Multi-degree-of-freedom mechanical arm energy consumption optimization method and system
By adopting the adaptive learning rate stochastic gradient descent algorithm and momentum concept in the energy consumption optimization of robotic arm, combined with the consideration of standby power consumption and friction power consumption, the problem of inaccurate energy consumption calculation and complex optimization algorithm in traditional methods is solved, and more efficient energy consumption optimization and model generalization performance is achieved.
Patent Information
- Application Number
- CN202311785121.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-22
- Publication Date
- 2025-06-24
AI Technical Summary
The traditional robotic arm energy consumption optimization method ignores the standby power consumption and friction power consumption of the robotic arm, resulting in inaccurate energy consumption calculations, and the optimization algorithm has problems such as complex hyperparameter adjustment, local optimal solution and large computing resource consumption.
The adaptive learning rate stochastic gradient descent algorithm is adopted, combined with the concept of momentum, and the learning rate is automatically adjusted to converge to the optimal solution of model parameters faster, and the standby power consumption and friction power consumption of the robot arm are taken into account to improve the accuracy of energy consumption calculation.
By reducing computing resource consumption, it quickly converges to the energy consumption-optimized robot arm control parameters, improves the accuracy and stability of the robot arm, and improves the generalization performance of the energy consumption model.
Smart Images
Figure CN120190813A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and particularly to a method and system for optimizing the energy consumption of a multi-degree-of-freedom robotic arm. Background Art
[0002] As an important part of modern automation and robotics, the problem of optimizing the energy consumption of a multi-degree-of-freedom robotic arm has always been a research hotspot. Traditional methods for optimizing the energy consumption of robotic arms usually use the sum of squared torques as the method for measuring energy consumption and optimize based on this method. However, this method may have a high error in some cases because it ignores the standby power consumption and frictional power consumption of the robotic arm.
[0003] Existing energy consumption models of robotic arms usually only consider the work power consumption of the robotic arm, that is, the energy consumed when the robotic arm performs tasks. However, the robotic arm still consumes energy when it is not performing tasks, such as the no-load power consumption of the motor. In addition, the movement of the robotic arm also faces frictional resistance, which also leads to an increase in energy consumption.
[0004] Secondly, in terms of algorithms, traditional optimization methods such as the gradient descent algorithm may have problems such as complex hyperparameter adjustment, getting stuck in local optimal solutions, and large computational resource consumption. These problems may limit the ability of the optimization algorithm to find the global optimal solution and increase the computational cost.
[0005] To overcome these defects, the present application proposes a method and system for optimizing the energy consumption of a multi-degree-of-freedom robotic arm. This method automatically adjusts the learning rate and combines the momentum concept by introducing the adaptive learning rate stochastic gradient descent algorithm to converge to the optimal solution of the model parameters faster. At the same time, this method also considers the standby power consumption and frictional power consumption of the robotic arm, improving the accuracy of energy consumption calculation. Summary of the Invention
[0006] The purpose of the present application is to provide a method and system for optimizing the energy consumption of a multi-degree-of-freedom robotic arm, aiming to solve the above problems.
[0007] To achieve the above purpose, the present application provides the following technical solutions:
[0008] The present application provides a method for optimizing the energy consumption of a multi-degree-of-freedom robotic arm, including:
[0009] Establishing an objective function for the energy consumption of the robotic arm;
[0010] Obtaining training data of the robotic arm movement to calculate the gradient of the objective function;
[0011] Performing iterative training according to the gradient to obtain the control parameters of the robotic arm after energy consumption optimization;
[0012] Apply the robotic arm control parameters to the robotic arm control for verification and parameter adjustment.
[0013] Further, in the step of obtaining the training data of the robotic arm movement and calculating the gradient of the objective function, the following steps are specifically included:
[0014] Calculate the gradient of the objective function with respect to the parameter θ i through the backpropagation algorithm. The calculation formula is:
[0015]
[0016] where g t,i is the gradient of the parameter θ i calculated from the time step t;
[0017] For the parameter θ i , accumulate the square of its historical gradient. The calculation formula is:
[0018] G t,ii = G t-1,ii +(g t,i ) 2
[0019] where G t,ii is the cumulative sum of the squared gradients of the parameter θ i at the time step t, and G t-1,ii is the cumulative sum of the squared gradients of the previous time step t;
[0020] Calculate the adaptive learning rate of the parameter according to the cumulative sum of the squared gradients. The calculation formula is:
[0021]
[0022] where η is the initial learning rate and ε is a constant.
[0023] Further, in the step of performing iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption, the following steps are specifically included:
[0024] Establish an energy consumption model of the robotic arm;
[0025] Use a loss function to calculate the difference between the predicted value and the actual energy consumption value of the energy consumption model;
[0026] Train the energy consumption model on the training set and adjust the parameters.
[0027] Further, in the step of using a loss function to calculate the difference between the predicted value and the actual energy consumption value of the energy consumption model, the following steps are specifically included:
[0028] The loss function includes but is not limited to the mean squared error (MSE), which is used to calculate the difference between the predicted value of the energy consumption model and the actual energy consumption value y. The definition of MSE is: between the predicted value of the energy consumption model and the actual energy consumption value y. The definition of MSE is:
[0029]
[0030] where N is the number of samples, is the predicted value of the energy consumption model for the i-th sample, and y i is the corresponding actual energy consumption value.
[0031] Furthermore, in the step of training the energy consumption model on the training set and adjusting the parameters, the following steps are specifically included:
[0032] Adjust the parameters through the backpropagation algorithm and the optimization algorithm to make the predicted value of the energy consumption model equal to the actual energy consumption value;
[0033] During iterative training, use the Adagrad algorithm to adaptively adjust the learning rate.
[0034] This application provides a multi-degree-of-freedom robotic arm energy consumption optimization system, including:
[0035] Calculation module: Establish the objective function of the robotic arm energy consumption; Obtain the training data of the robotic arm movement to calculate the gradient of the objective function;
[0036] Training module: Perform iterative training according to the gradient to obtain the robotic arm control parameters after energy consumption optimization;
[0037] Evaluation module: Apply the robotic arm control parameters to the robotic arm control for verification and parameter adjustment.
[0038] This application provides a device, which includes a processor and a memory coupled to the processor. Among them, the memory stores program instructions for implementing a multi-degree-of-freedom robotic arm energy consumption optimization method; the processor is used to execute the program instructions stored in the memory to achieve a multi-degree-of-freedom robotic arm energy consumption optimization.
[0039] This application provides a storage medium, which stores program instructions that can be run by a processor. The program instructions are used to execute a multi-degree-of-freedom robotic arm energy consumption optimization method.
[0040] This application provides a multi-degree-of-freedom robotic arm energy consumption optimization method and system, which has the following beneficial effects:
[0041] (1) This application aims to reduce the consumption of computing resources by calculating the gradient of the objective function, then perform iterative training based on the gradient to optimize and update the manipulator control parameters, improve the accuracy and stability of the manipulator, and finally apply the manipulator control parameters to the manipulator control for verification and parameter adjustment, further improving the generalization performance of the energy consumption model.
[0042] (2) Adagrad is used to maintain different learning rates for the energy consumption model parameters, and the learning rate is adaptively adjusted according to the historical gradients of the parameters. For different parameters, the learning rate can be dynamically adjusted according to their past update situations, so as to better adapt to the characteristics of the model. At the same time, for sparse data, Adagrad can adjust the learning rate according to the historical gradients of the parameters, so it can better adapt to the sparsity of the data. For parameters that are updated frequently, the learning rate is relatively reduced, making the update amplitude more stable. Description of the Drawings
[0043] Figure 1 It is a schematic flow chart of a multi-degree-of-freedom manipulator energy consumption optimization method and system according to Embodiment 1 of this application;
[0044] Figure 2 It is a schematic structural diagram of a multi-degree-of-freedom manipulator energy consumption optimization system according to Embodiment 2 of this application;
[0045] Figure 3 It is a schematic structural diagram of the device according to Embodiment 3 of this application;
[0046] Figure 4 It is a schematic structural diagram of the storage medium according to Embodiment 4 of this application. Detailed Embodiments
[0047] It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.
[0048] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.
[0049] Embodiment 1
[0050] Please refer to Figure 1 , which is a schematic flow chart of a multi-degree-of-freedom manipulator energy consumption optimization method according to Embodiment 1 of this application; the steps include:
[0051] S1: Establish an objective function for the manipulator energy consumption.
[0052] In this embodiment, establishing the objective function of the robotic arm's energy consumption requires considering multiple factors, including the geometry, dynamics, control strategy, etc. of the robotic arm. Generally speaking, the objective function can be defined as minimizing the total energy consumption of the robotic arm when performing a specific task.
[0053] The objective function is defined based on information such as the multi-degrees of freedom, dynamic characteristics, and task requirements of the robotic arm. For example, physical quantities related to the torque and energy consumption of the robotic arm can be used to establish the objective function. Establishing and optimizing the objective function of the robotic arm's energy consumption can effectively reduce energy consumption and improve the performance and efficiency of the robotic arm.
[0054] S2: Obtain the training data of the robotic arm's movement to calculate the gradient of the objective function.
[0055] In this embodiment, the control parameters of the robotic arm are initialized, and the parameters include the joint angles, speeds, accelerations, etc. of the robotic arm. At the same time, the initial value of the learning rate and other algorithm parameters are set, such as the joint angles and joint speeds and joint accelerations etc.
[0056] The training data of the robotic arm's movement includes the state information of the robotic arm and the corresponding energy consumption data, such as joint angles, speeds, and positions, etc. These data are used to calculate the gradient of the objective function.
[0057] For each parameter θ i , calculate the gradient of the objective function with respect to this parameter. It is achieved through the backpropagation algorithm, and the partial derivative of the loss function with respect to each parameter is calculated according to the chain rule.
[0058] The calculation formula is:
[0059]
[0060] where g t,i is the gradient of the parameter θ i calculated from the time step t;
[0061] For the parameter θ i , accumulate the square of its historical gradient, and the calculation formula is:
[0062] G t,ii = G t-1,ii +(g t,i ) 2
[0063] where G t,ii is the cumulative sum of the squares of the gradients of the parameter θ i at the time step t, and G t-1,ii is the cumulative sum of the squares of the gradients at the previous time step t;
[0064] Calculate the adaptive learning rate of the parameter according to the cumulative sum of the square of the gradient, and the calculation formula is:
[0065]
[0066] where η is the initial learning rate and ε is a constant.
[0067] By calculating the gradient of the objective function, find the optimal solution faster, thereby improving the optimization speed. During the optimization process, if the traditional gradient descent algorithm is used, it may be necessary to calculate the gradient of each parameter. By calculating the gradient of the objective function, the consumption of computing resources can be reduced.
[0068] The characteristic of the adaptive learning rate enables the learning rate to be dynamically adjusted according to the historical gradient of each parameter. For parameters that are updated frequently, their learning rates will relatively decrease, making the algorithm more stable; for parameters that are not updated frequently, the learning rates will relatively increase, improving the sensitivity of the algorithm. This helps to better adapt to the dynamic characteristics of the robotic arm during the training process, thereby optimizing energy consumption.
[0069] S3: Perform iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption.
[0070] In this embodiment, in the step of performing iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption, it includes steps S31 to S33, and the following specifically describes the steps.
[0071] S31: Establish an energy consumption model of the robotic arm.
[0072] Establish a mapping relationship between energy consumption, optimization variables, and related parameters. Specifically, it includes:
[0073] Collect training data on the robotic arm state information and corresponding energy consumption. The robotic arm state information may include joint angles, joint velocities, end effector positions, etc. Ensure the quality and diversity of the data to better capture the motion characteristics and energy consumption changes of the robotic arm.
[0074] Select an appropriate deep learning model structure to establish the robotic arm energy consumption model. In this application, a multi-layer perceptron (MLP) is used as the robotic arm energy consumption model, which includes two fully connected layers and an activation function ReLU. In actual applications, it needs to be adjusted according to specific problems. For example, if there is time series data, a recurrent neural network (RNN) or long short-term memory network (LSTM) may be used; if the data has a spatial structure, a convolutional neural network (CNN) can be considered. The selection of the model structure should consider the characteristics of the data and the complexity of the task.
[0075] S32: Use a loss function to calculate the gap between the predicted value and the actual energy consumption value of the energy consumption model.
[0076] The loss function includes but is not limited to the mean squared error (MSE), which is used to calculate the difference between the predicted value of the energy consumption model and the actual energy consumption value y. The definition of MSE is as follows:
[0077]
[0078] where N is the number of samples, is the predicted value of the energy consumption model for the i-th sample, and y i is the corresponding actual energy consumption value.
[0079] S33: Train the energy consumption model on the training set and adjust the parameters.
[0080] Adjust the parameters through the backpropagation algorithm and the optimization algorithm to make the predicted value of the energy consumption model equal to the actual energy consumption value.
[0081] During iterative training, the Adagrad algorithm is used to adaptively adjust the learning rate. Adagrad maintains a different learning rate for the parameters of the energy consumption model, and this learning rate is adaptively adjusted according to the historical gradients of the parameters. This enables the learning rate to be dynamically adjusted according to the past update situation of different parameters, thus better adapting to the characteristics of the model.
[0082] For sparse data, some parameters may be updated more frequently than others. Adagrad can adjust the learning rate according to the historical gradients of each parameter, so it can better adapt to the sparsity of the data. For frequently updated parameters, the learning rate is relatively reduced, making the update amplitude more stable.
[0083] The learning rate adjustment of Adagrad is based on the square root of the sum of the squares of the historical gradients of the parameters. When the gradient of a parameter is large, its learning rate will be correspondingly reduced; when the gradient is small, the learning rate will relatively increase. This adaptive adjustment helps to better balance the size of the learning rate in the early and late stages of training, improving the stability of training.
[0084] The implementation of Adagrad is relatively simple and does not require manual adjustment of the learning rate, reducing the burden of a common hyperparameter adjustment in deep learning. This is very beneficial for beginners or rapid experimental prototype development.
[0085] Adagrad usually performs well on small-scale datasets because it fully considers the historical gradient information of each parameter when updating the parameters, and can better adapt to the characteristics of small-scale data.
[0086] When adjusting parameters, heuristic search, random search, gradient descent and other methods can be used to find the optimal solution. These methods can guide the direction of parameter adjustment based on the gradient of the objective function or historical search information, thereby improving the search efficiency. After adjusting the parameters, the energy consumption of the robotic arm can be recalculated and evaluated. If it is found that the new parameter combination can reduce the energy consumption, the parameters can be further adjusted to optimize the energy consumption. By adjusting the parameters, the optimal robotic arm control parameters can be obtained, improving the accuracy and stability of the robotic arm, thereby reducing the energy consumption and improving the energy efficiency.
[0087] S4: Apply the robotic arm control parameters to the robotic arm control for verification and parameter adjustment.
[0088] In this embodiment, the performance of the model is evaluated using the validation set to observe whether the model is overfitting or underfitting. The model structure, hyperparameters, etc. are adjusted according to the verification results to further improve the generalization performance of the model.
[0089] It can be understood that this application mainly combines the randomness of SGD with the learning rate adaptability of the adaptive learning rate algorithm, so as to better balance fast convergence and stability and train the neural network model more efficiently. It should be noted that the formulas and parameter settings of specific adaptive learning rate algorithms may be adjusted according to the actual application situation. Therefore, in actual applications, it may be necessary to optimize and select parameters according to the specific tasks and datasets.
[0090] Adjustment of the learning rate:
[0091] Original setting: If the initial learning rate is set too high, it may cause the model to fail to converge or even diverge. For example, the learning rate is set to 0.1.
[0092] Optimization: Adjust the learning rate to a smaller value, such as 0.001, to improve the stability of the model and allow smaller step sizes for parameter updates.
[0093] Adjustment of the number of neurons in the hidden layer:
[0094] Original setting: Set a smaller number of neurons in the hidden layer, such as 10.
[0095] Optimization: Increase the number of neurons in the hidden layer, such as setting it to 128, to increase the representation ability of the model and improve the complexity of learning.
[0096] Adjustment of the number of iterations:
[0097] Original setting: Set a smaller number of iterations, such as 50 times.
[0098] Optimization: Increase the number of iterations, for example, set it to 200 times, to allow the model to observe the data and update the parameters more times, which helps to improve the performance of the model.
[0099] Optimization algorithm selection:
[0100] Original setting: Use the Adagrad optimization algorithm.
[0101] Optimization: Try other optimization algorithms, such as Adam or RMSprop, and observe their effects on model training, because different optimization algorithms may have different effects on different types of problems.
[0102] Introduction of regularization:
[0103] Original setting: The model does not use regularization.
[0104] Optimization: Introduce L1 or L2 regularization to prevent overfitting, especially when the training data is less or the model is more complex.
[0105] In summary, in Embodiment 1 of the present application, the gradient of the objective function is calculated to reduce the consumption of computing resources; then iterative training is performed according to the gradient to optimize and update the manipulator control parameters, improving the accuracy and stability of the manipulator; finally, the manipulator control parameters are applied to manipulator control for verification and parameter adjustment; further improving the generalization performance of the energy consumption model.
[0106] Embodiment 2
[0107] Please refer to Figure 2 , which is a schematic structural diagram of a multi-degree-of-freedom manipulator energy consumption optimization system according to Embodiment 2 of the present application; the specific content includes:
[0108] Calculation module: Establish the objective function of the manipulator energy consumption; obtain the training data of the manipulator movement to calculate the gradient of the objective function;
[0109] Training module: Perform iterative training according to the gradient to obtain the manipulator control parameters after energy consumption optimization;
[0110] Evaluation module: Apply the manipulator control parameters to manipulator control for verification and parameter adjustment.
[0111] In this embodiment, for example, there is a task of a manipulator grasping an object. The manipulator energy consumption data under different acceleration parameter settings has been collected, and the relevant parameters (object mass, action time, etc.) have been recorded. Now, the adaptive learning rate stochastic gradient descent algorithm is used to optimize the energy consumption of the manipulator.
[0112] First, establish an energy consumption model to establish the relationship between energy consumption and acceleration parameters, object mass, action time, etc. Next, preprocess the collected data, and initialize the acceleration parameters and the hyperparameters of the optimization algorithm.
[0113] Start iterative training. For each acceleration parameter setting, execute the grasping task of the robotic arm and record the energy consumption. Then calculate the loss function and the gradient. Next, use the adaptive learning rate algorithm to update the acceleration parameters to minimize the loss function as much as possible.
[0114] After multiple iterative trainings, obtain the optimized acceleration parameters. Apply these parameters to the control of the robotic arm, so as to reduce the energy consumption and improve the energy efficiency of the robotic arm in the grasping task.
[0115] In summary, in Embodiment 2 of the present application, the gradient of the objective function is obtained through the calculation module, and iterative training is performed by the training module to obtain the optimal robotic arm control parameters, making the control of the robotic arm more accurate and stable.
[0116] Embodiment 3
[0117] Please refer to Figure 3 , which is a schematic diagram of the device structure of Embodiment 3 of the present application. The device 50 includes a processor 51 and a memory 52 coupled to the processor 51.
[0118] The memory 52 stores program instructions for implementing the above-mentioned method for optimizing the energy consumption of a multi-degree-of-freedom robotic arm.
[0119] The processor 51 is used to execute the program instructions stored in the memory 52 to implement the optimization of the energy consumption of a multi-degree-of-freedom robotic arm.
[0120] Among them, the processor 51 can also be called a CPU (Central Processing Unit, central processing unit).
[0121] The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.
[0122] Embodiment 4
[0123] Please refer to Figure 4, which is a schematic structural diagram of the storage medium according to Embodiment 4 of the present application. The storage medium of the embodiment of the present application stores a program file 61 that can implement all the above methods. Among them, the program file 61 can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods according to various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, or devices such as computers, servers, mobile phones, and tablets.
[0124] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, device, article, or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such a process, device, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, device, article, or method including that element.
[0125] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.
[0126] Although the embodiments of the present application have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present application. The scope of the present application is defined by the appended claims and their equivalents.
[0127] Certainly, the present invention can also have other various embodiments. Based on this embodiment, other embodiments obtained by those of ordinary skill in the art without any creative labor belong to the scope protected by the present invention.
Claims
1. An energy consumption optimization method for a multi-degree-of-freedom robotic arm, characterized in that, Including: Establish an objective function for the energy consumption of the robotic arm; Obtain training data of the robotic arm's movement to calculate the gradient of the objective function; Perform iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption; Apply the robotic arm control parameters to the robotic arm control for verification and parameter adjustment.
2. The energy consumption optimization method of a multi-degree-of-freedom robotic arm according to claim 1, characterized in that In the step of obtaining training data of the robotic arm's movement to calculate the gradient of the objective function, it specifically includes the following steps: Calculate the gradient of the objective function with respect to the parameter θ by the backpropagation algorithm i The calculation formula is as follows: where g t,i is the gradient of parameter θ i calculated from time step t; For parameter θ i , accumulate the squares of its historical gradients, and the calculation formula is as follows: G t,ii = G t-1,ii + (g t,i ) 2 where G t,ii is the cumulative sum of the squared gradients of the parameter θ i at time step t, and G t-1,ii is the cumulative sum of the squared gradients at the previous time step t; Calculate the adaptive learning rate of the parameter according to the cumulative sum of the squared gradients, and the calculation formula is: where η is the initial learning rate and ε is a constant.
3. A multi-degree-of-freedom robotic arm energy consumption optimization method according to claim 1, characterized in that, In the step of performing iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption, it specifically includes the following steps: Establish an energy consumption model of the robotic arm; Use a loss function to calculate the difference between the predicted value and the actual energy consumption value of the energy consumption model; Train the energy consumption model on the training set and adjust the parameters.
4. A method for optimizing the energy consumption of a multi-degree-of-freedom robotic arm according to claim 3, characterized in that, In the step of using a loss function to calculate the difference between the predicted value and the actual energy consumption value of the energy consumption model, it specifically includes the following steps: The loss function includes but is not limited to the mean squared error (MSE), which is used to calculate the difference between the predicted value of the energy consumption model and the actual energy consumption value y. The definition of MSE is as follows: where N is the number of samples, is the predicted value of the energy consumption model for the i-th sample, y i is the corresponding actual energy consumption value.
5. A method for optimizing the energy consumption of a multi-degree-of-freedom robotic arm according to claim 3, characterized in that, In the step of training the energy consumption model on the training set and adjusting the parameters, it specifically includes the following steps: Adjust the parameters through the backpropagation algorithm and the optimization algorithm to make the predicted value of the energy consumption model equal to the actual energy consumption value; When performing iterative training, use the Adagrad algorithm to adaptively adjust the learning rate.
6. A system for an energy consumption optimization method of a multi-degree-of-freedom robotic arm according to claim 1, characterized in that, Including: Calculation module: Establish an objective function for the energy consumption of the robotic arm; Obtain training data of the robotic arm's movement to calculate the gradient of the objective function; Training module: Perform iterative training according to the gradient to obtain the robotic arm control parameters with optimized energy consumption; Evaluation module: Apply the robotic arm control parameters to the robotic arm control for verification and parameter adjustment.
7. A device, characterized in that, The device includes a processor and a memory coupled to the processor. Among them, the memory stores program instructions for implementing a multi-degree-of-freedom robotic arm energy consumption optimization method according to any one of claims 1-5; the processor is used to execute the program instructions stored in the memory to implement a multi-degree-of-freedom robotic arm energy consumption optimization.
8. A storage medium, characterized in that, Store program instructions that can be run by the processor, and the program instructions are used to execute a multi-degree-of-freedom robotic arm energy consumption optimization method according to any one of claims 1-5.