Data augmentation method and system based on attractors

Abstract

The embodiment of the invention provides a data augmentation method and system based on attractors, and belongs to the technical field of artificial intelligence. The method comprises the steps of obtaining action data; acquiring a mode action according to the action data; determining a balance point or a limit cycle according to the mode action, determining a series of balance points for non-repetitive actions, and determining a limit cycle for repetitive mode actions; according to the balance point or the limit cycle, a kinetic equation is constructed, the kinetic equation is divided into a tangential direction and a normal direction, and the tangential direction and the normal direction are evolution of the augmented data sampling points in the tangential direction and the normal direction of the mode action respectively; and completing data augmentation according to the kinetic equation. According to the method, data augmentation can be effectively carried out, more multivariate data support is provided for imitation learning, and the efficiency and effect of the pre-training stage are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an attractor-based data augmentation method and system. Background Art

[0002] In robot reinforcement learning, imitation learning can be used for pre-training to accelerate the convergence of strategies. However, existing imitation learning techniques have the following limitations in data augmentation:

[0003] Limited dataset size: The high cost of obtaining expert examples leads to limited dataset size and cannot effectively cover the wide state space that the robot may encounter.

[0004] Insufficient data diversity: Existing imitation learning methods tend to rely on a single expert path, resulting in weak model generalization ability and difficulty in adapting to dynamic or unseen environments.

[0005] Overfitting problem: With only a few expert examples, the model tends to overfit to specific action sequences, which affects its performance in new environments.

[0006] Insufficient generalization ability: Existing imitation learning methods can often only adapt to specific environments and are difficult to effectively transfer in changing environments or multi-task scenarios.

[0007] Low learning efficiency: A large number of expert examples are required to effectively learn complex strategies, especially in high-dimensional perception tasks (such as vision and speech), where the efficiency of imitation learning decreases significantly.

[0008] Difficulty in acquiring expert data: Robot imitation learning relies on expert demonstrations, but acquiring a large amount of high-quality expert data is costly.

[0009] Insufficient exploration of the state space: The pre-training stage of traditional imitation learning often only covers local areas of the state space, resulting in poor performance of the model when faced with new tasks.

[0010] Sample size limits generalization: The lack of sufficient and diverse example data limits the model's generalization ability. Summary of the invention

[0011] The main purpose of the embodiments of the present application is to provide a data augmentation method and system based on an attractor.

[0012] The technical solution adopted by the present invention is:

[0013] On the one hand, an embodiment of the present invention provides a data augmentation method based on an attractor, the method comprising the following steps:

[0014] Get action data;

[0015] According to the action data, obtaining a mode action;

[0016] Determining an equilibrium point or a limit cycle according to the mode action;

[0017] constructing a dynamic equation according to the equilibrium point or the limit cycle;

[0018] According to the kinetic equation, data augmentation is performed.

[0019] Furthermore, the step of obtaining the action data comprises the following steps:

[0020] The motion track of the target object is collected by using a motion capture device to obtain motion data; the motion capture device includes an optical tracking system, an inertial sensor or a depth camera.

[0021] Further, the acquiring of the mode action according to the action data comprises the following steps:

[0022] From the motion data, multiple pattern actions are selected based on robot kinematics and task requirement information.

[0023] Further, the determining of the equilibrium point or limit cycle according to the mode action includes:

[0024] Integrate multiple pattern actions to generate a target action sequence, forming an equilibrium point or a limit cycle.

[0025] Furthermore, constructing a dynamic equation according to the equilibrium point or the limit cycle comprises the following steps:

[0026] Determine the number of sampling points;

[0027] Determine the action space dimensions;

[0028] According to the dimension of the action space and the number of sampling points of the augmented data, a number of current state points are obtained;

[0029] Calculate the distance between each current state point and each original data point on the limit cycle to obtain the original data point closest to the limit cycle corresponding to each current state point;

[0030] According to the current state point and the original data point closest to the limit cycle, a dynamic equation is obtained. Further, according to the dynamic equation, data augmentation is completed, including the following steps:

[0031] Setting an iteration step size according to the distance between the current state point and the nearest original data point on the distance limit cycle;

[0032] The current state point is updated according to the dynamic equation and the iteration step to complete data augmentation.

[0033] On the other hand, an embodiment of the present invention further provides an attractor-based data augmentation system, the system comprising:

[0034] The first module is used to obtain action data;

[0035] The second module is used to obtain a mode action according to the action data;

[0036] A third module is used to determine the equilibrium point or limit cycle according to the mode action;

[0037] A fourth module is used to construct a dynamic equation according to the equilibrium point or the limit cycle;

[0038] The fifth module is used to complete data augmentation according to the kinetic equation.

[0039] Furthermore, the attractor-based data augmentation system of the embodiment of the present invention further includes:

[0040] An attractor construction module, used for constructing the attractor according to the action data;

[0041] Data augmentation module, used to sample in the action space and generate corresponding state trajectories;

[0042] The multi-attractor parallel mechanism module is used to utilize multiple attractor regions to generate corresponding state trajectories and action sequences.

[0043] On the other hand, an embodiment of the present invention further provides an attractor-based data augmentation device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the attractor-based data augmentation method as described above is implemented.

[0044] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the attractor-based data augmentation method as described above.

[0045] The embodiments of the present application include at least the following beneficial effects: The present application provides a data augmentation method and system based on an attractor. The present invention obtains action data; based on the action data, obtains pattern actions; based on the pattern actions, determines equilibrium points or limit cycles, wherein for non-repetitive actions, a series of equilibrium points are determined, and for repetitive pattern actions, limit cycles are determined; based on the equilibrium points or limit cycles, a dynamic equation is constructed, wherein the dynamic equation is divided into tangential and normal directions, which are respectively the evolution of the action sampling points in the tangential and normal directions of the pattern actions; based on the dynamic equation, data augmentation is completed. The present invention can effectively perform data augmentation, provide more diverse data support for imitation learning, and greatly improve the efficiency and effect of the pre-training stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of an attractor-based data augmentation method provided by an embodiment of the present invention;

[0047] Figure 2 is a schematic diagram of an attractor construction process provided by an embodiment of the present invention;

[0048] Figure 3 is an attractor effect diagram of each leg of a quadruped robot provided by an embodiment of the present invention;

[0049] Figure 4 is a schematic diagram of a data augmentation module process provided by an embodiment of the present invention;

[0050] Figure 5 This is a rendering of a current state point of each leg of a quadruped robot provided by an embodiment of the present invention iterated 20 times;

[0051] Figure 6 This is a rendering of 100 iterations of 20 current state points of one leg of a quadruped robot provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0053] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0054] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0056] The present invention takes into account that attractor theory is widely used in dynamic systems and chaotic systems to describe the behavior of system states tending to a specific region over time. The attractor-based augmentation method can effectively generate diverse samples, thereby providing more pre-training data for imitation learning and improving the generalization and learning efficiency of the model.

[0057] The present invention aims to solve the problems of insufficient data diversity, insufficient state space exploration and insufficient generalization ability in existing imitation learning through attractor-based data augmentation methods. By generating diverse samples through the attractor system, this method provides a richer expert data set for the pre-training stage of robot imitation learning, thereby improving the generalization ability and learning efficiency of the model.

[0058] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0059] On the one hand, an embodiment of the present invention provides a data augmentation method based on an attractor, referring to Figure 1 , the method comprises the following steps:

[0060] S100, obtaining action data;

[0061] S200, obtaining a mode action according to the action data;

[0062] S300, determining an equilibrium point or a limit cycle according to the mode action;

[0063] S400, constructing dynamic equations according to equilibrium points or limit cycles;

[0064] S500: Complete data augmentation based on the kinetic equation.

[0065] The embodiment of the present invention discloses step S100 of acquiring action data, including the following steps:

[0066] S110, using a motion capture device to collect a motion trajectory of the target object to obtain motion data; the motion capture device includes an optical tracking system, an inertial sensor, or a depth camera.

[0067] The embodiment of the present invention discloses that step S200 obtains a mode action according to the action data, including the following steps:

[0068] S210: Select multiple mode actions from the action data according to the robot kinematics and task requirement information.

[0069] The embodiment of the present invention discloses that step S300 determines the equilibrium point or limit cycle according to the mode action, including:

[0070] S310, integrating multiple pattern actions to generate a target action sequence to form an equilibrium point or a limit cycle.

[0071] The embodiment of the present invention discloses that step S400 constructs a dynamic equation according to an equilibrium point or a limit cycle, including the following steps:

[0072] S410, determining the number of sampling points;

[0073] S420, determining the action space dimension;

[0074] S430, obtaining a number of current state points according to the action space dimension and the number of action sampling points;

[0075] S440, calculating the distance between each current state point and each original data point on the limit cycle, and obtaining the original data point closest to the limit cycle corresponding to each current state point;

[0076] S450, obtaining a dynamic equation according to the current state point and the original data point closest to the limit cycle.

[0077] The embodiment of the present invention discloses that step S500 completes data augmentation according to the kinetic equation, including the following steps:

[0078] S510, setting an iteration step size according to the distance between the current state point and the nearest original data point on the distance limit cycle;

[0079] S520, update the current state point according to the dynamic equation and the iteration step size to complete data augmentation.

[0080] As an optional embodiment, the present invention includes:

[0081] 1. Attractor building blocks

[0082] Obtain motion data: Use motion capture equipment (such as optical tracking systems, inertial sensors, or depth cameras) to collect the motion trajectory of the target object and record relevant data.

[0083] Determine pattern actions: From the captured data, select multiple pattern actions based on robot kinematics and task requirements to form equilibrium points or limit cycles.

[0084] Determine equilibrium points or limit cycles: For non-repetitive actions, determine the series of equilibrium points; for repetitive pattern actions, determine the limit cycles as the robot's final behavior trajectory.

[0085] 2. Data augmentation module

[0086] Determine parameters: set the number of action sampling points m and the maximum number of iterations t.

[0087] Randomly generate action space sampling points:

[0088] Randomly select action sampling point a1 from the entire action space, and the number of samples is m1.

[0089] An action sampling point a2 is randomly selected within three times the range of the action data, and the number of samples is m2.

[0090] The original action data selects action sampling point a3, and the number of samples is m-m1-m2.

[0091] Get the nearest original data point on the limit cycle: the number of original data points is n, and calculate the distance l between each current state point and each original data point on the limit cycle i (i=1,2,…,n), determine the nearest original data point p on the limit cycle.

[0092] Determine the kinetic equation:

[0093] Calculate the normal velocity q, the current state point is in the direction of the nearest original data point on the limit cycle.

[0094] Calculate the tangential velocity t, the direction of the nearest original data point on the limit cycle itself.

[0095] The dynamic equation is the normal vector plus the tangent vector.

[0096] Determine the step size: based on the distance l between the current state point and the nearest original data point on the limit cycle min , set the iteration step size s.

[0097] Perform data augmentation generation:

[0098] Update the sampling points according to the dynamic equation and step size s.

[0099] The iterative process is repeated until the maximum number of iterations t is reached.

[0100] 3. Multi-attractor parallel mechanism

[0101] At the same time, multiple attractor regions are used to generate corresponding state trajectories and action sequences to ensure coverage of a variety of complex scenarios and enhance the generalization ability of the model.

[0102] On the other hand, an embodiment of the present invention further provides a data augmentation system based on an attractor, the system comprising:

[0103] The first module is used to obtain action data;

[0104] The second module is used to obtain the mode action according to the action data;

[0105] The third module is used to determine the equilibrium point or limit cycle according to the mode action;

[0106] The fourth module is used to construct dynamic equations based on equilibrium points or limit cycles;

[0107] The fifth module is used to complete data augmentation based on the kinetic equation.

[0108] The attractor-based data augmentation system provided by an embodiment of the present invention further includes:

[0109] An attractor construction module, used to construct attractors based on action data;

[0110] Data augmentation module, used to sample in the action space and generate corresponding state trajectories;

[0111] The multi-attractor parallel mechanism module is used to utilize multiple attractor regions to generate corresponding state trajectories and action sequences.

[0112] As an optional implementation, the present invention proposes a data augmentation method based on attractor theory to address the problems of poor generalization ability, low learning efficiency, and limited data size of existing imitation learning technologies, which is suitable for the pre-training stage of robot imitation learning. By utilizing the dynamic characteristics of attractors to generate diverse state trajectories and action sequences, the present invention can provide rich training data for imitation learning, thereby improving the performance of the model in complex environments.

[0113] The present invention first extracts representative data from the expert's demonstration data and transforms it into an attractor region. The attractor not only contains the expert's original trajectory, but also generates more diverse approximate expert behaviors through perturbations or random sampling within the attractor boundary. This attractor generation mechanism based on expert behavior can provide a wider range of state space exploration.

[0114] The system includes the following main modules:

[0115] Attractor building module: builds attractors based on action data.

[0116] Data augmentation module: Sample some samples in the action space and generate corresponding state trajectories.

[0117] Multi-attractor parallel mechanism: The present invention uses multiple attractor systems to generate data in parallel, and each attractor area corresponds to a different behavior strategy of the robot. The multi-attractor parallel mechanism ensures that the generated data can cover a variety of complex scenarios that the robot may face, avoiding the model's excessive reliance on a single trajectory in pre-training.

[0118] refer to Figure 2 , attractor building module process:

[0119] Attractors are stable regions in a dynamic system, and the system state converges to these regions. The present invention uses attractors to generate diverse samples, mainly including the following steps:

[0120] 1. Obtain motion data: Use motion capture equipment (such as optical tracking systems, inertial sensors, or depth cameras) to capture the motion trajectory of the target object and record relevant data.

[0121] 2. Determine the pattern action: Select the pattern action from the motion data based on the robot kinematics and task requirements.

[0122] 3. Determine the equilibrium point or limit cycle: For non-repetitive actions, determine the series of equilibrium points; for repetitive pattern actions, determine the limit cycle as the robot's final behavior trajectory. Figure 3 The attractor effect of each leg of the quadruped robot.

[0123] refer to Figure 4 , data augmentation module process:

[0124] 1. Determine the number of action sampling points and the maximum number of iterations.

[0125] 2. Randomly generate action space sampling points: randomly select action sampling points in the entire action space, randomly select action sampling points within three times the range of the motion capture data, and the remaining action sampling points are the original action data points, which serve as samples for this batch.

[0126] 3. Obtain the nearest original data point on the limit cycle: After obtaining the action sampling point, calculate the distance between each current state point and each original data point on the limit cycle to determine the nearest original data point on the limit cycle.

[0127] 4. Determine the dynamic equation of the current state point: The dynamic equation consists of two parts. The first part is the normal velocity, that is, the velocity of the current state point toward the nearest original data point on the limit cycle. The second part is the tangential velocity, that is, the velocity of the nearest original data point on the limit cycle itself.

[0128] 5. Determine the step size of each iteration: Determine the iteration step size based on the distance to the nearest original data point on the limit cycle.

[0129] 6. Data augmentation generation based on attractors: According to the dynamic equation and iteration step, perform one iteration to obtain a data augmentation, that is, update the current state point, and repeat process 4 and process 5 until the maximum number of iterations is reached. Figure 5 This is the effect diagram of 20 iterations of the current state point of each leg of the quadruped robot. Figure 6 The effect diagram of 100 iterations of the 20 current state points of one leg of a quadruped robot.

[0130] As an optional implementation, the data augmentation module process of the embodiment of the present invention is as follows:

[0131] 1. Determine the number of action sampling points m and the maximum number of iterations t.

[0132] 2. Randomly select the initial sampling point of the action space: randomly select the action sampling point a1 in the entire action space, the number of samples is m1, randomly select the initial sampling point a2 within three times the range of the action data, the number of samples is m2, and the remaining action sampling points are the original action data a3, the number of samples is m-m1-m2. In one embodiment, m1 and m2 can be m / 2 and m / 4 respectively. a1, a2, a3, etc. are samples of this batch.

[0133] 3. Get the nearest original data point on the limit cycle: The number of original data points is n. After obtaining the action sampling points, calculate the distance l between each current state point a (including a1, a2, a3, etc.) and each original data point on the limit cycle. i (i=1,2,…,n), where the current state point is closest to the original data point on the limit cycle is p, and the distance between the current state point and p is l p .

[0134] 4. According to the dynamic equation, determine the speed of the current state point: The dynamic equation consists of two parts. The first part is the normal velocity q, that is, the speed of the current state point toward the nearest original data point on the limit cycle. The second part is the tangential velocity t, that is, the direction of the nearest original data point on the limit cycle itself.

[0135] 5. Calculate the new state point: According to the dynamic equation and the iteration step s, perform one iteration to obtain a data augmentation point a′, a′=a+(q+t)s, and update the current state point.

[0136] 6. Attractor-based data augmentation generation: Repeat process 4 and process 5 until the maximum number of iterations t is reached.

[0137] Advantages of the technology of the present invention:

[0138] Significantly enhanced data diversity: Diversified samples generated based on attractors greatly improve the richness of the dataset and avoid overfitting to a single expert behavior.

[0139] Improve the efficiency of imitation learning: The attractor-enhanced dataset enables the imitation learning model to be exposed to more different behavior patterns during the pre-training stage, thereby accelerating the learning and optimization of the strategy.

[0140] Enhanced generalization ability: Through the diversified generation of attractor regions, the present invention effectively improves the generalization ability of the model in actual tasks, enabling it to cope with changes in different environments and tasks.

[0141] Reduced data dependence: Reduced dependence on large-scale expert data and expanded the coverage of the dataset through an attractor generation mechanism based on a small number of expert examples.

[0142] The key technologies of the present invention are:

[0143] Determine the normal velocity q and tangential velocity t in the dynamic equations.

[0144] The number of original data points is a scalar n, and the displacement of the current state point a from each original data point on the limit cycle is described as a loss d i (i=1,2,…,n), the distance between the current state point a and each original data point on the limit cycle is described as a scalar l i (i=1,2,…,n), where the current state point is closest to the original data point on the limit cycle is p, and the displacement of the current state point a from p is vector d p , the distance is a scalar l p , the velocity of each original data point on the limit cycle is described by the vector v i (i=1,2,…,n),||v i || represents vector v iThe 2-norm size of ||v|| max , the velocity of p is vector v p ,||v p || represents vector v p The 2-norm size, v p ′ is v p The unit vector, that is, v p ′=v p / ||v p In one embodiment, the distance threshold l t is a scalar, l t =60||v p ||, constant c is a scalar, c=0.001, constant factor μ is a scalar, μ=2000, coefficient k is a scalar, k=10, weight ω i is a scalar, ω i =||v i || / ||v|| max (i=1,2,…,n).

[0145] Normal velocity q is the normal velocity u multiplied by the normal unit vector n', that is, q = un'. Where:

[0146] The normal velocity u is the constant factor μ divided by the ω corresponding to each equilibrium point i / l i The sum of , and then divided by the number of equilibrium points n, that is

[0147] The calculation method of the normal unit vector n′ is the same as l p The size of

[0148] When l p Greater than or equal to l t hour:

[0149] First, calculate the attraction g of the i-th equilibrium point to the current state point a i , the step is weight ω i Multiply by the displacement d from the ith equilibrium point to the next equilibrium point i , and then divided by the distance l from the current state point to the i-th equilibrium point i The square of g i =ω i d i / l i 2 The normal unit vector n′ is the normalized sum of the attractive forces at each point, that is,

[0150] When l p Less than l t hour:

[0151] d p In v p The projection on is j p , i.e. j p =d p v p '. The normal unit vector n' is d p -j p After normalization, that is, n′=(d p -j p ) / ||d p -j p ||.

[0152] The calculation of the tangential velocity q needs to consider the distance between the current state point a and each equilibrium point and the velocity of each equilibrium point.

[0153] Drag force τ i is the velocity v of the ith equilibrium point i Divide by the square of the distance from the current state point to the equilibrium point, and then divide by the kl corresponding to the equilibrium point i +1, which is τ i =v i / (l i 2 (kl i +1)). The total drag force τ is the sum of the drag component forces, that is, The equivalent distance r is 1 / l corresponding to each equilibrium point i 2 The sum of Drag speed v t is the total drag force divided by the equivalent distance, or v t =τ / r.

[0154] When l p Greater than c and greater than l t hour:

[0155] The tangential velocity is the drag velocity, i.e. t = v t

[0156] When l p Greater than c and l p Less than l t hour:

[0157] The tangential velocity is the drag velocity v t The velocity v at the closest point p The drag force divided by the distance is v p The projection size on is b, that is, b = v t v p '. The tangential velocity t is b times v p The unit vector v p ′, that is, t = bvp ′.

[0158] When l p When less than c:

[0159] The tangential velocity is the velocity v of the closest point p , t = v p .

[0160] The protection points of the present invention are:

[0161] Pattern motion extraction and equilibrium point or limit cycle generation: Extract pattern motion from motion data and construct equilibrium points or limit cycles.

[0162] Multi-attractor parallel generation mechanism: Generate data in parallel in multiple attractor regions, improving data generation efficiency and coverage of the state space.

[0163] Modular Data Augmentation System: A modular data augmentation system is designed, including sampling point generation and data augmentation modules, to ensure efficient operation of the system and enhance learning effects and pre-training data sets for imitation learning.

[0164] Calculation of dynamic equations: including calculation of normal vector direction and tangent vector direction.

[0165] The data augmentation method based on attractor theory proposed in the present invention is particularly suitable for the pre-training stage of robot imitation learning. By generating diverse samples, the model can learn strategies more effectively, avoid overfitting, and improve generalization ability. This method can effectively perform data augmentation, provide more diverse data support for imitation learning, and effectively improve the efficiency and effect of the pre-training stage.

[0166] On the other hand, an embodiment of the present invention further provides an attractor-based data augmentation device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the attractor-based data augmentation method as described above is implemented.

[0167] The attractor-based data augmentation device according to the embodiment of the present invention includes a memory and a processor.

[0168] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0169] The memory may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage. Among them, ROM can store static data or instructions required by the processor or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at run time. In addition, the memory may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0170] The memory stores executable codes, and when the executable codes are processed by the processor, the processor can execute part or all of the methods described above.

[0171] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions for causing a computer to execute the attractor-based data augmentation method as described above.

[0172] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0173] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A data augmentation method based on attractors, characterized in that: The method comprises the following steps: Get action data; According to the action data, obtaining a mode action; Determining an equilibrium point or a limit cycle according to the mode action; constructing a dynamic equation according to the equilibrium point or the limit cycle; According to the kinetic equations, data augmentation is performed.

2. The method according to claim 1, characterized in that The step of obtaining the action data comprises the following steps: The motion track of the target object is collected by using a motion capture device to obtain motion data; the motion capture device includes an optical tracking system, an inertial sensor or a depth camera.

3. The method according to claim 1, characterized in that The step of obtaining the mode action according to the action data comprises the following steps: From the motion data, multiple pattern actions are selected based on robot kinematics and task requirement information.

4. The method according to claim 1, characterized in that Determining the equilibrium point or limit cycle according to the mode action includes: Integrate multiple pattern actions to generate a target action sequence, forming an equilibrium point or a limit cycle.

5. The method according to claim 1, characterized in that The constructing of the dynamic equation according to the equilibrium point or the limit cycle comprises the following steps: Determine the number of action sampling points; Determine the action space dimensions; According to the dimension of the action space and the number of action sampling points, a number of current state points are obtained; Calculate the distance between each current state point and each original data point on the limit cycle to obtain the original data point closest to the limit cycle corresponding to each current state point; A dynamic equation is obtained according to the current state point and the original data point closest to the limit cycle.

6. The method according to claim 5, characterized in that According to the kinetic equation, data augmentation is performed, comprising the following steps: Setting an iteration step size according to the distance between the current state point and the nearest original data point on the distance limit cycle; The current state point is updated according to the dynamic equation and the iteration step to complete data augmentation.

7. An attractor-based data augmentation system, characterized in that: The system comprises: The first module is used to obtain action data; The second module is used to obtain a mode action according to the action data; A third module is used to determine the equilibrium point or limit cycle according to the mode action; A fourth module is used to construct a dynamic equation according to the equilibrium point or the limit cycle; The fifth module is used to complete data augmentation according to the kinetic equation.

8. The system according to claim 7, characterized in that The system further comprises: An attractor construction module, used for constructing the attractor according to the action data; Data augmentation module, used to sample in the action space and generate corresponding state trajectories; The multi-attractor parallel mechanism module is used to utilize multiple attractor regions to generate corresponding state trajectories and action sequences.

9. A data augmentation device based on an attractor, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the attractor-based data augmentation method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the attractor-based data augmentation method according to any one of claims 1 to 6.

Images