Method for predicting life of battery of humanoid robot and storage medium

Abstract

The invention relates to a method for predicting the service life of a battery of a humanoid robot and a storage medium, and the method comprises the following steps: carrying out the multi-source data collection of the robot, and the multi-source data collection comprises the steps of battery parameter collection, use environment perception, task load recording and charging behavior analysis; carrying out preprocessing and feature extraction on the collected multi-source data; and outputting a prediction result according to the extracted features through a deep learning model architecture. Multi-source data is combined with a deep learning model to predict the battery life condition of the robot, the precision of a battery life prediction result is improved, and meanwhile, the method can adapt to different types of robots and variable use environments.

Description

Technical Field

[0001] The present application relates to the field of robotics technology, and in particular to a method for predicting the battery life of a humanoid robot and a storage medium. Background Art

[0002] Currently, the main methods for predicting robot battery power include:

[0003] (1) Prediction method based on battery discharge curve: By analyzing the battery discharge curve, a mathematical model is established to predict the remaining power. This method is more accurate in a static environment, but it is difficult to adapt to complex and changeable actual usage scenarios.

[0004] (2) Prediction method based on machine learning: Use historical data to train machine learning models, such as support vector machines (SVM) or random forests, to predict future battery power. This method can handle more complex data relationships, but has limited ability to handle long-term predictions and emergencies.

[0005] (3) Prediction method based on Kalman filter: The battery power is regarded as a dynamic system, and the Kalman filter algorithm is used to estimate and predict the battery status. This method can update the prediction results in real time, but its performance in nonlinear systems may not be ideal.

[0006] However, existing methods for predicting robot battery power have the following disadvantages:

[0007] Disadvantage 1:

[0008] Insufficient prediction accuracy: In complex actual usage scenarios, existing methods are difficult to accurately capture the dynamic characteristics of battery power changes, resulting in low prediction accuracy.

[0009] Disadvantage 2:

[0010] Poor adaptability: Existing methods are often optimized for specific scenarios or battery types, and are difficult to adapt to different types of robots and changing usage environments.

[0011] Disadvantage 3:

[0012] Limited long-term prediction capabilities: Most existing methods perform well in short-term predictions, but lack the ability to predict long-term trends in electricity consumption. Summary of the invention

[0013] In view of the above problems, the present application provides a method and storage medium for predicting the battery life of a humanoid robot, which solves the problems of insufficient prediction accuracy and poor adaptability in existing methods for predicting the battery power of a robot.

[0014] To achieve the above object, the inventor provides a method for predicting the battery life of a humanoid robot, comprising the following steps:

[0015] Perform multi-source data collection on the robot, including battery parameter collection, usage environment perception, task load recording and charging behavior analysis;

[0016] Preprocess and extract features of collected multi-source data;

[0017] The deep learning model architecture outputs prediction results based on the extracted features.

[0018] In some embodiments, the preprocessing includes data cleaning, data normalization, and time window division.

[0019] In some embodiments, outputting prediction results according to the extracted features through the deep learning model architecture specifically includes the following steps:

[0020] The long short-term memory network is used as the backbone to capture the long-term dependencies of battery life changes;

[0021] Based on the attention mechanism, different weights are assigned in the time dimension through the time attention mechanism, and the importance of each feature is dynamically adjusted in the feature dimension through the feature attention mechanism;

[0022] By using the temporal convolutional network structure in parallel, multi-scale temporal features can be extracted;

[0023] Based on the autoencoder module, the encoder compresses the input data into a low-dimensional representation, and then the decoder reconstructs the original data, extracts key features and removes noise;

[0024] The features extracted by the long short-term memory network, attention mechanism, temporal convolutional network and autoencoder module are integrated to form a comprehensive feature vector;

[0025] The fused summed feature vector is input into multiple fully connected layers, processed by nonlinear activation functions, and the prediction results are output.

[0026] In some embodiments, the processing by the nonlinear activation function and outputting the prediction result specifically comprises the following steps:

[0027] The output layer selects the corresponding nonlinear activation function according to the specific task, processes it through the selected nonlinear activation function, and outputs the prediction result.

[0028] In some embodiments, the prediction results include remaining life prediction results, capacity attenuation prediction results and fault warnings.

[0029] In some embodiments, the output layer selects a corresponding nonlinear activation function according to a specific task, and processes the output prediction result by the selected nonlinear activation function, specifically including the following steps:

[0030] The output layer predicts multiple tasks simultaneously based on a multi-task learning framework and outputs multiple prediction results.

[0031] In some embodiments, the following steps are also included:

[0032] The deep learning module is dynamically adjusted through an adaptive prediction strategy, which includes an online learning module, a transfer learning mechanism, an active learning strategy, and integrated learning.

[0033] In some embodiments, the following steps are also included:

[0034] The output prediction results are optimized and calibrated.

[0035] In some embodiments, the following steps are also included:

[0036] Based on the output prediction results, feature importance analysis, decision path visualization and case analysis system are established.

[0037] Another technical solution is provided, a storage medium storing a computer program, wherein the computer program, when executed by a processor, executes the steps in the method for predicting the battery life of a humanoid robot as described above.

[0038] Different from the existing technology, the above technical solution collects multi-source data on the robot, including battery parameter collection, usage environment perception, task load recording and charging behavior analysis, and then preprocesses and extracts features of the collected multi-source data, inputs the extracted features into the deep learning model for prediction, and outputs the prediction results. By using multi-source data combined with the deep learning model to predict the battery life of the robot, the accuracy of the battery life prediction results is improved, and it can adapt to different types of robots and changing usage environments.

[0039] The above-mentioned records related to the invention content are only an overview of the technical solution of the present application. In order to enable ordinary technicians in the field to more clearly understand the technical solution of the present application, and then implement it according to the text of the specification and the contents recorded in the drawings, and to make the above-mentioned purpose and other purposes, features and advantages of the present application easier to understand, the following is an explanation in combination with the specific implementation mode and drawings of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings are only used to illustrate the principles, implementation methods, applications, characteristics and effects of the specific embodiments of the present application and other related contents, and shall not be considered as limitations of the present application.

[0041] In the drawings of the specification:

[0042] Figure 1 A schematic diagram of a flow chart of a method for predicting the battery life of a humanoid robot according to a specific implementation method;

[0043] Figure 2 A schematic diagram of a flow chart of step S130 described in the specific implementation method;

[0044] Figure 3 Another schematic diagram of the process of predicting the battery life of a humanoid robot according to the specific implementation method;

[0045] Figure 4 Another schematic diagram of the process of predicting the battery life of a humanoid robot according to the specific implementation method;

[0046] Figure 5 A schematic diagram of the structure of the storage medium described in the specific implementation method.

[0047] The reference numerals in the above drawings are described as follows:

[0048] 510. Storage medium,

[0049] 520. Processor. DETAILED DESCRIPTION

[0050] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects that can be achieved, the following is a detailed description of the specific embodiments listed in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0051] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.

[0052] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.

[0053] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.

[0054] In the present application, terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0055] Without further limitations, in this application, the words "include", "comprises", "has" or other similar expressions used in the sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0056] Similar to the understanding in the Examination Guidelines, in this application, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of this application, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0057] In the description of the embodiments of the present application, space-related expressions used, such as "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or position relationship based on the orientation or position relationship shown in the specific embodiments or drawings, and are only for the convenience of describing the specific embodiments of the present application or facilitating the reader's understanding, and do not indicate or imply that the referred device or component must have a specific position, a specific orientation, or be constructed or operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0058] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0059] See also Figure 1 This embodiment provides a method for predicting the battery life of a humanoid robot, comprising the following steps:

[0060] Step S110: performing multi-source data collection on the robot, wherein the multi-source data collection includes battery parameter collection, usage environment perception, task load recording, and charging behavior analysis;

[0061] Step S120: preprocessing and feature extraction of the collected multi-source data;

[0062] Step S130: Output prediction results based on the extracted features through the deep learning model architecture.

[0063] By collecting multi-source data on the robot, including battery parameter collection, usage environment perception, task load recording and charging behavior analysis, and then preprocessing and feature extraction of the collected multi-source data, the extracted features are input into the deep learning model for prediction, and the prediction results are output. By using multi-source data combined with the deep learning model to predict the battery life of the robot, the accuracy of the battery life prediction results is improved, and it can adapt to different types of robots and changing usage environments.

[0064] In some embodiments, the multi-source data acquisition module includes battery parameter acquisition, usage environment perception, task load recording, and task load recording.

[0065] Battery parameter collection: Obtain the robot's historical data from the data platform, and collect the current robot's data in real time. The collected data includes but is not limited to battery voltage, current, temperature, charge and discharge times, battery pack SOC and other parameters.

[0066] Use environmental perception: Collect external factor data such as ambient temperature, humidity, vibration, etc. through environmental sensors to evaluate the impact of the environment on battery life.

[0067] Task load recording: Record the energy consumption patterns and duration of the robot when performing different tasks, and analyze the impact of different tasks on battery consumption.

[0068] Charging behavior analysis: Record the time, frequency, and charge amount of each charge to identify the impact of charging patterns on battery life.

[0069] The preprocessing includes data cleaning, data normalization and time window division.

[0070] Data preprocessing and feature engineering are implemented as follows:

[0071] Data cleaning: Use moving median filtering to remove outliers and noise, process missing data, and ensure data quality.

[0072] V c lean(t)=median(V(tk),...,V(t),...,V(t+k))

[0073] Data normalization: Min-Max normalization is performed on data of different scales to unify the data into the same range and improve the efficiency of model training.

[0074] X n orm=(XX m in) / (X m ax-X m in) Time window division: According to the prediction task requirements, the sliding window method is used to divide the time series data into fixed-size time windows (window size is w, step size is s) to capture the local characteristics of the data.

[0075] Feature extraction: Extract high-level features from raw data, such as battery state of health (SOH), capacity decay rate, etc., to enhance the model’s predictive capabilities.

[0076] See also Figure 2 In some embodiments, outputting prediction results according to the extracted features through the deep learning model architecture specifically includes the following steps:

[0077] S210: Capturing long-term dependencies of battery life variations using a long short-term memory network as the backbone;

[0078] S220: Based on the attention mechanism, different weights are assigned in the time dimension through the time attention mechanism, and the importance of each feature is dynamically adjusted in the feature dimension through the feature attention mechanism;

[0079] S230: Extracting multi-scale temporal features by using the temporal convolutional network structure in parallel;

[0080] S240: Based on the autoencoder module, the input data is compressed into a low-dimensional representation through the encoder, and then the original data is reconstructed through the decoder to extract key features and remove noise;

[0081] S250: Fuse the features extracted by the long short-term memory network, attention mechanism, temporal convolutional network and autoencoder module to form a comprehensive feature vector;

[0082] S260: Input the fused sum feature vector into multiple fully connected layers, process it through a nonlinear activation function, and output the prediction result.

[0083] The framework of the deep learning network used is as follows:

[0084] 1. LSTM backbone network: Long Short-Term Memory (LSTM) is used as the backbone to capture the long-term dependencies of battery life changes and enhance the modeling capabilities of time series data. LSTM can effectively process and predict long-term trends and cyclical changes in time series through its memory units.

[0085] The core formula is:

[0086] f t =σ(W f ·[h(t-1),x t ]+b f )

[0087] i t =σ(W i ·[h(t-1),x t ]+b i )

[0088] o t =σ(W o ·[h(t-1),x t ]+b o )

[0089]

[0090] h t =o t *tanh(c t )

[0091] 2. Attention mechanism: Introduce time attention and feature attention mechanisms to improve the model's sensitivity to key time points and important features. Specifically, the time attention mechanism is used to assign different weights in the time dimension to highlight the most critical time steps for life prediction; the feature attention mechanism dynamically adjusts the importance of each feature in the feature dimension to ensure that the model can focus on parameters that have a greater impact on the prediction results. Time attention mechanism formula:

[0092]

[0093] 3. Temporal Convolutional Network (TCN) branch: The Temporal Convolutional Network (TCN) structure is used in parallel to extract multi-scale temporal features. TCN can capture dependencies in different time ranges through causal convolution and dilated convolution, enhancing the model's ability to recognize short-term and long-term patterns.

[0094] The TCN architecture consists of multiple causal convolutional layers, with each layer's dilation rate increasing gradually to cover a longer range of temporal dependencies. Each convolutional layer is followed by a nonlinear activation function (such as ReLU) and a normalization layer (such as batch normalization) to stabilize the training process.

[0095] F(s)=(x*f)(s)=∑ i x(si)·f(i)

[0096] 4. Residual Connections: Adding residual connections to the network can alleviate the gradient vanishing problem and improve the training effect and stability of the model. Residual connections allow signals to be transmitted directly in the network, thereby promoting effective training of deeper networks. The formula is:

[0097] Output=Layer(x)+x

[0098] Among them, x is the input and Layer(x) is the output after a series of transformations.

[0099] 5. Autoencoder module: The integrated autoencoder module is used for data noise reduction and feature compression to improve the generalization ability of the model and its adaptability to new data. The autoencoder compresses the input data into a low-dimensional representation through the encoder, and then reconstructs the original data through the decoder, thereby extracting key features and removing noise.

[0100] The autoencoder structure consists of symmetrical encoder and decoder parts, and the dimension of the hidden layer in the middle is greatly reduced compared to the input to achieve effective feature compression.

[0101] The loss function uses the reconstruction error (such as mean squared error) to optimize the performance of the autoencoder.

[0102] 6. Fusion layer: The features extracted by the LSTM backbone network, attention mechanism, TCN branch and autoencoder module are integrated to form a comprehensive feature vector, which provides rich information support for subsequent prediction tasks. The fusion method can be concatenation or weighted summation, and the specific choice is adjusted according to the experimental results.

[0103] 7. Fully connected layer and output layer: The fused feature vector is input into multiple fully connected layers, processed by nonlinear activation functions (such as ReLU), and finally the prediction results are output. The processing by nonlinear activation functions and outputting the prediction results specifically include the following steps:

[0104] The output layer selects the corresponding nonlinear activation function according to the specific task, processes it through the selected nonlinear activation function, and outputs the prediction result.

[0105] The output layer uses the corresponding activation function (such as linear activation for regression tasks) according to the specific task (such as RUL prediction, capacity decay prediction, etc.).

[0106] 8. Loss function and optimization algorithm: The weighted loss function in multi-task learning is used, combined with mean square error (MSE) and cross-entropy loss (Cross-EntropyLoss), and weighted according to the importance of different tasks. The optimization algorithm uses the Adam optimizer to accelerate convergence and improve training stability.

[0107] Multi-task loss function formula:

[0108] L=αL RUL +βL Capacity +γL Fault

[0109] Among them, α, β, and γ are the weight coefficients of task loss, and their specific values ​​are adjusted according to task priority and experiments.

[0110] 9. Regularization and prevention of overfitting: During model training, methods such as Dropout and L2 regularization are used to prevent model overfitting and improve generalization ability. Dropout randomly discards some neuron connections, while L2 regularization limits model complexity by penalizing the size of weights.

[0111] 10. Model evaluation and validation: Use cross-validation and independent test sets to evaluate the model to ensure that it performs consistently on different datasets. Evaluation metrics include mean square error (MSE), mean absolute error (MAE), and prediction accuracy.

[0112] In some embodiments, the prediction results include remaining life prediction results, capacity attenuation prediction results and fault warnings.

[0113] In some embodiments, the output layer selects a corresponding nonlinear activation function according to a specific task, and processes the output prediction result by the selected nonlinear activation function, specifically including the following steps:

[0114] The output layer predicts multiple tasks simultaneously based on a multi-task learning framework and outputs multiple prediction results.

[0115] The Multi-Task Learning (MTL) framework is used to solve multiple related tasks at the same time to improve the overall prediction performance and the collaborative ability of the model. Specifically, it includes the following aspects:

[0116] 1. Remaining Useful Life (RUL): Predict the remaining useful life of the battery, that is, the time the battery can continue to work normally from the current moment. The model is implemented through the following process:

[0117] Output layer design: An independent regression output layer is used to output the specific value of RUL.

[0118] Loss function: Mean Squared Error (MSE) is used as the main loss function to optimize the accuracy of RUL prediction.

[0119] Feature enhancement: Combine time series features with extracted advanced features (such as SOH and capacity decay rate) to improve the accuracy of RUL prediction.

[0120] 2. Capacity decay prediction: predict the decay trend of battery capacity over time and evaluate the changes in battery health status. The model is implemented through the following process:

[0121] Output layer design: An independent regression output layer is used to output the battery capacity at a specific time point in the future.

[0122] Loss function: A combination of mean square error (MSE) and absolute error (MAE) is used to balance different aspects of prediction error.

[0123] Feature Utilization: Leverage historical capacity data and related features to capture dynamic patterns of capacity decay.

[0124] C(t+Δt)=C(t)+f c ap(h t , α t , z t )*Δt

[0125] 3. Fault warning: predict possible battery failures and abnormal conditions, take preventive measures in advance, and avoid robot downtime due to battery problems. The model is implemented through:

[0126] Output layer design: Classification output layer, outputs the probability or category of fault occurrence (such as normal, minor fault, serious fault).

[0127] Loss function: Cross-Entropy Loss is used to optimize the accuracy of the classification task.

[0128] Data enhancement: Combined with anomaly detection technology, the model's ability to identify rare failure events is enhanced.

[0129] 4. Task sharing layer: Design a shared parameter layer so that multiple tasks can promote each other and improve the overall prediction performance and the collaborative ability of the model.

[0130] See also Figure 3 , in some embodiments, further comprising the following steps:

[0131] Step S140: Dynamically adjust the deep learning module through an adaptive prediction strategy, wherein the adaptive prediction strategy includes an online learning module, a transfer learning mechanism, an active learning strategy, and integrated learning.

[0132] Online learning module: Adopting incremental learning algorithm, the model can adapt to new data patterns in real time and maintain high prediction accuracy. Transfer learning mechanism: Using transfer learning technology, the knowledge learned on one type of battery is transferred to a new type of battery, shortening the model training time and improving the generalization ability. Active learning strategy: Identify samples with high uncertainty, actively request manual labeling, and optimize the learning efficiency and prediction accuracy of the model. Ensemble learning: Combine the prediction results of multiple basic models, generate the final prediction through weighted average or voting mechanism, and improve the robustness and accuracy of the prediction.

[0133] See also Figure 3 , in some embodiments, further comprising the following steps:

[0134] Step S150: Optimize and calibrate the output prediction results.

[0135] Optimization and calibration of forecast results include:

[0136] Confidence interval estimation: Provide a confidence interval for each prediction result to quantify the uncertainty of the prediction and enhance the credibility of the result.

[0137] Forecast bias correction: By analyzing historical forecast errors, dynamically adjust forecast results, reduce system bias and improve forecast accuracy.

[0138] Multi-time scale fusion: Combine short-term, medium-term and long-term forecast results to generate more comprehensive and stable lifespan forecasts to meet the needs of different scenarios.

[0139] See also Figure 3 , in some embodiments, further comprising the following steps:

[0140] Step S160: Perform feature importance analysis, decision path visualization and establish a case analysis system based on the output prediction results.

[0141] The interpretability analysis module implements the following functions:

[0142] Feature Importance Analysis: Identify the factors that have the greatest impact on battery life prediction, provide a basis for decision making, and improve model transparency.

[0143] Decision path visualization: Visualization technology is used to display the decision process and key nodes of the model to help users understand the prediction results.

[0144] Case analysis system: Establish a typical case library to provide explanations and references for abnormal prediction results and support subsequent optimization and improvement.

[0145] See also Figure 4 The method for predicting the battery life of a humanoid robot provided in the above embodiment includes: a multi-source data acquisition module, data preprocessing and feature engineering, a deep learning module framework, a multi-task learning framework, an adaptive prediction strategy, a prediction result optimization and calibration, and an interpretability analysis module. The method combines deep learning technology, multi-source data fusion, and an adaptive prediction strategy, and can effectively solve the problems existing in the prior art. It has the following advantages:

[0146] Advantage 1: Improve prediction accuracy and real-time performance

[0147] The present invention adopts multi-source data fusion and advanced deep learning models, such as LSTM combined with attention mechanism and TCN branch, which can more comprehensively capture the complex patterns of battery life changes and significantly improve the accuracy of prediction. At the same time, through online learning and adaptive prediction strategies, real-time prediction is achieved to meet the high requirements of the robot system for response speed.

[0148] Advantage 2: Enhanced long-term forecasting capabilities

[0149] Through a multi-task learning framework and integrated learning methods, the present invention can not only predict the remaining service life, but also evaluate the capacity attenuation trend and warn of potential failures, providing more comprehensive and long-term prediction capabilities, helping users plan maintenance and replacement strategies in advance, extending the robot's operating time, and reducing maintenance costs.

[0150] Advantage 3: Improve the generalization ability and adaptability of the model

[0151] The introduction of transfer learning and active learning strategies enables the prediction model to quickly adapt to different types and brands of batteries, improving the generalization ability of the model. In addition, the design of residual connections and autoencoder modules enhances the adaptability and robustness of the model to new data, ensuring that it can still provide stable prediction performance under changing usage environments.

[0152] See also Figure 5 In another embodiment, a storage medium 510 stores a computer program, and when the computer program is executed by a processor 520, the steps in the method for predicting the battery life of a humanoid robot in the above embodiment are executed.

[0153] By collecting multi-source data on the robot, including battery parameter collection, usage environment perception, task load recording and charging behavior analysis, and then preprocessing and feature extraction of the collected multi-source data, the extracted features are input into the deep learning model for prediction, and the prediction results are output. By using multi-source data combined with the deep learning model to predict the battery life of the robot, the accuracy of the battery life prediction results is improved, and it can adapt to different types of robots and changing usage environments.

[0154] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of this application, this does not limit the scope of patent protection of this application. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of this application using the contents recorded in the specification and drawings of this application, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the scope of patent protection of this application.

Claims

1. A method for predicting the battery life of a humanoid robot, characterized in that: The following steps are involved: Perform multi-source data collection on the robot, including battery parameter collection, usage environment perception, task load recording and charging behavior analysis; Preprocess and extract features of collected multi-source data; The deep learning model architecture outputs prediction results based on the extracted features.

2. The method for predicting the battery life of a humanoid robot according to claim 1, characterized in that: The preprocessing includes data cleaning, data normalization and time window division.

3. The method for predicting the battery life of a humanoid robot according to claim 1, characterized in that: Outputting prediction results according to the extracted features through the deep learning model architecture specifically includes the following steps: The long short-term memory network is used as the backbone to capture the long-term dependencies of battery life changes; Based on the attention mechanism, different weights are assigned in the time dimension through the time attention mechanism, and the importance of each feature is dynamically adjusted in the feature dimension through the feature attention mechanism; By using the temporal convolutional network structure in parallel, multi-scale temporal features can be extracted; Based on the autoencoder module, the encoder compresses the input data into a low-dimensional representation, and then the decoder reconstructs the original data, extracts key features and removes noise; The features extracted by the long short-term memory network, attention mechanism, temporal convolutional network and autoencoder module are integrated to form a comprehensive feature vector; The fused summed feature vector is input into multiple fully connected layers, processed by nonlinear activation functions, and the prediction results are output.

4. The method for predicting the battery life of a humanoid robot according to claim 3, characterized in that: The processing by the nonlinear activation function and the output prediction result specifically include the following steps: The output layer selects the corresponding nonlinear activation function according to the specific task, processes it through the selected nonlinear activation function, and outputs the prediction result.

5. The method for predicting the battery life of a humanoid robot according to claim 4, characterized in that: The prediction results include remaining life prediction results, capacity attenuation prediction results and fault warning results.

6. The method for predicting the battery life of a humanoid robot according to claim 4, characterized in that: The output layer selects a corresponding nonlinear activation function according to a specific task, processes the output prediction result through the selected nonlinear activation function, and specifically includes the following steps: The output layer predicts multiple tasks simultaneously based on a multi-task learning framework and outputs multiple prediction results.

7. The method for predicting the battery life of a humanoid robot according to claim 1, characterized in that: The following steps are also included: The deep learning module is dynamically adjusted through an adaptive prediction strategy, which includes an online learning module, a transfer learning mechanism, an active learning strategy, and integrated learning.

8. The method for predicting the battery life of a humanoid robot according to claim 1, characterized in that: The following steps are also included: The output prediction results are optimized and calibrated.

9. The method for predicting the battery life of a humanoid robot according to claim 1, characterized in that: The following steps are also included: Based on the output prediction results, feature importance analysis, decision path visualization and case analysis system are established.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting the battery life of a humanoid robot as described in any one of claims 1 to 9 are executed.

Images