Predictive robot path planning method and device, electronic equipment and medium

Abstract

This invention relates to a prediction-based robotic arm path planning method, apparatus, electronic device, and medium. The method involves real-time acquisition of hand position data and storage in a raw dataset, which also includes the current position data of the robotic arm, target position data, and actual obstacle position data. Based on the raw dataset, an Informer model is trained to obtain model weights and output predicted hand position data. The predicted hand position is used as a virtual obstacle, and a predicted repulsive potential field is obtained based on the predicted hand position data. Based on the predicted repulsive potential field, the repulsive potential field of the actual obstacle, and the gravitational potential field of the target position, the current resultant force is calculated to optimize the robotic arm's trajectory, control the robotic arm's movement, and perform sorting. The target position is the endpoint of the robotic arm's movement. Compared with existing technologies, this invention has advantages such as higher prediction accuracy, faster calculation speed, and higher human-machine collaboration efficiency.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a prediction-based method, apparatus, electronic device, and medium for planning the path of a robotic arm. Background Technology

[0002] Path planning is one of the core problems in robotics. It enables robots to plan a feasible path within a workspace. After years of development, industrial robot motion planning has become relatively mature and is widely used in sorting, welding, grinding, and other fields. However, collaborative robots typically operate in complex environments, facing dynamic obstacles (such as workbenches) and dynamic obstacles (such as other robots), requiring suitable path planning methods to ensure the robot successfully completes its tasks. Obstacle avoidance is the process where, when an obstacle is detected in a collaborative scenario, the robotic arm can avoid it in the fastest time and at the lowest cost, and continue moving towards the target point after successfully avoiding the obstacle. The artificial potential field method has advantages such as simple and clear algorithm structure and fast response speed, therefore, it has been widely used in the field of obstacle avoidance path planning for robotic arms and has gradually become the mainstream obstacle avoidance algorithm. However, when the artificial potential field method is applied to obstacle avoidance for robotic arms, two problems arise: the target unreachability problem and the local minima problem.

[0003] In existing technologies, researchers primarily address the local minima problem by setting virtual guide points (including virtual target points and obstacles) to guide the robotic arm away from the local minimum point. Some have proposed randomly generating a sub-target point at the local minimum point, which can generate an attractive force to help the robot escape the local minimum location. Besides setting sub-target points, virtual obstacle points can also be set, dynamically added in three ways depending on the angle between the line connecting the robot and the target point and the sides of the obstacle. While existing technologies can solve the local minimum problem, they suffer from high randomness, making the planned motion trajectory of the robotic arm's end effector uncontrollable. This leads to a lack of user safety and low collaboration efficiency in human-computer interaction and collaboration. Chinese patent CN115502961A discloses a human-machine online obstacle avoidance method based on prediction of human arm motion information. This method obtains the coordinates of key points on a human arm to acquire its motion information. This information is then input into an LSTM-BP model to predict the arm's movement. The LSTM-BP model includes a Long Short-Term Memory neural network and a Backpropagation neural network. The human arm is treated as an obstacle; a repulsive field is obtained based on the predicted arm motion. The robot's endpoint is used as the target point; a gravitational field is obtained based on the target point, thereby determining the robot's end-effector pose and controlling its movement. This patent improves the safety of human-robot interaction, but suffers from low prediction accuracy and slow computation speed. Therefore, improving human-robot collaboration efficiency while ensuring safety in robotic arm path planning is a problem that needs to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art, such as low prediction accuracy and slow calculation speed, and to provide a prediction-based robotic arm path planning method, device, electronic device and medium.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] According to a first aspect of the present invention, a prediction-based robotic arm path planning method is provided, the method comprising the following steps:

[0007] S1, collect hand position data in real time and store it in the raw dataset, which also includes the current position data of the robotic arm, the target position data and the actual obstacle position data;

[0008] S2, based on the original dataset, train the Informer model, obtain the model weights, and output the predicted hand position data;

[0009] S3, using the predicted hand position as a virtual obstacle, obtains the predicted repulsive potential field based on the predicted hand position data;

[0010] S4 calculates the current resultant force based on the predicted repulsive potential field, the actual repulsive potential field of the obstacle, and the gravitational potential field of the target position, optimizes the robot arm's motion trajectory, controls the robot arm's movement, and performs sorting.

[0011] The target position is the endpoint of the robotic arm's movement.

[0012] Preferably, the process of training the Informer model in S2 includes:

[0013] S21, Organize and preprocess the collected hand position data, and store it in the original dataset;

[0014] S22, divide the original dataset into training set, validation set and test set, and obtain the time series input features contained in the data;

[0015] S23. Based on the time series input features of the training set data, select a loss function, train the Informer model, and calculate the prediction results through forward propagation;

[0016] S24, calculate and monitor the loss value during the training process, optimize the loss function through backpropagation, and update the weight parameters of the Informer model;

[0017] S25, outputs the predicted hand position data;

[0018] The Informer model is constructed by an encoder and a decoder. The validation set data is used to select the training time and avoid overfitting of the predicted hand position data. The test set data is used to verify the predictive ability of the Informer model on real data.

[0019] Preferably, the expression for the predicted repulsive force is:

[0020]

[0021] Where, k p It is the repulsion coefficient, p1(X) represents the distance between the robotic arm and the predicted obstacle, p p It is a threshold used to determine the range of predicted obstacle effects on the robotic arm.

[0022] Preferably, the repulsive force expression of the actual obstacle is:

[0023]

[0024] Where, k r It is the repulsion coefficient, p2(X) represents the distance between the robotic arm and the actual obstacle, p o It is a threshold used to determine the range of influence of actual obstacles on the robotic arm;

[0025] The gravitational expression for the target location is:

[0026]

[0027] Where, k a X is the gravitational coefficient, where X represents the current position of the robotic arm. g Indicates the location of the target, U attr (X) represents the gravitational potential field at the target location.

[0028] Preferably, the process of optimizing the robotic arm's motion trajectory in S4 includes using Cartesian space path planning to plan an intuitive and controllable motion path for the robotic arm's end effector.

[0029] Preferably, the original dataset also includes head pose data for training the Informer model and predicting hand positions.

[0030] Preferably, the process of real-time acquisition of hand position data includes acquiring hand position data using a depth camera with an acquisition frequency of 30 frames per second.

[0031] According to a second aspect of the present invention, a prediction-based robotic arm path planning device is provided. This device implements the method described above, comprising a hand position data acquisition module, an Informer model training module, a predicted repulsive force acquisition module, and a motion trajectory optimization module. The hand position data acquisition module is used to acquire hand position data in real time and store it in an original dataset. The Informer model training module is used to train an Informer model using the original dataset, obtain model weights, and output predicted hand position data. The predicted repulsive force acquisition module is used to treat the predicted hand position as a virtual obstacle, obtain a predicted repulsive potential field based on the predicted hand position data, and then obtain the predicted repulsive force. The motion trajectory optimization module is used to calculate the current resultant force based on the predicted repulsive potential field, the repulsive potential field of the actual obstacle, and the gravitational potential field of the target position, optimize the robotic arm motion trajectory, control the robotic arm movement, and perform sorting.

[0032] According to a third aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0033] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. This invention uses the hand position information predicted by the Informer model to increase the predicted repulsive force, which can predict obstacles with higher prediction accuracy and shorter training time. It can improve the flexibility and obstacle avoidance ability of the robotic arm, with better real-time performance and higher accuracy. It can improve the safety of path planning and the efficiency of model training, thereby improving the efficiency of human-machine collaboration.

[0036] 2. This invention introduces predicted repulsive force into the traditional artificial potential field method and combines it with an improved repulsive force function, which solves the local minima problem and the problem of unreachable targets near obstacles in the traditional artificial potential field method;

[0037] 3. This invention utilizes the long-sequence data processing capability of the Informer model, which helps the robotic arm to make longer-term path planning, thereby improving the smoothness of the robotic arm's motion and its dynamic obstacle avoidance capability. Attached Figure Description

[0038] Figure 1 This is a schematic flowchart of the method of the present invention;

[0039] Figure 2 This is a schematic diagram of the trajectory of hand movement in Embodiment 1 of the present invention;

[0040] Figure 3 This is a schematic diagram of the Informer model in Embodiment 1 of the present invention;

[0041] Figure 4 This is a schematic diagram comparing the actual trajectory and the predicted trajectory of hand movement in Embodiment 1 of the present invention;

[0042] Figure 5 This is a schematic diagram of the prediction-based robotic arm path planning algorithm in Embodiment 1 of the present invention;

[0043] Figure 6 This is a schematic diagram of the prediction-based robotic arm path planning trajectory in Embodiment 1 of the present invention;

[0044] Figure 7 This is a schematic diagram of path planning for the traditional artificial potential field method and the prediction-based improved artificial potential field method in Embodiment 1 of the present invention. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0046] Example 1

[0047] like Figure 1 As shown in the figure, this embodiment provides a prediction-based robotic arm path planning method, and the specific execution steps are as follows:

[0048] Step S1: Create a dataset, use the ZED2i depth camera object detection parameters, enable the object detection module, and acquire object data. For example... Figure 2 As shown, real-time hand position data of the operator during sorting is collected, specifically the coordinates of the operator's right hand in the camera coordinate system. Time information ("data") is added, and the data is saved as a CSV file to create the training set, which serves as the original dataset for the neural network. The original dataset also includes data on the current position of the robotic arm, the target position, and the actual obstacle positions. The depth camera captures data at a frequency of 30 frames per second, with "data" added to the first column.

[0049] Step S2 involves normalizing the original data, dividing the dataset into training, validation, and test sets, and extracting the time-series input features. A suitable loss function is selected for the regression task. Using the time-series input features from the training set, the Informer model is trained with a custom time feature encoding frequency. Model parameters are initialized, and parameter training and iteration are performed within the network to ultimately obtain the model weight file. During training, the Informer model performs forward propagation to calculate prediction results and outputs predicted hand position data. Simultaneously, the loss value is calculated and monitored during training. The loss function is optimized through backpropagation, and the Informer model's weight parameters are updated to achieve better model performance. The output results are continuously optimized in subsequent training. The validation set data is used to evaluate the model's performance, and an appropriate training time is selected to avoid overfitting of the hand position prediction data. The test set data is used to evaluate the model's performance and verify the Informer model's predictive ability on real-world data.

[0050] like Figure 3 As shown, the Informer model is an efficient long-time series prediction model based on an improvement of the Transformer model. Its structure includes an encoder and a decoder. The encoder receives extremely long input data (Inputs: X). feed_en This paper replaces the traditional self-attention layer with the proposed probabilistic sparse self-attention layer. The blue-filled area in the figure represents self-attention distillation, an operation used for feature compression. The encoder part improves the robustness of the algorithm by stacking the probabilistic sparse self-attention layer and the self-attention distillation layer. The decoder receives a series of long sequence inputs (Inputs: X...). feed_de ={X tokenThe model is processed by filling the predicted target position with 0, and then passing it through a masked multi-head probSpare self-attention layer to finally generate the predicted output, which is the predicted hand position information in this embodiment. In actual model training, the key parameters are set as follows: --model informer, --features M, --seq_len 48, --label_len 24, --pred_len 12, and --inverse True. Here, --modelinformer indicates that the experimental model is an Informer model, --features M indicates that the prediction task is multivariate prediction of multivariate, --seq_len 48 indicates that the input sequence length of the Informer encoder is 48, --label_len 24 indicates that the initial label length of the Informer decoder is 24, --pred_len 12 indicates that the predicted sequence length is 12, and --inverseTrue indicates that the output result is inversely normalized.

[0051] Step S3: Using the predicted hand position as a virtual obstacle, obtain the predicted hand position information to get the predicted repulsive force. During model training, the prediction function is called to load the trained Informer weight file, predict the hand's movement over a future period, and inversely normalize the output hand position data to obtain the predicted value. The predicted hand position coordinates are transformed to the robotic arm base coordinate system through the robotic arm camera coordinate transformation, serving as obstacle prediction information in the robotic arm path planning, such as... Figure 4 As shown. Compared to the traditional LSTM model, the Informer model has higher prediction accuracy, improving accuracy by 17%. Due to the parallel computing capabilities of the Informer, training time is shorter, computation time is shorter, and computation efficiency is improved by 10%. Higher obstacle prediction accuracy and shorter computation time allow the robotic arm to quickly plan obstacle avoidance routes in advance, resulting in smoother robotic arm movements and fewer iterations. Based on the acquired obstacle prediction information, a predicted repulsion force is added, and the expression for the predicted repulsion force is:

[0052]

[0053] Where, k p It is the repulsion coefficient, p1(X) represents the distance between the robotic arm and the predicted obstacle, p p It is a threshold used to determine the range of predicted obstacle effects on the robotic arm.

[0054] Step S4: Based on the predicted repulsive potential field, the actual repulsive potential field of the obstacle, and the gravitational potential field of the target position, calculate the current resultant force. Through Cartesian space path planning, plan an intuitive and controllable motion path for the robotic arm's end effector, optimize the robotic arm's motion trajectory, and thus control the actual movement of the robotic arm for sorting. The specific process is as follows: Figure 5 As shown in the figure. The target position is the endpoint of the robotic arm's movement.

[0055] Since the artificial potential field method is based on a potential field model, the representations of attraction and repulsion need to be built upon the potential field. The gravitational potential field function at the target location is:

[0056]

[0057] Since the gravitational function is the negative gradient of the gravitational potential function, the gravitational function at the target location is:

[0058]

[0059] Where, k a X is the gravitational coefficient, where X represents the current position of the robotic arm. g Indicates the location of the target, U attr (X) represents the gravitational potential field at the target location.

[0060] Similar to the gravitational potential field function, the improved repulsive potential field function for a real obstacle is:

[0061]

[0062] Therefore, the repulsive force function of the actual obstacle is:

[0063]

[0064] Where, k r It is the repulsion coefficient, p2(X) represents the distance between the robotic arm and the actual obstacle, p o It is a threshold used to determine the range of influence of actual obstacles on the robotic arm, also known as the repulsion range.

[0065] Compared to the repulsive potential field function of the traditional artificial potential field method, the repulsive force of the actual obstacle on the robotic arm in the improved repulsive potential field function is weakened by adding a distance term between the target and the robotic arm. Simultaneously, a new force is added on the robotic arm from the target, and this force increases as the distance between the robotic arm and the target decreases. However, in some scenarios (where the actual obstacle is between the robotic arm and the target point), if the direction of the robotic arm's movement is not changed, simply improving the repulsive force function will not allow the robotic arm to reach the target point. Therefore, combining the predicted repulsive force obtained in step 3, the improved artificial potential field method based on prediction is used for robotic arm path planning. In this case, the resultant force formula of the improved artificial potential field based on prediction is:

[0066] F res (X)=F attr (X)+F rep (X)+F pre (X)

[0067] like Figure 6 As shown, the improved artificial potential field method based on prediction establishes a virtual repulsive potential field at the predicted obstacle location, allowing the robotic arm to consider obstacle movement in advance during path planning and move away from obstacles ahead of time, thereby reducing the possibility of collisions between the robotic arm and obstacles. Furthermore, introducing the Informer model for obstacle prediction results in higher accuracy, making the parameters closer to reality and resulting in smoother robotic arm movements. Figure 7 As shown in the figure, through theoretical experiments in Matlab, it is obtained that the prediction-based improved artificial potential field method can take into account the movement of obstacles and move away from obstacles in advance. The number of iterations is reduced by 14.91% compared with the traditional artificial potential field method, thus enabling the robotic arm to efficiently avoid obstacles and complete the obstacle avoidance motion planning. Therefore, the prediction-based improved artificial potential field method can solve two defects of the traditional artificial potential field method: the local minima problem and the problem of unreachable targets near obstacles.

[0068] Finally, the trajectory data obtained by the prediction-based improved artificial potential field method will be broadcast through the node of the ROS robot operating system, and the data will be subscribed to by the MoveIt! robot arm trajectory planning node and the robot arm's underlying driver to realize the sorting and obstacle avoidance work of the collaborative robot arm.

[0069] Example 2

[0070] This embodiment is basically the same as the method provided in Embodiment 1, the difference being that the original dataset also includes the user's head pose data, which is used to train the Informer model and predict hand positions. Since in a sorting scenario, the human eye looks at the item first, and then the hand grasps it; head movement precedes hand movement. Therefore, combining head pose and hand position data to predict hand positions yields more accurate prediction results. Furthermore, the trajectory data obtained through iterative iteration using the improved artificial potential field method based on prediction is broadcast through the ROS robot operating system nodes, and the data is subscribed to by the MoveIt! robotic arm trajectory planning node and the underlying robotic arm driver, enabling collaborative robotic arm sorting and obstacle avoidance.

[0071] Example 3

[0072] This embodiment provides a prediction-based robotic arm path planning device, which includes a hand position data acquisition module, an Informer model training module, a predicted repulsive force acquisition module, and a motion trajectory optimization module. The hand position data acquisition module collects hand position data in real time and stores it in the original dataset; the Informer model training module trains the Informer model using the original dataset, obtains model weights, and outputs predicted hand position data; the predicted repulsive force acquisition module uses the predicted hand position as a virtual obstacle, obtains the predicted repulsive potential field based on the predicted hand position data, and then obtains the predicted repulsive force; the motion trajectory optimization module calculates the current resultant force based on the predicted repulsive potential field, the repulsive potential field of the actual obstacle, and the gravitational potential field of the target position, optimizes the robotic arm's motion trajectory, controls the robotic arm's movement, and performs sorting.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the aforementioned method embodiments 1 and 2, and will not be repeated here.

[0074] Example 4

[0075] The electronic device provided in this embodiment includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from storage units into random access memory (RAM). The RAM can also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0076] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0077] The processing unit executes the various methods and processes described above, such as steps S1 to S4 in Embodiment 1. For example, in some embodiments, steps S1 to S4 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of steps S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 by any other suitable means (e.g., by means of firmware).

[0078] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0079] Example 5

[0080] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described herein. The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0081] In the context of this invention, a computer-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0082] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A prediction-based robotic arm path planning method, characterized in that, Includes the following steps: S1, collect the hand position data of the person working in real time and store it into the raw dataset. The raw dataset also includes the current position data of the robotic arm, the target position data and the actual obstacle position data. S2, based on the original dataset, train the Informer model, obtain the model weights, and output the predicted hand position data; S3, using the predicted hand position as a virtual obstacle, obtains the predicted repulsive potential field based on the predicted hand position data; S4 calculates the current resultant force based on the predicted repulsive potential field, the actual repulsive potential field of the obstacle, and the gravitational potential field of the target position, optimizes the robot arm's motion trajectory, controls the robot arm's movement, and performs sorting. The target position is the endpoint of the robotic arm's movement; The expression for the predicted repulsive force is: in, It is the repulsion coefficient. This indicates the distance between the robotic arm and the predicted obstacle. It is a threshold used to determine the predicted range of the obstacle's impact on the robotic arm; The expression for the repulsive force of the actual obstacle is: in, It is the repulsion coefficient. This indicates the distance between the robotic arm and the actual obstacle. It is a threshold used to determine the range of influence of actual obstacles on the robotic arm; The gravitational expression for the target location is: in, It is the gravitational coefficient. Indicates the current position of the robotic arm. Indicates the location of the target. This represents the gravitational potential field at the target location.

2. The prediction-based robotic arm path planning method according to claim 1, characterized in that, The process of training the Informer model in S2 includes: S21, Organize and preprocess the collected hand position data, and store it in the original dataset; S22, divide the original dataset into training set, validation set and test set, and obtain the time series input features contained in the data; S23. Based on the time series input features of the training set data, select a loss function, train the Informer model, and calculate the prediction results through forward propagation; S24, calculate and monitor the loss value during the training process, optimize the loss function through backpropagation, and update the weight parameters of the Informer model; S25, outputs the predicted hand position data; The Informer model is constructed by an encoder and a decoder. The validation set data is used to select the training time and avoid overfitting of the predicted hand position data. The test set data is used to verify the predictive ability of the Informer model on real data.

3. The prediction-based robotic arm path planning method according to claim 1, characterized in that, The process of optimizing the robotic arm's motion trajectory in S4 includes using Cartesian space path planning to plan the motion path for the robotic arm's end effector.

4. The prediction-based robotic arm path planning method according to claim 1, characterized in that, The original dataset also includes head pose data, which is used to train the Informer model and predict hand positions.

5. The prediction-based robotic arm path planning method according to claim 1, characterized in that, The process of acquiring real-time hand position data includes using a depth camera with an acquisition frequency of 30 frames per second to acquire hand position data.

6. A prediction-based robotic arm path planning device, characterized in that, The device is used to implement the method according to any one of claims 1-5, and includes a hand position data acquisition module, an Informer model training module, a predicted repulsion force acquisition module, and a motion trajectory optimization module. The hand position data acquisition module is used to acquire hand position data in real time and store it into the raw dataset; The Informer model training module is used to train the Informer model using the original dataset, obtain the model weights, and output the predicted hand position data. The predicted repulsive force acquisition module is used to take the predicted hand position as a virtual obstacle, obtain the predicted repulsive force potential field based on the predicted hand position data, and then obtain the predicted repulsive force. The motion trajectory optimization module is used to calculate the current resultant force based on the predicted repulsive potential field, the actual repulsive potential field of the obstacle, and the gravitational potential field of the target position, optimize the motion trajectory of the robotic arm, control the movement of the robotic arm, and perform sorting.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-5.

Images