, current data of the power system, running state data of the motor (such as the speed of the motor, torque feedback data), output instructions of the control system, feedback signals of the control system, or posture information of the humanoid robot. The posture information can be used to determine whether the humanoid robot is mechanically stuck.

[0048] In another example, the current fault condition of the humanoid robot is that the humanoid robot falls frequently, and the current fault information may include at least one of the following items of the humanoid robot: measurement data of a balance sensor (such as a measurement value of a gyroscope and / or a measurement value of an accelerometer), parameter information of a gait planning module, output data of a gait planning module, motion data of a leg joint (such as an angle, angular velocity and / or force data of a leg joint), or image data of a visual sensor. The image data of the visual sensor can be used to determine whether the humanoid robot falls due to a visual perception error.

[0049] S202, based on the current fault information, multiple event searches are performed in a pre-built fault tree related to the humanoid robot. After each event search, the probability of occurrence of the searched fault event under the current fault symptom is predicted to obtain the predicted probability of occurrence of the searched fault event. The current fault symptom is determined based on the current fault information.

[0050] The fault tree is constructed based on the fault event information related to the humanoid robot, and includes multiple fault events related to the humanoid robot and the event causal relationship between the multiple fault events. Specifically, in the fault tree, multiple nodes represent multiple fault events, and the edge between two nodes represents the event causal relationship between the fault events corresponding to the two nodes.

[0051] Among them, the distribution of multiple fault events in the fault event information in the fault tree is determined according to the event impact degrees corresponding to the multiple fault events and the event causal relationship between the multiple fault events. In the fault tree, multiple fault events are distributed according to the hierarchy of top events, intermediate events corresponding to the top events, and sub-events corresponding to the intermediate events. There is a causal relationship between the top event and the intermediate event corresponding to the top event, the top event is the result, and the intermediate event corresponding to the top event is the cause of the top event; there is a causal relationship between the intermediate event and the sub-event corresponding to the intermediate event, the intermediate event is the result, and the sub-event corresponding to the intermediate event is the cause of the intermediate event.

[0052] There can be multiple fault trees, and the top events of different fault trees are different. For example, the top event of one fault tree is that the humanoid robot completely loses its ability to move, and the top event of another fault tree is that the humanoid robot frequently falls. In the event search process, event searches can be performed separately in multiple fault trees.

[0053] The current fault symptom may be extracted from the current fault information. For example, words related to the fault symptom are identified in the current fault information to obtain the current fault symptom. The current fault symptom may include excessive battery temperature, unresponsive control system, etc.

[0054] The probability of a fault event occurring under the current fault symptom refers to the probability of a fault event occurring when the robot exhibits the current fault symptom.

[0055] In this embodiment, before performing an event search, the current fault symptom of the humanoid robot can be determined based on the current fault information of the humanoid robot. The current fault symptom can be one or more symptoms. During an event search, the corresponding fault event can be searched by matching the current fault information with the fault event corresponding to the node in the fault tree. For the searched fault event, the probability of occurrence of the fault event under the current fault symptom is predicted to obtain the predicted probability of occurrence of the searched fault event. In this way, the above event search process is repeatedly executed to perform multiple event searches and probability predictions to obtain the predicted probability of occurrence corresponding to the multiple searched fault events.

[0056] Optionally, the current fault information is preprocessed before multiple event searches, thereby improving the accuracy of the fault event search through the preprocessed current fault information and fault tree.

[0057] Furthermore, the preprocessing of the current fault information includes at least one of the following: data cleaning, data conversion or data feature extraction. Thus, the information quality of the current fault information is improved through one or more operations of data cleaning, data conversion and data feature extraction.

[0058] Among them, data cleaning of the current fault information may include: removing noise and outliers in the current fault information to avoid interference of noise and outliers in the search for fault events and the determination of current fault symptoms, thereby improving the accuracy of fault event search and current fault symptoms.

[0059] The data conversion of the current fault information may include: converting the data format in the current fault information into a target format, and / or normalizing the values ​​involved in the current fault information. Thus, based on the current fault information with a unified format or normalized processing, the accuracy of the fault event search and the current fault symptoms is improved.

[0060] Among them, data feature extraction is performed on the current fault information, and key information reflecting the current fault of the humanoid robot can be extracted from the current fault information, such as fault characteristics reflecting the main manifestations of the humanoid robot's fault and abnormal data changes, thereby reducing the redundancy of the current fault information, improving the accuracy of fault event search and current fault symptoms, and also improving the efficiency of fault event search.

[0061] S203, generating fault diagnosis information of the humanoid robot according to the predicted occurrence probability of the searched fault event, where the fault diagnosis information is used to assist the fault handling of the humanoid robot.

[0062] In one implementation of this step, the fault diagnosis information of the humanoid robot includes multiple fault events and the predicted occurrence probabilities corresponding to the multiple fault events. The multiple fault events and the predicted occurrence probabilities corresponding to the multiple fault events can assist professionals or diagnostic programs in determining the cause of the fault of the humanoid robot, and further assist professionals or repair programs in troubleshooting the humanoid robot. The multiple fault events in the fault diagnosis information are fault events obtained by multiple searches in the fault tree, or the multiple fault events are screened from the fault events obtained by multiple searches based on the predicted occurrence probabilities of the fault events obtained by multiple searches, for example, the fault events whose predicted occurrence probabilities are greater than or equal to the probability threshold are screened as the fault events in the fault diagnosis information.

[0063] In another implementation of this step, a fault decision can be made based on the predicted probability of occurrence of the fault event obtained by the search, and the fault diagnosis information of the humanoid robot can be obtained. The fault diagnosis information includes a fault decision result, and the fault decision result can include the fault event finally determined to have occurred, or the fault decision result can include multiple fault events and the diagnosis information corresponding to the multiple fault events (such as the diagnosis accuracy of the fault event, further diagnosis suggestions, etc.).

[0064] In the embodiment of the present application, the fault tree provides sufficient fault events to comprehensively and accurately search for fault events that match the current fault information of the humanoid robot. These searched fault events are all fault events that may currently occur in the humanoid robot. By predicting the probability of occurrence of the searched fault events under the current fault symptoms of the humanoid robot, the probability of the fault event occurring in the humanoid robot can be more accurately inferred, thereby improving the accuracy of fault diagnosis of the humanoid robot.

[0065] The following provides corresponding embodiments regarding the probability prediction of fault events.

[0066] In some embodiments, after each search, the Bayesian theorem can be used to predict the probability of occurrence of the searched fault event under the current fault symptom to obtain the predicted probability of occurrence of the searched fault event. Thus, the accuracy of predicting the probability of occurrence of the searched fault event under the current fault symptom is improved through the Bayesian theorem.

[0067] In one implementation, reference probability information may be obtained, and based on the reference probability information, the occurrence probability of the searched fault event under the current fault symptom may be predicted using the Bayesian theorem to obtain the predicted occurrence probability of the searched fault event.

[0068] The reference probability information includes: the prior probability of the searched fault event, the probability of the current fault symptom occurring when the searched fault event occurs, the prior probability corresponding to multiple fault events in the fault tree (i.e., the fault events corresponding to multiple nodes in the fault tree), and the probability of the current fault symptom occurring when the multiple fault events occur. The prior probability of the fault event is the probability of the humanoid robot having the fault event without any preconditions.

[0069] Taking the i-th search as an example, the reference probability information includes: the prior probability of the fault event searched for the i-th time, the probability of occurrence of the current fault symptom when the fault event searched for the i-th time occurs, the prior probabilities corresponding to multiple fault events in the fault tree, and the probability of occurrence of the current fault symptom when the multiple fault events occur respectively.

[0070] The reference probability information can be obtained based on statistics of historical failure data and experimental data.

[0071] In this implementation, after each search, the prior probability of the searched fault event, the probability of occurrence of the current fault symptom when the searched fault event occurs, the prior probabilities corresponding to multiple fault events in the fault tree, and the probability of occurrence of the current fault symptom when the multiple fault events occur respectively, can be input into the probability prediction formula corresponding to the Bayesian theorem. The probability of occurrence of the searched fault event under the current fault symptom is predicted by the probability prediction formula to obtain the predicted occurrence probability of the searched fault event.

[0072] Thus, by using the probability calculation capability of the Bayesian algorithm and the data support provided by the reference probability information, the accuracy of predicting the probability of occurrence of each fault event under the current fault symptom is improved, and the probability of occurrence of each fault event under the current fault symptom provides a quantitative basis for the fault location of the humanoid robot (i.e., locating the fault event and fault cause of the humanoid robot), making the fault diagnosis information of the humanoid robot more scientific, objective and accurate. In the case where the same symptoms exist in the fault symptoms caused by multiple fault events (for example, the motor fault of the humanoid robot and the heat dissipation system fault of the humanoid robot may both cause the fault symptom of excessive motor temperature), the fault event with the highest probability (i.e., the most likely) can be determined among the multiple fault events through the above-mentioned probability prediction process, avoiding the subjectivity and uncertainty of judgment based on experience alone.

[0073] Optionally, the probability prediction formula corresponding to Bayes' theorem is expressed as:

[0074]

[0075] Among them, F i represents the fault event found for the i-th time, S represents the current fault symptom, and P(F i |S) indicates a fault event F i The probability of occurrence under the current fault symptom S, P(F i ) refers to the fault event F i The prior probability, P(S|F i ) indicates that in the event of a fault F i The probability of occurrence of the current fault symptom S in the event of a fault.

[0076] Among them, F j represents the jth fault event in the fault tree, that is, the fault event corresponding to the jth node in the fault tree, P(S|F j ) indicates that in the event of a fault F j The probability of occurrence of the current fault symptom S in the event of a fault, P(Fj ) refers to the fault event F j The prior probability of .

[0077] As an example, the fault event of motor failure is represented by F motor The fault symptom of motor overtemperature is represented by S temp It is known that the probability of motor overheating in the event of a motor failure is P(S temp |F motor )=0.8, the prior probability of motor failure is P(F motor )=0.05. Similarly, the probability of other fault events being related to the symptom of motor overtemperature can also be obtained by statistics or estimation of historical probability data and / or experimental data. Using these probabilities and the above formula, the probability of motor fault P(F motor |S temp ).

[0078] In some embodiments, the probability of occurrence of a fault event under the current fault symptoms can be predicted based on a fuzzy reasoning system to obtain the predicted probability of occurrence of the fault event. The fuzzy reasoning system includes a fuzzy rule base, and the probability of occurrence of a fault event under the current fault symptoms is calculated by fuzzy reasoning according to the fuzzy rules in the fuzzy rule base. The fuzzy rule base is determined based on expert knowledge and a large amount of experimental data. With the support of a large amount of data, it can better handle the uncertainty and ambiguity between fault symptoms and fault events. It is suitable for handling some complex fault relationships that are difficult to describe with precise mathematical models, and improves the accuracy of predicting the probability of occurrence of a fault event under the current fault symptoms.

[0079] In this embodiment, a fuzzy rule base is determined based on expert knowledge related to the humanoid robot and a large amount of experimental data, and a fuzzy reasoning system is constructed based on the fuzzy rule base. The current fault symptoms of the humanoid robot are input into the fuzzy reasoning system; fuzzy reasoning calculations are performed in the fuzzy reasoning system based on the fuzzy rules in the fuzzy rule base to obtain fuzzy probability values ​​corresponding to multiple fault events; the fuzzy probability values ​​corresponding to the multiple fault events are defuzzified to obtain the occurrence probabilities of the multiple fault events under the current fault symptoms, that is, the predicted occurrence probabilities corresponding to the multiple fault events.

[0080] In one example, for the diagnosis of motor failure of a humanoid robot, the input variable of the fuzzy inference system can be at least one of the following fault symptoms: the fuzzy level of the motor temperature of the humanoid robot (such as low temperature, medium temperature, high temperature), the fuzzy change rate of the motor current of the humanoid robot (such as slow change, rapid change, etc.), and the fuzzy amplitude of the motor vibration of the humanoid robot (such as slight, medium, severe). The fault symptoms are input into the fuzzy inference system, and the fuzzy rules in the fuzzy rule base, such as "if the motor temperature is high, the motor current changes rapidly, and the motor vibration amplitude is severe, then the possibility of motor failure is very high", are inferred based on the fuzzy rules.

[0081] Regarding event search in a fault tree, the following embodiments are provided.

[0082] In some embodiments, multiple event searches can be performed in the fault tree according to the current fault information through a set search strategy. The set search strategy includes a depth-first search strategy (DFS). Thus, the fault tree is traversed through the depth-first search strategy to find fault events matching the current fault information from the fault tree, thereby improving the accuracy and efficiency of the fault event search.

[0083] Figure 3 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 2 .like Figure 3 As shown, the method of the embodiment of the present application includes:

[0084] S301, obtaining current fault information of the humanoid robot.

[0085] The implementation principle and technical effects of S301 may refer to the aforementioned embodiments and will not be described in detail.

[0086] S302: According to the current fault information and through the set search strategy, an i-th event search is performed in the fault tree to obtain the i-th searched fault event.

[0087] like Figure 3 As shown, S302 includes: S3021, when i is equal to 1, according to the current fault information and the depth-first search strategy, start searching from the top event of the fault tree to obtain the fault event obtained by the first search; S3022, when i is greater than 1, according to the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, determine the starting node of the i-th search, and according to the current fault information, start searching from the starting node to obtain the fault event corresponding to the i-th search.

[0088] In this embodiment, when i is equal to 1, starting from the top event of the fault tree, according to the depth-first search strategy, the intermediate events and sub-events of the fault tree are accessed in sequence, and the accessed fault events are matched with the current fault information to search for the fault event that matches the current fault event, that is, the fault event searched for the i-th time is obtained. When i is greater than 1, the next layer event (intermediate event or sub-event) connected to the fault event searched for the i-1th time can be first determined as the starting node of the i-th search according to the predicted occurrence probability of the fault event searched for the i-1th time, or the starting node of the i-th search can be determined according to the depth-first search strategy; from the starting node, according to the depth-first search strategy, the fault events of the fault tree are accessed in sequence, and the accessed fault events are matched with the current fault information to search for the fault event that matches the current fault event, that is, the fault event searched for the i-th time is obtained. Therefore, the fault tree search is performed in combination with the depth-first search strategy and the predicted occurrence probability of the fault event searched last time, so that the next layer event related to certain fault events is focused on during the search process, thereby improving the accuracy and efficiency of the fault event search.

[0089] Optionally, S3022 includes: if the predicted probability of occurrence of the fault event searched for the i-1th time is greater than or equal to the probability threshold, then determining the starting node of the i-th search to be the next intermediate event or the next sub-event connected to the fault event searched for the i-1th time, otherwise determining the starting node of the i-th search according to the depth-first search strategy. Thus, the fault event with a higher predicted probability of occurrence is regarded as a key suspect, and the subordinate events of the fault event are searched first, otherwise the in-depth search of its subordinate events can be skipped first, and other branches of the fault tree can be searched in depth-first order, effectively improving the accuracy and efficiency of the fault event search.

[0090] In one example, the top event of the fault tree is the complete loss of mobility of the humanoid robot. During the search process, the predicted probability of occurrence of intermediate events such as the humanoid robot's power system failure, the humanoid robot's control system paralysis, and the humanoid robot's transmission system damage can be calculated first. If the probability calculation result of the power system failure is high, the predicted probability of occurrence of its sub-events such as power failure, motor driver failure, and motor failure is further analyzed. Finally, the most likely cause of the humanoid robot's failure can be determined based on the probability of these sub-events.

[0091] S303 , predicting the occurrence probability of the fault event searched for the i-th time under the current fault symptom based on the historical fault information, and obtaining the predicted occurrence probability of the fault event searched for the i-th time.

[0092] The implementation principle and technical effects of S303 may refer to the aforementioned embodiments and will not be described in detail.

[0093] S304, determining whether the i-th search satisfies the search end condition.

[0094] If the i-th search satisfies the search end condition, S306 is executed, otherwise S305 is executed.

[0095] The search end condition may be that all possible causes and paths of the top event in the fault tree have been traversed, that is, all nodes and edges in the fault tree have been traversed. Alternatively, the search end condition may be that the number of searches is greater than or equal to a number threshold.

[0096] S305, adding 1 to i.

[0097] In this embodiment, after i is incremented by 1, the process jumps to S302.

[0098] S306, generating fault diagnosis information of the humanoid robot according to the predicted occurrence probability of the searched fault event, and the fault diagnosis information is used to assist the fault handling of the humanoid robot.

[0099] In this embodiment, after completing multiple searches, fault diagnosis information of the humanoid robot is generated according to the predicted occurrence probabilities corresponding to the multiple fault events searched multiple times. The specific implementation principles and technical effects can be referred to the aforementioned embodiments and will not be repeated here.

[0100] In an embodiment of the present application, a fault tree is used to represent various fault events of a humanoid robot and the causal relationship between the various fault events. By searching for fault events in the fault tree and making probability predictions for fault events, the accuracy of fault diagnosis of the humanoid robot is improved. During the search process, the depth-first search strategy and the predicted probability of occurrence of the fault events searched in the previous step are combined to perform the next step of searching for fault events, thereby improving the search efficiency and accuracy of fault events.

[0101] In some embodiments, an artificial intelligence search algorithm may be used to perform multiple event searches in a fault tree to improve the search efficiency and accuracy of fault events.

[0102] Optionally, the artificial intelligence search algorithm is a genetic algorithm.

[0103] Optionally, the artificial intelligence search algorithm is a particle swarm optimization algorithm.

[0104] Among them, when the artificial intelligence algorithm is a genetic algorithm or a particle swarm optimization algorithm, the search path in the fault tree can be encoded as individuals in the artificial intelligence algorithm (the population in the genetic algorithm or the particles in the particle swarm optimization algorithm), and each individual represents a possible troubleshooting order. For example, in a fault tree in which a humanoid robot completely loses the ability to move, an individual may represent a search path that first checks the power system of the humanoid robot, then checks the control system of the humanoid robot, and finally checks the transmission system of the humanoid robot.

[0105] Furthermore, the fitness function (i.e., the objective function to be optimized) in the artificial intelligence algorithm can be determined based on the predicted occurrence probability of the searched fault events and the constraints, and the fitness function can be optimized and solved by searching for fault events, so that the artificial intelligence algorithm tends to choose a search path with a high fault probability and low detection cost, thereby improving the diagnostic efficiency and accuracy of the humanoid robot. Among them, the constraints include constraints on fault detection costs and constraints on fault detection events.

[0106] The following embodiments are provided with respect to making a fault decision based on the predicted probability of occurrence of a fault event.

[0107] Figure 4 A schematic diagram of a humanoid robot fault diagnosis method provided in an embodiment of the present application Figure 3 .like Figure 4 As shown, the method of the embodiment of the present application includes:

[0108] S401, obtaining current fault information of the humanoid robot.

[0109] S402, based on the current fault information, multiple event searches are performed in a pre-constructed fault tree, and after each event search, the occurrence probability of the searched fault event under the current fault symptom is predicted to obtain the predicted occurrence probability of the searched fault event.

[0110] The implementation principles and technical effects of S401 to S402 may refer to the aforementioned embodiments and will not be described in detail.

[0111] S403, obtaining the occurrence probability of the associated fault symptoms of the searched fault event when the searched fault event occurs.

[0112] There may be multiple current fault symptoms, some or all of which may be related to the searched fault event. The associated fault symptoms of the searched fault event may be some or all of the current fault symptoms.

[0113] In this embodiment, the occurrence probability of the associated fault symptom of the searched fault event when the searched fault event occurs can be obtained from the reference probability information. The reference probability information can refer to the relevant description of the above embodiment and will not be repeated here.

[0114] S404, determining the information entropy of the searched fault event according to the occurrence probability of the associated fault symptoms of the searched fault event and the number of the associated fault symptoms of the searched fault event, where the information entropy is used to indicate the degree of uncertainty of the searched fault event.

[0115] Among them, the greater the information entropy of the fault event, the higher the uncertainty of the fault event, that is, the higher the uncertainty of the humanoid robot occurring the fault event.

[0116] In this embodiment, for each searched fault event, the information entropy of the fault event can be calculated by combining the predicted occurrence probability of the fault event and the number of fault symptoms associated with the fault event. The calculation formula of the information entropy is as follows:

[0117]

[0118] Among them, F i represents the fault event found in the ith search, S k Indicates the current fault symptom that is related to F i The kth fault symptom associated, P(S k |F i ) indicates that in the event of a fault F i Associated fault symptoms in case of occurrence S k The probability of occurrence of the fault event F i The number of associated fault symptoms, H(F i ) indicates a fault event F i The greater the information entropy, the more likely the humanoid robot will experience a failure event F i The higher the uncertainty.

[0119] S405: Generate fault diagnosis information according to the predicted occurrence probability of the searched fault event and the information entropy of the searched fault event.

[0120] In this embodiment, the fault diagnosis information may include the predicted occurrence probability of the searched fault event and the information entropy of the searched fault event.

[0121] Optionally, in the process of generating fault diagnosis information, the degree of uncertainty of the occurrence of the searched fault event in the humanoid robot can be judged based on the information entropy of the searched fault event, and troubleshooting suggestions for the fault event can be generated based on the degree of uncertainty, and the troubleshooting suggestions can be written into the fault diagnosis information.

[0122] For example, in the fault diagnosis of abnormal robot motion, if the information entropy of an intermediate event is calculated to be large, it means that the sub-events under the intermediate event are more complex or the relevant data are insufficient. The following troubleshooting suggestions can be generated: It is recommended to further check the relevant sensor data or perform more fault tests to reduce the uncertainty of the intermediate event.

[0123] In the embodiment of the present application, after searching for fault events based on the fault tree, the probability of occurrence of the fault event under the current fault symptom is predicted, the information entropy of the fault event is determined based on the probability of occurrence of the associated fault symptoms under the fault event, and the fault diagnosis information is generated by combining the predicted probability of occurrence of the fault event and the information entropy of the fault event. Thus, by combining multi-source fault information such as the fault tree, probability prediction, and information entropy, the humanoid robot is analyzed and decided on the fault, which enhances the diagnostic capability and adaptability of complex fault situations and ensures that the humanoid robot can be effectively diagnosed under various complex working conditions.

[0124] In some embodiments, the current fault information of the humanoid robot can be input into the deep learning model, and features of the current fault information of the humanoid robot can be extracted in the deep learning model. Fault decisions are made based on the extracted features to obtain fault diagnosis results. Thus, the deep learning model is used to automatically learn the deep features in the fault data and learn fault diagnosis, so as to realize automatic fault diagnosis of the humanoid robot. Since the deep learning model has strong self-learning and generalization capabilities, it can be used to process different types of fault diagnoses with higher complexity and improve the accuracy of diagnosis.

[0125] Among them, the deep learning model is obtained through training with a large amount of training data, and the training data includes historical fault information used for training and fault diagnosis results corresponding to the historical fault information, wherein the fault diagnosis results may include the fault event that occurred, the predicted probability of occurrence (or confidence) corresponding to the fault event, the fault location and / or the degree of fault.

[0126] Optionally, the deep learning model is a convolutional neural network or a deep belief network.

[0127] In one example, for a visual system failure of a humanoid robot, in a deep learning model, feature extraction is performed on the sensor data of the image sensor of the humanoid robot, and based on the extracted features, the specific fault event such as lens blur, image sensor damage, or image processing algorithm error is identified, and the confidence level corresponding to the fault event is given.

[0128] In some embodiments, after the fault diagnosis information of the humanoid robot is generated, a fault diagnosis report of the humanoid robot may be generated based on the current fault information and the fault diagnosis information. The fault diagnosis report may include: the time of the fault, the description of the fault phenomenon, the fault diagnosis process and fault diagnosis method, the cause of the fault, and the corresponding fault repair suggestions. Thus, the problem that the fault diagnosis results of the humanoid robot are simply recorded and it is difficult to provide effective data support for the subsequent maintenance, planning, design improvement, etc. of the humanoid robot is solved.

[0129] Among them, the fault occurrence time and the fault phenomenon description can be obtained from the current fault information. The fault diagnosis process may include at least one of the following: the search path of the fault tree, the searched fault events, the predicted occurrence probability of the searched fault events, and the information entropy of the searched fault events. The fault diagnosis method may include: a search method (such as depth-first search, a fault tree search based on an artificial intelligence algorithm) and / or a decision method (such as a decision method based on information entropy, a decision method based on a deep learning model). The cause of the fault is determined among the searched fault events, for example, the fault event with the highest predicted probability of occurrence is selected as the determined fault cause, or the fault cause is determined in the fault event in combination with the predicted probability of occurrence of the fault event and the information entropy of the fault event. After determining the cause of the fault, the fault repair suggestion corresponding to the fault cause can be obtained from the fault tree.

[0130] In one example, if the fault diagnosis result of the humanoid robot is that the motor winding short circuit causes the humanoid robot to completely lose the ability to move, then the fault diagnosis report may include at least one of the following: the position of the short-circuited motor, the short-circuit situation (such as the number of phases of the short-circuited winding, the estimated position of the short-circuit point, etc.), repair suggestions for replacing the motor or repairing the motor winding, or measures to prevent similar faults from happening again (such as strengthening daily maintenance of the motor, regularly checking the insulation performance of the motor, etc.).

[0131] Optionally, the fault diagnosis report is stored in the fault database of the robot. Thus, the fault diagnosis report can be used to diagnose the faults of the humanoid robot in the future to improve the accuracy of the fault diagnosis; the fault diagnosis report can also be used to conduct statistical analysis on the historical faults of the humanoid robot in the future to summarize the fault patterns of the humanoid robot and provide data support for the subsequent design improvement, maintenance plan formulation and fault prediction of the humanoid robot.

[0132] Figure 5 The following is a flow chart of the process of constructing a fault tree according to an embodiment of the present application. Figure 5 As shown in Figure 1, the fault tree construction process includes:

[0133] S501, obtaining fault event information related to the humanoid robot, where the fault event information includes multiple fault events used to construct a fault tree, event influencing factors corresponding to the multiple fault events, and event causal relationships between the multiple fault events.

[0134] The multiple fault events may be historical fault events that have occurred, or may be fault events that may occur in the humanoid robot according to relevant knowledge, experience or experiments of the humanoid robot. The event impact factors corresponding to the multiple fault events respectively indicate the event impact degrees corresponding to the multiple fault events respectively. In other words, the event impact factors corresponding to the fault events indicate the impact degrees of the fault events on the humanoid robot.

[0135] In this embodiment, based on the physical structure and / or motion state of the humanoid robot, multiple fault events related to the humanoid robot can be obtained, and event influencing factors corresponding to the multiple fault events and event causal relationships between the multiple fault events can be determined.

[0136] For example, the multiple fault events obtained include: the humanoid robot completely loses its ability to move, the humanoid robot frequently falls during movement, the humanoid robot's joint movement is abnormal, the humanoid robot's sensor data is abnormal, the humanoid robot's communication failure, and the humanoid robot's power system failure. Among them, joint movement abnormalities, such as: joint jamming, excessive shaking, inability to reach a specified angle, etc.; sensor data abnormalities, such as: blurred image data of the visual sensor, loss of image data of the visual sensor, excessive deviation of force sensor readings, etc.; communication failures, such as: interruption or error in data transmission between the humanoid robot and the external control system or other equipment; power system failures, such as: the battery power of the humanoid robot lasts too short, charging abnormalities, etc.

[0137] Optionally, after obtaining multiple fault events, the multiple fault events may be classified and sorted to obtain the categories to which the multiple fault events belong respectively, and the event impact factors corresponding to the multiple fault events and the event causal relationships between the multiple fault events may be determined according to the categories to which the multiple fault events belong respectively. Thus, the accuracy of the event impact factors and the event causal relationships may be improved through fault event classification.

[0138] In this optional method, multiple fault categories and subcategories corresponding to the multiple fault categories can be determined according to the composition structure of the humanoid robot. Multiple fault events are classified according to the multiple fault categories and subcategories corresponding to the multiple fault categories. Afterwards, the event impact factor of the fault event can be determined according to the impact of the category to which the fault event belongs on the humanoid robot; the event causal relationship between the multiple fault events can be determined according to the correlation between the categories to which the fault events belong.

[0139] Furthermore, according to the mechanical structure, electrical system, control system and software system of the humanoid robot, it can be determined that the fault categories include mechanical structure failure, electrical system failure, control system failure and software system failure; for mechanical structure failure, it can be determined that its corresponding subcategories include at least one of the following: leg structure failure, arm structure failure or torso connection failure; for electrical system failure, it can be determined that its corresponding subcategories include at least one of the following: motor failure, circuit short circuit, open circuit or electronic component damage; for control system failure, it can be determined that its corresponding subcategories include controller freeze, control signal distortion, etc.; for software system failure, it can be determined that its corresponding subcategories include program crash, abnormal behavior caused by algorithm abnormality, etc.

[0140] S502: Determine a top event in a fault tree from among the multiple fault events according to event impact factors respectively corresponding to the multiple fault events.

[0141] In this embodiment, by comparing the event impact factors corresponding to the event impact factors corresponding to the multiple fault events, a fault event whose event impact factor meets the conditions can be selected from the multiple fault events, and the fault event can be determined as the top event of the fault tree. Among them, selecting the fault event whose event impact factor meets the conditions from the multiple fault events may include: selecting the fault event with the largest event impact factor from the multiple fault events; or selecting the fault event whose event impact factor is greater than the impact threshold from the multiple fault events. If there are multiple fault events whose event impact factors meet the conditions, multiple fault trees can be constructed, or a fault event can be randomly selected from the fault events whose event impact factors meet the conditions as the top event of the fault tree.

[0142] For example, the complete loss of mobility of a humanoid robot can be considered a top event because the failure event renders the robot unable to perform any tasks, reflecting the failure of the robot's overall functionality.

[0143] For another example, frequent falls of humanoid robots during movement can also be regarded as top events, which means that there may be serious problems with the balance control, gait planning or perception system of the humanoid robot, which seriously affects the mobility and safety of the humanoid robot.

[0144] S503, according to the event causal relationship between multiple fault events, determine the intermediate events corresponding to the top event, the logical relationship between the top event and the intermediate events, the sub-events corresponding to the intermediate events, and the logical relationship between the intermediate events and the sub-events in the multiple fault events.

[0145] In this embodiment, after determining the top event, the fault event that has an event causal relationship with the top event is determined to be an intermediate event, and the fault event that has an event causal relationship with the intermediate event is determined to be a sub-event. In the event causal relationship, one intermediate event may lead to the occurrence of the top event, or multiple intermediate events may lead to the occurrence of the top event together. Therefore, the logical relationship between the top event and the intermediate event corresponding to the top event can be determined based on the event causal relationship between the top event and the intermediate event corresponding to the top event, and the logical relationship between the intermediate event and the sub-event corresponding to the intermediate event can be determined based on the event causal relationship between the intermediate event and the sub-event corresponding to the intermediate event.

[0146] Among them, the types of logical relationships include logical "and" and logical "or". The logical "and" between the top event and the intermediate event means that multiple intermediate events will trigger the top event, and the logical "or" between the top event and the intermediate event means that one intermediate event will trigger the top event. The logical "and" and logical "or" between the intermediate event and the sub-event are similar. It can be seen that the logical relationship between multiple events in the fault tree can be determined based on the causal relationship between the multiple events.

[0147] For example, taking the top event of the humanoid robot completely losing its ability to move as an example, its intermediate events include: power system failure, control system paralysis, transmission system damage, etc. Taking the top event of the humanoid robot frequently falling during movement as an example, its intermediate events include: balance sensor failure, gait algorithm error, leg mechanical structure damage, etc. Taking the intermediate event of the humanoid robot's power system failure as an example, its sub-events include: power failure (such as battery exhaustion, power module damage), motor driver failure (such as failure to drive the motor to operate normally), motor failure (such as motor winding short circuit, open circuit, bearing damage), etc.

[0148] S504, draw the nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events, and construct a fault tree.

[0149] Among them, the logic gate symbols include "OR gate" and "AND gate". The "OR gate" represents the logical relationship "OR". Specifically, the "OR" gate indicates that the occurrence of any one of the multiple lower-level events corresponding to the upper-level event may lead to the occurrence of the upper-level event. The "AND gate" represents the logical relationship "AND". Specifically, the "AND gate" indicates that the occurrence of multiple lower-level events corresponding to the upper-level event will lead to the occurrence of the upper-level event.

[0150] In this embodiment, after determining the top event, intermediate events, sub-events and the logical relationship between events, the nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events can be drawn. In the drawing process, starting from the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events are drawn step by step downward to obtain a fault tree.

[0151] Optionally, in the process of searching for fault events in the fault tree, if the fault event searched for the i-1th time is an intermediate event and the logical relationship between the fault event searched for the i-1th time and the top event is a logical "AND", then the start node of the i-th search is determined in the unsearched fault event corresponding to the top event; if the fault event searched for the i-1th time is an intermediate event and the logical relationship between the fault event searched for the i-1th time and the top event is a logical "OR", then the start node of the i-th search is determined in the sub-event corresponding to the fault event searched for the i-1th time. Thus, considering that the top event has a greater impact on the humanoid robot, when the logical relationship between the top event and the intermediate event is a logical "AND", the intermediate event that determines whether the top event occurs is searched first.

[0152] Optionally, after the fault tree performs multiple fault event searches, the predicted occurrence probability of at least one fault event among the multiple searched fault events can be updated based on the predicted occurrence probabilities corresponding to the multiple searched fault events and the logical relationship between the multiple searched fault events to improve the accuracy of the predicted occurrence probability of the fault event.

[0153] In this optional method, if the multiple searched fault events include top events and intermediate events, the predicted occurrence probability of the top event can be updated according to the predicted occurrence probability of the intermediate event and the logical relationship between the top event and the intermediate event; if the multiple searched fault events include intermediate events and sub-events, the predicted occurrence probability of the intermediate event can be updated according to the predicted occurrence probability of the sub-event and the logical relationship between the intermediate event and the sub-event. Thus, the accuracy of the predicted occurrence probability of the fault event is improved by using the logical relationship between the events.

[0154] In the process of updating the predicted occurrence probability of the top event according to the predicted occurrence probability of the intermediate event and the logical relationship between the top event and the intermediate event: if the top event corresponds to multiple intermediate events and the logical relationship between the top event and the multiple intermediate events is a logical "and", the predicted occurrence probabilities corresponding to the multiple intermediate events can be multiplied to obtain a first value (that is, the product of the predicted occurrence probabilities corresponding to the multiple intermediate events); the predicted occurrence probability of the top event is determined as the first value, or, according to the first value, the predicted occurrence probability of the top event is adjusted.

[0155] Among them, according to the first numerical value, the predicted probability of occurrence of the top event is adjusted in the following manner: the first numerical value and the predicted probability of occurrence of the top event are weighted to obtain a weighted result, and the predicted probability of occurrence of the top event is updated to the weighted result; or, if the first numerical value is smaller than the predicted probability of occurrence of the top event, the predicted probability of occurrence of the top event is reduced according to the difference between the first numerical value and the predicted probability of occurrence of the top event; if the first numerical value is larger than the predicted probability of occurrence of the top event, the predicted probability of occurrence of the top event is increased according to the difference between the first numerical value and the predicted probability of occurrence of the top event.

[0156] In the process of updating the predicted occurrence probability of the top event based on the predicted occurrence probability of the intermediate event and the logical relationship between the top event and the intermediate event: if the top event corresponds to multiple intermediate events and the logical relationship between the top event and the multiple intermediate events is a logical "or", the predicted occurrence probabilities of the multiple intermediate events can be added to obtain a second value; the predicted occurrence probability of the top event is determined as the second value, or, based on the second value, the predicted occurrence probability of the top event is adjusted.

[0157] Among them, according to the second numerical value, the predicted occurrence probability of the top event is adjusted in the following manner: the second numerical value and the predicted occurrence probability of the top event are weighted to obtain a weighted result, and the predicted occurrence probability of the top event is updated to the weighted result; or, if the second numerical value is smaller than the predicted occurrence probability of the top event, the predicted occurrence probability of the top event is reduced according to the difference between the second numerical value and the predicted occurrence probability of the top event; if the second numerical value is larger than the predicted occurrence probability of the top event, the predicted occurrence probability of the top event is increased according to the difference between the second numerical value and the predicted occurrence probability of the top event.

[0158] Among them, the predicted occurrence probability of the intermediate event is updated according to the predicted occurrence probability of the sub-event and the logical relationship between the intermediate event and the sub-event. The above description of the implementation method of "updating the predicted occurrence probability of the top event according to the predicted occurrence probability of the intermediate event and the logical relationship between the top event and the intermediate event" can be referred to and no further details will be given.

[0159] Optionally, the fault event information also includes event attribute information corresponding to multiple fault events and / or the mutual influence between multiple fault events, such as Figure 5 As shown, the fault tree construction process also includes S505 and / or S506, wherein:

[0160] S505, adding corresponding event attribute information to multiple nodes in the fault tree according to event attribute information corresponding to multiple fault events, wherein the event attribute information includes at least one of the following: fault description content, related fault symptoms, or related fault detection methods.

[0161] For example, for the motor failure event, the fault description is "the internal winding of the motor is open or short-circuited", and the fault symptoms include abnormal noise during motor operation, excessive motor temperature, unstable motor speed, and the motor being unable to rotate. The fault detection methods include using a multimeter to measure the motor winding resistance, checking whether the motor current is abnormal, and detecting the motor temperature with an infrared thermometer.

[0162] In this optional method, the event attribute information corresponding to the fault event can be added as the event attribute information of the node corresponding to the fault event in the fault tree, thereby improving the information richness of the fault tree. During the fault event search process, the event attribute information of the node can be used to match the current fault information of the humanoid robot, thereby improving the accuracy of the fault event search.

[0163] S506, adding relationship annotation information to multiple nodes in the fault tree according to the mutual influence between the multiple fault events, where the relationship annotation information is used to indicate the mutual influence between the fault events corresponding to the multiple nodes in the fault tree.

[0164] In this embodiment, relationship annotation information can be added to the nodes that have mutual influence in the fault tree by adding text annotations and / or marks (such as dotted lines, special symbols), and the mutual influence between the fault events corresponding to the nodes can be indicated by the relationship annotation information.

[0165] For example, in a humanoid robot, some electrical faults may cause damage to some mechanical parts, or some abnormalities in the software system may cause some erroneous operations in the control system. In the fault tree, the node corresponding to the electrical fault and the node corresponding to the damage to the mechanical parts affected by the electrical fault are connected by a dotted line, and the node corresponding to the software system abnormality and the node corresponding to the control system operation abnormality affected by the software system abnormality are connected by a dotted line.

[0166] Therefore, by marking the event attribute information of the fault event and the mutual influence between the fault events in the fault tree, the information richness in the fault tree is improved, more comprehensive information is provided for the fault diagnosis of the humanoid robot, and the fault diagnosis accuracy of the humanoid robot is improved.

[0167] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0168] Figure 6 A schematic diagram of the structure of a fault diagnosis device for a humanoid robot provided in an embodiment of the present application, as shown in FIG. Figure 6 As shown, the humanoid robot fault diagnosis device 600 of the embodiment of the present application includes: an acquisition module 601, a probability prediction module 602 and a diagnosis information generation module 603. Among them:

[0169] An acquisition module 601 is used to acquire the current fault information of the humanoid robot; a probability prediction module 602 is used to perform multiple event searches in a pre-constructed fault tree related to the humanoid robot based on the current fault information, and after each event search, predict the probability of occurrence of the searched fault event under the current fault symptom to obtain the predicted probability of occurrence of the searched fault event, and the current fault symptom is determined based on the current fault information; a diagnosis information generation module 603 is used to generate fault diagnosis information of the humanoid robot based on the predicted probability of occurrence of the searched fault event, and the fault diagnosis information is used to assist the fault handling of the humanoid robot; wherein, the fault tree is constructed based on the fault event information related to the humanoid robot, the fault event information includes multiple fault events and event causal relationships between the multiple fault events, and the distribution of the multiple fault events in the fault tree is determined based on the event impact levels corresponding to the multiple fault events and the event causal relationships between the multiple fault events.

[0170] In some embodiments, the probability prediction module 602 is specifically used to: obtain reference probability information, the reference probability information includes the prior probability of the searched fault event, the probability of occurrence of the current fault symptom when the searched fault event occurs, the prior probabilities corresponding to multiple fault events in the fault tree, and the probability of occurrence of the current fault symptom when multiple fault events occur respectively; based on the reference probability information, the probability of occurrence of the searched fault event under the current fault symptom is predicted by Bayesian theorem to obtain the predicted occurrence probability.

[0171] In some embodiments, the probability prediction module 602 is specifically used to: perform multiple event searches in the fault tree according to the current fault information through a set search strategy, and the set search strategy includes a depth-first search strategy.

[0172] In some embodiments, multiple fault events in a fault tree are distributed according to a hierarchical structure of top events, intermediate events, and sub-events, and the i-th search of multiple event searches includes: when i is equal to 1, based on the current fault information and a depth-first search strategy, the search is started from the top event of the fault tree to obtain the fault event obtained by the first search; when i is greater than 1, based on the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, the starting node of the i-th search is determined, and based on the current fault information, the search is started from the starting node to obtain the fault event corresponding to the i-th search.

[0173] In some embodiments, the starting node of the i-th search is determined based on the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, including: if the predicted probability of occurrence of the fault event searched for the i-1th time is greater than or equal to the probability threshold, then the starting node is determined to be the next layer of intermediate event or the next layer of sub-event connected to the fault event searched for the i-1th time, otherwise the starting node is determined according to the depth-first search strategy.

[0174] In some embodiments, the diagnostic information generation module 603 is specifically used to: obtain the probability of occurrence of the associated fault symptoms of the searched fault event when the searched fault event occurs; determine the information entropy of the searched fault event based on the probability of occurrence of the associated fault symptoms of the searched fault event and the number of associated fault symptoms of the searched fault event, the information entropy is used to indicate the degree of uncertainty of the searched fault event; generate fault diagnosis information based on the predicted probability of occurrence of the searched fault event and the information entropy of the searched fault event.

[0175] In some embodiments, the fault tree construction process includes: obtaining fault event information, the fault event information also includes event impact factors corresponding to multiple fault events, the event impact factors corresponding to multiple fault events respectively represent the event impact degrees corresponding to the multiple fault events respectively; determining the top event in the fault tree among multiple fault events according to the event impact factors corresponding to the multiple fault events respectively; determining the intermediate event corresponding to the top event, the logical relationship between the top event and the intermediate event, the sub-event corresponding to the intermediate event, and the logical relationship between the intermediate event and the sub-event among multiple fault events according to the event causal relationship between the multiple fault events; drawing the nodes corresponding to the top event, the nodes corresponding to the intermediate event, the nodes corresponding to the sub-event, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate event, the edges between the nodes corresponding to the intermediate event and the nodes corresponding to the sub-event, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate event, and the logic gate symbols corresponding to the logical relationship between the intermediate event and the sub-event, to construct a fault tree.

[0176] In some embodiments, the fault event information also includes event attribute information corresponding to multiple fault events and / or the mutual influence between multiple fault events. The nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events are drawn. After the fault tree is constructed, it also includes: according to the event attribute information corresponding to the multiple fault events, corresponding event attribute information is added to the multiple nodes in the fault tree, and the event attribute information includes at least one of the following: fault description content, related fault symptoms or related fault detection methods; and / or, according to the mutual influence between the multiple fault events, relationship annotation information is added to the multiple nodes in the fault tree, and the relationship annotation information is used to indicate the mutual influence between the fault events corresponding to the multiple nodes.

[0177] The device of this embodiment can be used to execute the technical solution of any of the above-mentioned method embodiments. Its implementation principle and technical effects are similar and will not be repeated here.

[0178] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. For example, the electronic device can be provided as a server or a computer. Figure 7 , the electronic device 700 includes a processing component 701, which further includes one or more processors, and a memory resource represented by a memory 702, for storing instructions executable by the processing component 701, such as an application. The application stored in the memory 702 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 701 is configured to execute instructions to perform any of the above method embodiments.

[0179] The electronic device 700 may further include a power supply component 703 configured to perform power management of the electronic device 700, a wired or wireless network interface 704 configured to connect the electronic device 700 to a network, and an input / output (I / O) interface 705. The electronic device 700 may operate based on an operating system stored in the memory 702, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0180] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above-mentioned method for diagnosing a humanoid robot fault is implemented.

[0181] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned method for diagnosing the faults of a humanoid robot.

[0182] The computer-readable storage medium mentioned above 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. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0183] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also be present in the fault diagnosis device of the humanoid robot as discrete components.

[0184] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0185] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for diagnosing a humanoid robot fault, characterized in that: include: Get the current fault information of the humanoid robot; According to the current fault information, multiple event searches are performed in a pre-constructed fault tree related to the humanoid robot, and after each event search, the occurrence probability of the searched fault event under the current fault symptom is predicted to obtain the predicted occurrence probability of the searched fault event, wherein the current fault symptom is determined according to the current fault information; generating fault diagnosis information of the humanoid robot according to the predicted occurrence probability of the searched fault event, wherein the fault diagnosis information is used to assist in fault handling of the humanoid robot; Among them, the fault tree is constructed based on fault event information related to the humanoid robot, the fault event information includes multiple fault events and event causal relationships between the multiple fault events, and the distribution of the multiple fault events in the fault tree is determined based on the event impact degrees corresponding to the multiple fault events and the event causal relationships.

2. The humanoid robot fault diagnosis method according to claim 1, characterized in that: The predicting the occurrence probability of the searched fault event under the current fault symptom to obtain the predicted occurrence probability of the searched fault event includes: Acquire reference probability information, the reference probability information including a priori probability of the searched fault event, probability of occurrence of the current fault symptom when the searched fault event occurs, a priori probabilities corresponding to a plurality of fault events in the fault tree, and probability of occurrence of the current fault symptom when the plurality of fault events occur respectively; According to the reference probability information, the occurrence probability of the searched fault event under the current fault symptom is predicted by Bayesian theorem to obtain the predicted occurrence probability.

3. The method for diagnosing a humanoid robot fault according to claim 1, characterized in that: The method further comprises: performing multiple event searches in a pre-built fault tree related to the humanoid robot according to the current fault information, including: According to the current fault information, multiple event searches are performed in the fault tree through a set search strategy, and the set search strategy includes a depth-first search strategy.

4. The method for diagnosing a humanoid robot fault according to claim 3, characterized in that: The multiple fault events in the fault tree are distributed according to a hierarchical structure of a top event, an intermediate event corresponding to the top event, and a sub-event corresponding to the intermediate event; The i-th search of the multiple event searches includes: When i is equal to 1, according to the current fault information and the depth-first search strategy, the search starts from the top event of the fault tree to obtain the fault event obtained by the first search; When i is greater than 1, the starting node of the i-th search is determined according to the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time, and the search is started from the starting node according to the current fault information to obtain the fault event corresponding to the i-th search.

5. The method for diagnosing a fault of a humanoid robot according to claim 4, characterized in that: The step of determining the starting node for the i-th search according to the depth-first search strategy and the predicted probability of occurrence of the fault event searched for the i-1th time includes: If the predicted occurrence probability of the fault event searched for the i-1th time is greater than or equal to the probability threshold, the starting node is determined to be the next layer of intermediate event or the next layer of sub-event connected to the fault event searched for the i-1th time, otherwise the starting node is determined according to the depth-first search strategy.

6. The method for diagnosing a humanoid robot fault according to any one of claims 1 to 5, characterized in that: Generating fault diagnosis information of the humanoid robot according to the predicted occurrence probability of the searched fault event includes: Obtaining the occurrence probability of the associated fault symptom of the searched fault event when the searched fault event occurs; Determining the information entropy of the searched fault event according to the occurrence probability of the associated fault symptom of the searched fault event and the number of the associated fault symptoms of the searched fault event, wherein the information entropy is used to indicate the degree of uncertainty of the searched fault event; The fault diagnosis information is generated according to the predicted occurrence probability of the searched fault event and the information entropy of the searched fault event.

7. The method for diagnosing a humanoid robot fault according to any one of claims 1 to 5, characterized in that: The fault tree construction process includes: Acquire the fault event information, wherein the fault event information further includes event impact factors corresponding to the multiple fault events respectively, and the event impact factors corresponding to the multiple fault events respectively represent event impact degrees corresponding to the multiple fault events respectively; Determining a top event in the fault tree from among the multiple fault events according to event impact factors respectively corresponding to the multiple fault events; According to the event causal relationship between the multiple fault events, determine, among the multiple fault events, an intermediate event corresponding to the top event, a logical relationship between the top event and the intermediate event, a sub-event corresponding to the intermediate event, and a logical relationship between the intermediate event and the sub-event; The nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events are drawn to construct the fault tree.

8. The method for diagnosing a fault of a humanoid robot according to claim 7, characterized in that: The fault event information also includes event attribute information corresponding to the multiple fault events and / or mutual influence between the multiple fault events. After the nodes corresponding to the top event, the nodes corresponding to the intermediate events, the nodes corresponding to the sub-events, the edges between the nodes corresponding to the top event and the nodes corresponding to the intermediate events, the edges between the nodes corresponding to the intermediate events and the nodes corresponding to the sub-events, the logic gate symbols corresponding to the logical relationship between the top event and the intermediate events, and the logic gate symbols corresponding to the logical relationship between the intermediate events and the sub-events are drawn, and the fault tree is constructed, further comprising: According to the event attribute information respectively corresponding to the multiple fault events, corresponding event attribute information is added to the multiple nodes in the fault tree, wherein the event attribute information includes at least one of the following: fault description content, related fault symptoms or related fault detection methods; And / or, according to the mutual influence between the multiple fault events, relationship annotation information is added to the multiple nodes in the fault tree, wherein the relationship annotation information is used to indicate the mutual influence between the fault events corresponding to the multiple nodes.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the fault diagnosis method for a humanoid robot according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the fault diagnosis method for a humanoid robot according to any one of claims 1 to 8 is implemented.

Images