A robot posture and emergency synchronous identification method

Abstract

The present application relates to the field of robot technology, and specifically provides a robot posture and emergency synchronous identification method, the method comprising: constructing an observation signal matrix of a current time window; obtaining a trigger probability of an emergency occurring in the current time window and a category probability corresponding to each category of emergency according to robot posture information of a previous time window; decomposing the observation signal matrix into a low-rank posture signal matrix and a sparse event signal matrix by solving a joint optimization model containing a dynamic compensation term and a penalty weight; obtaining the robot posture of the current time window based on the posture signal matrix; and determining the category of emergency of the current time window in combination with an event dictionary and the category probability. The present application guides event identification through posture estimation results, and event identification results correct posture estimation in return, forming a synchronous identification system that mutually enhances and cooperates.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a method for synchronously recognizing robot posture and sudden events. Background Technology

[0002] Distributed fiber optic sensing technology is widely used for state perception of large structures such as underwater robots due to its advantages such as resistance to electromagnetic interference, high sensitivity, and ability to achieve full-structure health monitoring. The data collected by sensors typically includes slowly varying signals caused by changes in the robot's own posture (such as navigation and robotic arm movements), as well as signals from sudden events generated by interactions with the external environment (such as collisions and scratches). Accurately separating and identifying these two types of signals from the mixed signals is a key challenge for achieving intelligent perception in robots.

[0003] Existing signal processing methods, such as those based on Robust Principal Component Analysis (RPCA), typically assume that attitude signals are low-rank matrices and event signals are sparse matrices, separating them through optimization. However, these methods suffer from two core drawbacks: 1. Limitations of the static low-rank assumption: Traditional RPCA models assume the low-rank background is static or essentially invariant. In real-world applications, the robot's continuous motion causes the attitude background signal to drift slowly but continuously, gradually destroying its strict low-rank properties and leading to singular value diffusion. Algorithms easily misjudge this slow, cumulative change as sparse events, resulting in numerous false alarms. 2. Lack of contextual association in event recognition: After separating sparse event signals, dictionary learning or pattern matching methods are typically used to classify events. This approach is isolated and does not consider the robot's specific state at the time of the event. For example, when a robotic arm extends its arm, its motion signal pattern may be very similar to a minor collision, leading to misjudgment. In reality, the probability of a specific event occurring varies significantly depending on the robot's posture. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for synchronously recognizing robot posture and sudden events. This method can adapt to the slow-changing posture characteristics of robots, integrate the robot's own state prior information for intelligent reasoning, and improve the robustness and physical consistency of recognition.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0006] This invention provides a method for synchronously recognizing robot posture and sudden events, including:

[0007] Acquire robot observation signals collected by distributed sensors and construct the observation signal matrix for the current time window;

[0008] Based on the robot's posture information from the previous time window, obtain the trigger probability of a sudden event occurring in the current time window and the category probability corresponding to each type of sudden event;

[0009] By solving a joint optimization model that includes a dynamic compensation term and penalty weights, the observed signal matrix is ​​decomposed into a low-rank attitude signal matrix and a sparse event signal matrix; wherein, the dynamic compensation term is used to constrain the continuity of the attitude signal matrix in adjacent time windows; and the penalty weights are obtained by mapping the trigger probability.

[0010] Based on the attitude signal matrix, the robot attitude in the current time window is obtained;

[0011] By combining the event dictionary and the category probabilities, sparse optimization is performed on the event signal matrix to determine the category of sudden events in the current time window.

[0012] In some embodiments of the present invention, the probability of a sudden event occurring in the current time window is obtained based on the robot posture information of the previous time window, including:

[0013] Based on the robot's posture information from the previous time window, the linear velocity of the sensor node on the motion trajectory in the current time window is obtained;

[0014] Based on the linear velocity and the distance from the sensor node to the nearest environmental obstacle, a risk score for the sensor node is calculated.

[0015] The risk score is normalized to obtain the probability of a sudden event occurring at the sensor node in the current time window.

[0016] In some embodiments of the present invention, the linear velocity of the sensor node on the motion trajectory of the current time window is obtained based on the robot posture information of the previous time window, including:

[0017] Based on the robot's posture information from the previous time window, the robot joint angle corresponding to the sensor node in the current time window is calculated using the following formula:

[0018] ;

[0019] in, Indicates the first time window of the current time window The sampling time of the first sampling moment The robot joint angles corresponding to each sensor node; Indicates the current time window; Indicates the previous time window; This indicates the last sampling time of the previous time window. Robot posture information corresponding to each sensor node; This indicates the number of sampling points in each time window; Indicates the first The number of sensor nodes in the current time window angular velocity at each sampling time; Indicates the sampling time interval;

[0020] Among them, the robot posture information corresponding to the sensor node is the robot joint angle corresponding to the sensor node;

[0021] The linear velocity of the sensor node is calculated based on the robot joint angle corresponding to the sensor node, using the following formula:

[0022] ;

[0023] in, Indicates the first The number of sensor nodes in the current time window The linear velocity at each sampling time; Indicates the first The position of each sensor node relative to the Jacobian matrix of the robot joint space.

[0024] In some embodiments of the present invention, the risk score of the sensor node is calculated using the following formula:

[0025] ;

[0026] in, Indicates the first The number of sensor nodes in the current time window Risk score at each sampling time; Indicates the first The number of sensor nodes in the current time window The location of each sampling time; Indicates and The nearest environmental obstacle; express and The distance; This represents the speed weighting coefficient; This represents the distance weighting coefficient; Represents a tiny constant;

[0027] The normalized calculation formula for the risk score is as follows:

[0028] ;

[0029] in, Indicates the first The number of sensor nodes in the current time window The probability of a sudden event occurring at each sampling moment; This indicates a preset risk threshold; This represents the sensitivity coefficient.

[0030] In some embodiments of the present invention, the joint optimization model including dynamic compensation terms and penalty weights is as follows:

[0031] ;

[0032] in, Represents the attitude signal matrix; Represents the event signal matrix; Represents the observed signal matrix; Represents the nuclear norm; Represents the L1 norm; Denotes the Frobenius norm; Represents the penalty weight matrix. , Represents a tiny positive number; It represents the Hadamardi (or Hadama) stack; , and Represents the regularization parameter; This indicates a dynamic compensation item.

[0033] In some embodiments of the present invention, the robot posture for the current time window is obtained based on the posture signal matrix, including:

[0034] Extract the smoothed strain signal from the low-rank attitude signal matrix;

[0035] The smooth strain signal is converted into the robot's posture information using a pre-calibrated strain-attitude mapping model.

[0036] The attitude information is input into the state estimator to obtain the optimal robot attitude estimate for the current time window.

[0037] In some embodiments of the present invention, the categories of emergencies include collision events, scratch events, and joint limitation events;

[0038] Obtain the category probability for each type of emergency event within the current time window, including:

[0039] The formula for calculating the category probability of the collision event is as follows:

[0040] ;

[0041] in, This represents the probability of a collision occurring with the robot. This indicates the robot's pose information for the previous time window; This represents the linear velocity magnitude of the robot's end effector within the current time window; Indicates the speed threshold; An adjustment coefficient indicating the steepness of the curve;

[0042] The formula for calculating the category probability of the scratch event is as follows:

[0043] ;

[0044] in, This indicates the probability of a scratch incident occurring on the robot. Indicates task status factor; This represents the tangential velocity component of the robot's end effector within the current time window; Represents a tiny positive number;

[0045] The formula for calculating the category probability of the joint limiting event is as follows:

[0046] ;

[0047] in, This indicates the probability of a robot experiencing a joint limitation event. Indicates the first The robot joint angles corresponding to each sensor node; Indicates the first The physical limit angle of the robot joint corresponding to each sensor node; Indicates the number of sensor nodes; This represents the joint limit sensitivity coefficient.

[0048] In some embodiments of the present invention, the event signal matrix is ​​sparsified by combining the event dictionary and the category probabilities to determine the category of sudden events in the current time window, including:

[0049] Construct an event dictionary containing typical signal patterns of different types of emergencies;

[0050] Extract potential event signal segments to be identified from the event signal matrix;

[0051] Using the reciprocal of the category probability as the weight, the potential event signal fragment is weighted sparsely encoded to obtain a sparse representation of the potential event signal fragment in the event dictionary;

[0052] Based on the sparse representation, determine the category of emergencies in the current time window.

[0053] In some embodiments of the present invention, the loss function of the weighted sparse coding is as follows:

[0054] ;

[0055] in, Represents the sparse coefficient vector to be solved; Represents the regularization parameter; express The One element; This indicates the total number of emergency categories; Represents an event dictionary; Indicates an emergency The class probability; This represents a potential event signal segment to be identified.

[0056] In some embodiments of the present invention, the method further includes: feeding back the identified sudden event category information of the current time window to the state estimator to correct the prediction results of the state estimator.

[0057] The robot posture and sudden event synchronous recognition method provided by this invention constructs a complete closed-loop system of "perception-separation-estimation-recognition-feedback". Posture estimation results guide event recognition, while event recognition results, in turn, correct posture estimation, forming a mutually reinforcing and collaborative intelligent perception framework that comprehensively improves the robot's robustness and intelligence level in perceiving its own and its environment's state. By introducing a dynamic compensation term, the slow changes in posture signals caused by robot motion are effectively compensated, avoiding misjudging normal slow movements of the robot (such as underwater robot posture adjustments and slow turns) as sparse events, significantly reducing the false alarm rate. By introducing weighted sparse coding based on posture prior probability, this invention integrates the robot's state prior information into the event classification process, making event recognition no longer an isolated pattern matching process, but an intelligent decision-making process with physical meaning and logical reasoning, greatly reducing confusion and misjudgment caused by similar signal patterns. Attached Figure Description

[0058] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other embodiments based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating a method for synchronously recognizing robot posture and sudden events according to an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0061] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0062] This invention uses an example of a Remote Operated Vehicle (ROV) equipped with distributed fiber optic sensors performing an underwater operation to illustrate the application of this method.

[0063] In order to fully perceive the attitude of the underwater robot and external emergencies, it is necessary to construct a distributed sensor network covering key structural areas. The distribution of optical fibers on the robot includes: (1) Main frame: Multiple optical fibers are arranged along the longitudinal main beam and transverse stiffeners of the underwater robot's main frame to form a sensing grid. This layout can effectively monitor the bending and torsional deformation of the underwater robot as a whole, thereby reproducing the navigation attitude (such as pitch and roll). (2) Robotic arm: Two optical fibers are symmetrically arranged along the axial direction on the surface of each link of the underwater robot's robotic arm (e.g., top and bottom). This differential arrangement can accurately measure the bending strain of the link, thereby calculating the joint torque and end load. Near the joint, the sensor density can be increased to improve the measurement accuracy. (3) Key connection parts: In the base, thruster bracket, and other parts where stress concentration is likely to occur, the optical fibers are arranged in a ring or cross pattern to monitor the load state and vibration at the connection.

[0064] The specific design method of the distributed sensor in this embodiment of the invention is as follows:

[0065] Sensor selection: Fiber Bragg gratings (FBGs) are used as the core sensing element. Multiple FBGs are connected in series on a single optical fiber using ultraviolet light writing technology. Each FBG has a unique central reflection wavelength, forming a wavelength division multiplexing sensor array.

[0066] Array design: Depending on monitoring requirements, a single optical fiber can integrate 10-30 FBG sensors with a wavelength spacing of 2-5 nanometers, operating within a 1550-nanometer communication window. The entire ROV system can deploy 4-8 such fiber optic arrays, with a total number of sensing points ranging from tens to hundreds.

[0067] Packaging and Installation: To protect the fragile optical fiber and ensure effective strain transmission, the FBG sensor is first encapsulated on a flexible substrate such as polyimide to form a patch sensor. Then, these sensor patches are firmly adhered to the pre-treated locations on the robot structure surface using a specialized epoxy resin adhesive. For cases requiring embedding in composite material structures, a pre-embedding process can be employed.

[0068] like Figure 1 As shown in the figure, this embodiment of the invention provides a method for synchronously recognizing robot posture and sudden events. Figure 1 This is a flowchart illustrating a method for synchronously recognizing robot posture and sudden events. This flowchart only shows the logical sequence of the method in this embodiment of the invention. Provided there are no conflicts, other possible embodiments of the invention may use different sequences. Figure 1 Complete the steps shown or described in the order indicated.

[0069] See Figure 1 The method of this invention specifically includes the following steps:

[0070] Step S101: Obtain robot observation signals collected by distributed sensors and construct the observation signal matrix for the current time window.

[0071] The underwater robot's fiber optic demodulator scans and acquires the center wavelength offset data of all FBGs at a high speed of 100Hz. Within a fixed-length sliding time window (e.g., 1 second, corresponding to 100 sampling points), the data is... The wavelength offset data collected by each sensor node (which has been converted into strain values ​​through calibration coefficients) are combined into an observation signal matrix.

[0072] The observed signal matrix for the current time window is represented as follows: ,in, Index for the current time window, The number of sensor nodes. This represents the number of sampling points within a time window.

[0073] Step S102: Based on the robot posture information of the previous time window, obtain the trigger probability of the sudden event in the current time window and the category probability corresponding to each type of sudden event.

[0074] Specifically, the probability of a sudden event occurring in the current time window is obtained based on the robot's posture information from the previous time window, including steps S1021 to S1023.

[0075] Step S1021: Based on the robot's posture information from the previous time window, obtain the linear velocity of the sensor node on the motion trajectory in the current time window, including:

[0076] Based on the robot's posture information from the previous time window, the robot joint angle corresponding to the sensor node in the current time window is calculated using the following formula:

[0077] ;

[0078] in, Indicates the first time window of the current time window. The sampling time of the first sampling moment The robot joint angles corresponding to each sensor node; Indicates the current time window; Indicates the previous time window; This indicates the last sampling time of the previous time window. Robot posture information corresponding to each sensor node; This indicates the number of sampling points in each time window; Indicates the first The sensor node in the current time window angular velocity at each sampling time; Indicates the sampling time interval.

[0079] In a robotic system, the robot's posture in space is continuously acquired by reading data from all sensor nodes in real time and inputting it into a kinematic model. In this embodiment of the invention, the robot posture information corresponding to a sensor node is the robot joint angle corresponding to that sensor node. Indicates the number of the previous time window The sampling time of the first sampling moment The robot joint angles corresponding to each sensor node. .

[0080] The linear velocity of the sensor node is calculated based on the robot joint angle corresponding to the sensor node, using the following formula:

[0081] ;

[0082] in, Indicates the first The sensor node in the current time window The linear velocity at each sampling time; Indicates the first The position of each sensor node relative to the Jacobian matrix of the robot joint space.

[0083] Step S1022: Calculate the risk score of the sensor node based on the linear velocity and the distance from the sensor node to the nearest environmental obstacle.

[0084] Risk scoring is used to quantify the potential triggers of sudden events. This invention comprehensively considers two factors: "intensity of movement" and "distance of interaction with the environment," thus defining the risk score as follows:

[0085] ;

[0086] in, Indicates the first The sensor node in the current time window Risk score at each sampling time; Indicates the first The sensor node in the current time window The location of each sampling time; Indicates and The nearest environmental obstacle; express and The Euclidean distance is used; the closer the distance, the higher the risk. If an environmental map is unavailable, this item can be omitted or only the distance to the structure itself can be calculated. This represents the speed weighting coefficient; This represents the distance weighting coefficient; This represents a tiny constant used to prevent the denominator from being zero.

[0087] Step S1023: Normalize the risk score to obtain the probability of triggering a sudden event at the sensor node in the current time window.

[0088] In this embodiment of the invention, the risk score is mapped to the interval (0, 1) using the Sigmoid function, and the normalized calculation formula is as follows:

[0089] ;

[0090] in, Indicates the first The sensor node in the current time window The probability of a sudden event occurring at each sampling moment; This represents a preset risk threshold, which is used when the calculated risk score is... When the risk threshold is exceeded, the probability value will rise rapidly and approach 1; This represents the sensitivity coefficient, used to control the steepness of the probability curve. The larger the threshold, the more sensitive the system is to risks exceeding the threshold.

[0091] After that Mapped to penalty weight matrix One of the elements, , This represents a tiny positive number, used to avoid the penalty weight from reaching an infinite value, thus improving the stability of subsequent optimization solutions.

[0092] For underwater robots, the categories of emergencies include collision events, scratch events, and joint limit events, and can be set to be more specific, such as frontal collisions, lateral scratches, and joint self-locking.

[0093] In this embodiment of the invention, the category probabilities corresponding to collision events, scratch events, and joint limiting events are designed, including:

[0094] The formula for calculating the category probability of a collision event is as follows:

[0095] ;

[0096] in, This represents the probability of a collision occurring with the robot. This indicates the robot's pose information for the previous time window; This represents the linear velocity magnitude of the robot's end effector within the current time window; Indicates the speed threshold; An adjustment coefficient indicating the steepness of the curve.

[0097] The formula for calculating the category probability of a scratch event is as follows:

[0098] ;

[0099] in, This indicates the probability of a scratch incident occurring on the robot. This represents the task status factor; if the robot control system is in "precision operation mode". If in "Fast Movement Mode", ; This represents the tangential velocity component of the robot's end effector within the current time window; It represents a tiny positive number.

[0100] The formula for calculating the category probability of joint limitation events is as follows:

[0101] ;

[0102] in, This indicates the probability of a robot experiencing a joint limitation event. Indicates the first The robot joint angles corresponding to each sensor node; Indicates the first The physical limit angle of the robot joint corresponding to each sensor node; Indicates the number of joints in the robot; This represents the joint limit sensitivity coefficient.

[0103] Step S103: By solving a joint optimization model that includes dynamic compensation terms and penalty weights, the observation signal matrix is ​​decomposed into a low-rank attitude signal matrix and a sparse event signal matrix.

[0104] Underwater robots possess inherently low-rank attitude signals, while external, sudden event signals exhibit typical sparsity. During normal navigation or operation, the motion of an underwater robot is dominated by a few degrees of freedom (such as forward and backward movement, surfacing and diving, and turning). The signals collected by sensors distributed throughout the robot, although high-dimensional, are not independent but rather linearly combined from these finite motion patterns. Mathematically, this manifests as a strong correlation between the time series of readings from all sensors, resulting in a very low-rank observation matrix. This low-rank property captures the high redundancy and coordinated variation of attitude signals in the spatial dimension. External events such as collisions and scratches, on the other hand, are typically local, sudden, and transient. They last only a very short time (sparser than time) and directly affect only a few nearby sensors in the spatial dimension (sparser than space). Therefore, the traces left by such events in the observation signal matrix deviate significantly from the normal background; sparsity precisely describes this "occasional and local" characteristic.

[0105] The joint optimization model, which includes dynamic compensation terms and penalty weights, is as follows:

[0106] ;

[0107] in, Represents the attitude signal matrix; Represents the event signal matrix; Represents the observed signal matrix; Represents the nuclear norm; Represents the L1 norm; Denotes the Frobenius norm; Represents the penalty weight matrix; It represents the Hadamardi (or Hadama) stack; , and Represents the regularization parameter; This represents the dynamic compensation term, which constrains the attitude signal matrix within the current time window. The attitude signal matrix obtained from the previous time window The continuity of the signal is maintained, thereby effectively compensating for the slow changes in attitude signals caused by robot motion.

[0108] In this embodiment of the invention, the joint optimization model is solved using the Alternating Direction Multiplier Method (ADMM) algorithm to obtain... and The specific steps are as follows:

[0109] Step 3.1: Introduce auxiliary variables , Define the augmented Lagrange function:

[0110] ;

[0111] in , For Lagrange multiplier matrices, Let be the penalty function; let the iteration count be... .

[0112] Step 3.2: Update :

[0113] ;

[0114] Step 3.3: Update :

[0115] ;

[0116] Step 3.4: Update :

[0117] ;

[0118] Step 3.5: Update :

[0119] ;

[0120] Step 3.6: Update the Lagrange multipliers:

[0121] ;

[0122] ;

[0123] Step 3.7: Convergence check. Stop iteration when the following conditions are met simultaneously:

[0124] ;

[0125] in, This is the preset tolerance threshold.

[0126] If the convergence condition is not met, then let Then proceed to step 3.2 to continue iterative updates.

[0127] Step S104: Based on the attitude signal matrix, obtain the robot attitude in the current time window.

[0128] Smooth strain signals are extracted from the low-rank attitude signal matrix; the smooth strain signals are converted into robot attitude information through a pre-calibrated strain-attitude mapping model; the attitude information is input into the state estimator to obtain the optimal robot attitude estimate for the current time window.

[0129] In this embodiment of the invention, the state estimator uses an extended Kalman filter (EKF).

[0130] Step S105: Combine the event dictionary and category probabilities to perform sparse optimization on the event signal matrix to determine the category of sudden events in the current time window, specifically including steps S1051 to S1054.

[0131] Step S1051: Construct an event dictionary containing typical signal patterns of different types of emergencies.

[0132] Through prior experiments or simulations, waveforms of corresponding standard strain modes generated by the FBG sensor array during different types of sudden events are collected to construct an event dictionary. The event dictionary is represented as follows: Each atom represents a typical response pattern on the sensor to a type of sudden event (such as collision or scratch).

[0133] Step S1052: From the event signal matrix Extract potential event signal fragments to be identified ( The List).

[0134] In a sparse event signal matrix, if the energy of a certain column exceeds a preset threshold, it is determined to be a potential event.

[0135] Step S1053: Using the reciprocal of the class probability as the weight, perform weighted sparse coding on the latent event signal fragment to obtain the sparse representation of the latent event signal fragment on the event dictionary.

[0136] The loss function for weighted sparse coding is as follows:

[0137] ;

[0138] in, Represents the sparse coefficient vector to be solved; Represents the regularization parameter; express The One element; This indicates the total number of emergency categories; Represents an event dictionary; Indicates an emergency The class probability; This represents a potential event signal segment to be identified.

[0139] In this embodiment of the invention, the steps for solving the loss function are as follows:

[0140] Step 5.1: Initialize the sparse coefficient vector residual Iteration counting .

[0141] Step 5.2: For each event type Calculate the gradient:

[0142] ;

[0143] Step 5.3: For each event type Update the soft threshold:

[0144] ;

[0145] in For the event The weight, This is a soft threshold function.

[0146] Step 5.4: Convergence judgment, let the convergence threshold be... ,like If the iteration stops, output the result. The optimal estimate. Conversely, let Then return to step 5.2 to continue the iteration.

[0147] Step S1054: Determine the category of sudden events in the current time window based on the sparse representation.

[0148] Sparse representation means that the vast majority of emergencies in the dictionary have a weight of 0 or close to 0, while only a few related emergencies have a larger weight, which can be used to determine the category of emergencies in the current time window.

[0149] The system reports the type of emergency and its location within the current time window to the ROV's autonomous control system, which can then immediately execute avoidance or protective shutdown actions.

[0150] At the same time, this information about the sudden event can be used to update the state estimator. For example, an unexpected collision means the action of an external force, which can be used to correct the estimates of velocity and position.

[0151] The current calculation For the next time window The calculations are performed in a loop throughout the entire process.

[0152] An embodiment of the present invention also provides a non-transitory machine-readable medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform the robot posture and sudden event synchronization recognition method of the embodiments of the present invention.

[0153] An embodiment of the present invention also provides a computer program product, including a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform the robot posture and sudden event synchronization recognition method of the embodiments of the present invention.

[0154] An embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the robot posture and sudden event synchronization recognition method of the present invention.

[0155] refer to Figure 2 The present invention will now describe a structural block diagram of an electronic device that can serve as an embodiment of the present invention, serving as an example of a hardware device applicable to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0156] like Figure 2 As shown, the electronic device includes a computing unit 101, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 102 or a computer program loaded from a storage unit 108 into a random access memory (RAM) 103. The RAM 103 may also store various programs and data required for the operation of the electronic device. The computing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0157] Multiple components in the electronic device are connected to I / O interface 105, including: input unit 106, output unit 107, storage unit 108, and communication unit 109. Input unit 106 can be any type of device capable of inputting information into the electronic device. Input unit 106 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 107 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 108 may include, but is not limited to, disks and optical discs. Communication unit 109 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, and / or wireless communication transceivers, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0158] The computing unit 101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 101 include, but are not limited to, CPUs, graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 101 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention can be implemented as computer programs tangibly contained in a machine-readable medium, such as storage unit 108. In some embodiments, part or all of the computer program can be loaded and / or installed on an electronic device via ROM 102 and / or communication unit 109. In some embodiments, the computing unit 101 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0159] Computer programs for implementing the methods of embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0160] In the context of embodiments of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, or infrared systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0161] It should be noted that the term "comprising" and its variations used in the embodiments of this invention are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "a plurality" mentioned in the embodiments of this invention are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0162] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this invention are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0163] The steps described in the method embodiments provided by the present invention can be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0164] The term "embodiment" in this specification refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily imply the same embodiment, nor does it imply independence or alternativeity from other embodiments. The various embodiments in this specification are described in a related manner, with reference to each other for similar or identical parts. In particular, for apparatus, device, and system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and relevant details are referred to in the description of the method embodiments.

[0165] The above embodiments merely illustrate several implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for synchronously recognizing robot posture and sudden events, characterized in that, include: Acquire robot observation signals collected by distributed sensors and construct the observation signal matrix for the current time window; Based on the robot's posture information from the previous time window, obtain the trigger probability of a sudden event occurring in the current time window and the category probability corresponding to each type of sudden event; By solving a joint optimization model that includes a dynamic compensation term and penalty weights, the observed signal matrix is ​​decomposed into a low-rank attitude signal matrix and a sparse event signal matrix; wherein, the dynamic compensation term is used to constrain the continuity of the attitude signal matrix in adjacent time windows; and the penalty weights are obtained by mapping the trigger probability. Based on the attitude signal matrix, the robot attitude in the current time window is obtained; By combining the event dictionary and the category probabilities, sparse optimization is performed on the event signal matrix to determine the category of sudden events in the current time window; The joint optimization model is as follows: ； in, Represents the attitude signal matrix; Represents the event signal matrix; Represents the observed signal matrix; Represents the nuclear norm; Represents the L1 norm; Denotes the Frobenius norm; Represents the penalty weight matrix; It represents the Hadamardi (or Hadama) stack; , and Represents the regularization parameter; This indicates a dynamic compensation item.

2. The method for synchronously recognizing robot posture and sudden events according to claim 1, characterized in that, Based on the robot's posture information from the previous time window, obtain the probability of triggering a sudden event in the current time window, including: Based on the robot's posture information from the previous time window, the linear velocity of the sensor node on the motion trajectory in the current time window is obtained; Based on the linear velocity and the distance from the sensor node to the nearest environmental obstacle, a risk score for the sensor node is calculated. The risk score is normalized to obtain the probability of a sudden event occurring at the sensor node in the current time window.

3. The method for synchronously recognizing robot posture and sudden events according to claim 2, characterized in that, Based on the robot's posture information from the previous time window, the linear velocity of the sensor node on the motion trajectory in the current time window is obtained, including: Based on the robot's posture information from the previous time window, the robot joint angle corresponding to the sensor node in the current time window is calculated using the following formula: ； in, Indicates the first time window of the current time window. The sampling time of the first sampling moment The robot joint angles corresponding to each sensor node; Indicates the current time window; Indicates the previous time window; This indicates the last sampling time of the previous time window. Robot posture information corresponding to each sensor node; This indicates the number of sampling points in each time window; Indicates the first The sensor node in the current time window angular velocity at each sampling time; Indicates the sampling time interval; Among them, the robot posture information corresponding to the sensor node is the robot joint angle corresponding to the sensor node; The linear velocity of the sensor node is calculated based on the robot joint angle corresponding to the sensor node, using the following formula: ； in, Indicates the first The sensor node in the current time window The linear velocity at each sampling time; Indicates the first The position of each sensor node relative to the Jacobian matrix of the robot joint space.

4. The method for synchronously recognizing robot posture and sudden events according to claim 3, characterized in that, The formula for calculating the risk score of the sensor node is as follows: ； in, Indicates the first The sensor node in the current time window Risk score at each sampling time; Indicates the first The sensor node in the current time window The location of each sampling time; Indicates and The nearest environmental obstacle; express and The distance; This represents the speed weighting coefficient; This represents the distance weighting coefficient; Represents a tiny constant; The normalized calculation formula for the risk score is as follows: ； in, Indicates the first The sensor node in the current time window The probability of a sudden event occurring at each sampling moment; This indicates a preset risk threshold; This represents the sensitivity coefficient.

5. The method for synchronously recognizing robot posture and sudden events according to claim 4, characterized in that, The penalty weight matrix The calculation formula is as follows: ； in, It represents a tiny positive number.

6. The method for synchronously recognizing robot posture and sudden events according to claim 1, characterized in that, Based on the attitude signal matrix, the robot attitude for the current time window is obtained, including: Extract the smoothed strain signal from the low-rank attitude signal matrix; The smooth strain signal is converted into the robot's posture information using a pre-calibrated strain-attitude mapping model. The attitude information is input into the state estimator to obtain the optimal robot attitude estimate for the current time window.

7. The method for synchronously recognizing robot posture and sudden events according to claim 3, characterized in that, The emergency incident categories include collision incidents, scratch incidents, and joint limitation incidents; Obtain the category probability for each type of emergency event within the current time window, including: The formula for calculating the category probability of the collision event is as follows: ； in, This represents the probability of a collision occurring with the robot. This indicates the robot's pose information for the previous time window; This represents the linear velocity magnitude of the robot's end effector within the current time window; Indicates the speed threshold; An adjustment coefficient indicating the steepness of the curve; The formula for calculating the category probability of the scratch event is as follows: ； in, This indicates the probability of a scratch incident occurring on the robot. Indicates task status factor; This represents the tangential velocity component of the robot's end effector within the current time window; Represents a tiny positive number; The formula for calculating the category probability of the joint limiting event is as follows: ； in, This represents the probability of a robot experiencing a joint limitation event. Indicates the first The robot joint angles corresponding to each sensor node; Indicates the first The physical limit angle of the robot joint corresponding to each sensor node; Indicates the number of sensor nodes; This represents the joint limit sensitivity coefficient.

8. The method for synchronously recognizing robot posture and sudden events according to claim 1, characterized in that, By combining the event dictionary and the category probabilities, sparse optimization is performed on the event signal matrix to determine the category of sudden events in the current time window, including: Construct an event dictionary containing typical signal patterns of different types of emergencies; Extract potential event signal segments to be identified from the event signal matrix; Using the reciprocal of the category probability as the weight, the potential event signal fragment is weighted sparsely encoded to obtain a sparse representation of the potential event signal fragment in the event dictionary; Based on the sparse representation, determine the category of emergencies in the current time window.

9. The method for synchronously recognizing robot posture and sudden events according to claim 8, characterized in that, The loss function of the weighted sparse coding is as follows: ； in, Represents the sparse coefficient vector to be solved; Represents the regularization parameter; express The One element; This indicates the total number of emergency categories; Represents an event dictionary; Indicates an emergency The class probability; This represents a potential event signal segment to be identified.

10. The method for synchronously recognizing robot posture and sudden events according to claim 6, characterized in that, Also includes: The identified event category information for the current time window is fed back to the state estimator to correct the state estimator's prediction results.

Images