Non-contact early warning methods, robotic arm collision avoidance methods and systems

Abstract

This application provides a non-contact early warning method, a robotic arm collision avoidance method, and a system, relating to the field of robot safety control technology. The non-contact early warning method includes: during the target robot's task execution, determining at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by various capacitance sensors, each capacitance sensor being installed on the target robot; for each capacitance sensor, triggering an obstacle warning when the capacitance characteristics corresponding to the sensor meet the impending contact condition, where the impending contact condition indicates that the target robot is about to contact an obstacle. This application can achieve proactive early warning before collisions, improving safety; it is unaffected by lighting, occlusion, or object material, suitable for complex scenarios, and has strong environmental adaptability; it has low cost and low computational overhead, requiring no complex vision system or deep learning model, making deployment simple and suitable for large-scale applications; it has strong real-time performance, with fast capacitance sensing response speed, meeting the obstacle avoidance requirements of highly dynamic environments.

Description

Technical Field

[0001] This application relates to the field of robot safety control technology, and in particular to a non-contact early warning method, a robotic arm collision avoidance method and system. Background Technology

[0002] With the increasing demand for dexterous operations by humanoid robots, the safe operation of robotic arms in complex environments has become a key technological challenge. Traditional collision avoidance methods mainly rely on force sensors, vision systems, or lidar, but these solutions suffer from problems such as response delays, high costs, or poor environmental adaptability. Furthermore, force sensors require physical contact to trigger, making non-contact warnings impossible; vision systems are significantly affected by lighting and occlusion; and lidar struggles to detect transparent or highly reflective objects. Therefore, an effective solution is urgently needed to address these issues. Summary of the Invention

[0003] To address the aforementioned issues, this application provides a non-contact early warning method, a robotic arm collision avoidance method, and a system.

[0004] This application provides a non-contact early warning method, including: During the task execution process of the target robot, at least two dimensions of capacitance characteristics are determined based on the target capacitance values ​​collected by each capacitance sensor, and each capacitance sensor is installed on the target robot. For each of the capacitive sensors, if the capacitive characteristics corresponding to the capacitive sensor meet the impending contact condition, an obstacle warning is triggered, wherein the impending contact condition indicates that the target robot is about to come into contact with an obstacle.

[0005] According to a non-contact early warning method provided in this application, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the changing trend of adjacent capacitance sensors; The method for determining capacitance characteristics in at least two dimensions based on the target capacitance values ​​collected by each capacitance sensor includes: For each of the capacitive sensors, local extrema calculation is performed on the target capacitance value collected by the capacitive sensor to obtain the peak value corresponding to the capacitive sensor; By selecting a fixed time window, gradient feature statistics are performed on the target capacitance value collected by the capacitance sensor to obtain the gradient feature corresponding to the capacitance sensor. Based on the target capacitance value collected by the capacitance sensor and the target capacitance value collected by the adjacent capacitance sensors, a sign consistency judgment is performed to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor.

[0006] According to a non-contact early warning method provided in this application, the step of performing a sign consistency judgment based on the target capacitance value collected by the capacitance sensor and the target capacitance values ​​collected by adjacent capacitance sensors to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor includes: The target capacitance value collected by the capacitance sensor is calculated by differential value to obtain a first differential value, and the target capacitance value collected by the adjacent capacitance sensor is calculated by differential value to obtain a second differential value; Perform a sign consistency check on the first difference value and the second difference value; If the signs are the same, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensors are determined to be in the same direction. If the symbols are inconsistent, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensor are determined to be opposite.

[0007] According to a non-contact early warning method provided in this application, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the changing trend of adjacent capacitance sensors; Before triggering an obstacle warning when the capacitance characteristics corresponding to each capacitance sensor meet the impending contact condition, the method further includes: The consistency ratio of the changing trends of adjacent capacitive sensors corresponding to the capacitive sensor is determined to be in the same direction; Determine whether the peak value, gradient feature, and consistency ratio corresponding to the capacitive sensor meet the upcoming contact conditions.

[0008] According to a non-contact early warning method provided in this application, the peak value is a minimum value; The conditions for impending contact are that the minimum value corresponding to the capacitive sensor is less than a specified peak threshold, the gradient feature corresponding to the capacitive sensor is greater than a specified gradient feature threshold, and the consistency ratio corresponding to the capacitive sensor is greater than a specified ratio threshold. The specified peak threshold is either a general peak threshold or a dedicated peak threshold corresponding to the capacitive sensor; The specified gradient feature threshold is either a general gradient feature threshold or a dedicated gradient feature threshold corresponding to the capacitive sensor. The specified ratio threshold is either a general ratio threshold or a specific ratio threshold corresponding to the capacitance sensor.

[0009] According to a non-contact early warning method provided in this application, the process of determining the dedicated peak threshold corresponding to the capacitive sensor includes: Local extremum calculations are performed on the target data corresponding to the capacitance sensor to obtain multiple candidate peak values ​​corresponding to the capacitance sensor. The target data is the historical capacitance value collected by the capacitance sensor, or the target data includes the historical capacitance value collected by the capacitance sensor and the target capacitance value. Select the n target peaks with the smallest values ​​from the plurality of candidate peaks, where n is a positive integer; Based on the n target peak values, a dedicated peak threshold corresponding to the capacitive sensor is determined.

[0010] According to a non-contact early warning method provided in this application, the process of determining the dedicated gradient feature threshold corresponding to the capacitive sensor includes: Gradient feature statistics are performed on the target data corresponding to the capacitive sensor to obtain multiple candidate gradient features corresponding to the capacitive sensor. Local extremum calculations are performed on the multiple candidate gradient features to obtain multiple candidate peak values ​​corresponding to the capacitive sensor; Select the m target gradient peaks with the smallest values ​​from the plurality of candidate peaks, where m is a positive integer; Based on the m target gradient peaks, the dedicated gradient features corresponding to the capacitive sensor are determined.

[0011] According to the non-contact early warning method provided in this application, each of the capacitive sensors is arranged in a non-uniform grid. And / or, the capacitive sensor density in the first region of the target robot surface is greater than the capacitive sensor density in the second region of the target robot surface, wherein the first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold; And / or, each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the target robot's shell.

[0012] This application also provides a collision avoidance method for a robotic arm, including: During the task performed by the robotic arm of the target robot, at least two dimensions of capacitance characteristics are determined based on the target capacitance values ​​collected by each capacitance sensor, and each capacitance sensor is installed on the robotic arm. For each of the capacitive sensors, collision avoidance control is performed on the robotic arm based on the capacitive characteristics and impending contact conditions corresponding to the capacitive sensors, wherein the impending contact conditions indicate that the robotic arm is about to come into contact with an obstacle.

[0013] According to the collision avoidance method for a robotic arm provided in this application, the collision avoidance control of the robotic arm based on the capacitance characteristics corresponding to the capacitance sensors and the impending contact conditions includes: When the capacitance characteristics corresponding to the capacitance sensors meet the conditions for imminent contact, the robotic arm is controlled to avoid obstacles.

[0014] According to the collision avoidance method for a robotic arm provided in this application, before determining at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor during the task execution of the target robot's robotic arm, the method further includes: Receive the task to be executed; The task is broken down into multiple sub-tasks; Based on the multiple sub-tasks, the robotic arm is controlled to perform the task.

[0015] According to the collision avoidance method for a robotic arm provided in this application, the multiple sub-tasks include an object recognition task, an object pose calculation task, and a robotic arm motion control task. The step of controlling the robotic arm to perform the task based on the multiple sub-tasks includes: Based on the object recognition task, the target object corresponding to the task in the task scene is identified and segmented to obtain a segmented image; Based on the object pose calculation task, the coordinate information and depth information of the segmented image are processed to obtain the pose information of the target object. Based on the robotic arm motion control task, the pose information is analyzed to obtain the motion angles and / or positions of each joint on the robotic arm; Based on the motion angles and / or positions of each joint, the robotic arm is controlled to perform the task.

[0016] According to the collision avoidance method for a robotic arm provided in this application, the method further includes: When the capacitance characteristics corresponding to the capacitance sensor meet the conditions for imminent contact, an obstacle warning is broadcast.

[0017] According to the collision avoidance method for a robotic arm provided in this application, each of the capacitive sensors is arranged in a non-uniform grid. And / or, the density of capacitive sensors in the first region of the robotic arm surface is greater than the density of capacitive sensors in the second region of the robotic arm surface, wherein the first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold; And / or, each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the robotic arm housing.

[0018] This application also provides a robotic arm collision avoidance system, including: The tactile sensing module is configured to collect the target capacitance value based on various capacitance sensors during the task performed by the robotic arm of the target robot. The main control module is configured to determine at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor, wherein each capacitance sensor is mounted on the robotic arm. For each of the capacitive sensors, collision avoidance control is performed on the robotic arm based on the capacitive characteristics and impending contact conditions corresponding to the capacitive sensors, wherein the impending contact conditions indicate that the robotic arm is about to come into contact with an obstacle.

[0019] According to the present application, a robotic arm anti-collision system further includes a human-machine communication module, an instruction parsing module, an object detection module, an object pose calculation module, and a robotic arm control module. The human-machine communication module is configured to receive the task to be executed and send the task to the instruction parsing module; The instruction parsing module is configured to split the task into an object recognition task, an object pose calculation task, and a robotic arm motion control task, and then distribute the tasks. The object detection module is configured to identify and segment the target object corresponding to the task in the task scene based on the object recognition task, and obtain a segmented image; and send the coordinate information of the segmented image to the object pose calculation module. The object pose calculation module is configured to process the coordinate information and depth information based on the object pose calculation task to obtain the pose information of the target object; and send the pose information to the robotic arm control module. The robotic arm control module is configured to parse the pose information based on the robotic arm motion control task to obtain the motion angles and / or positions of each joint on the robotic arm; and to control the robotic arm to perform the task based on the motion angles and / or positions of each joint.

[0020] According to the robotic arm anti-collision system provided in this application, the robotic arm control module is further configured to determine the state information of the robotic arm based on the capacitance characteristics corresponding to each of the capacitance sensors and the impending contact condition; and send the state information to the human-machine communication module. The human-machine communication module is also configured to broadcast an obstacle warning when the status information indicates that contact is imminent.

[0021] According to the robotic arm anti-collision system provided in this application, the robotic arm control module is further configured to determine that the state information of the robotic arm is normal when none of the capacitance features corresponding to each of the capacitance sensors meet the impending contact condition; and to determine that the state information of the robotic arm is about to make contact when there are capacitance features corresponding to each of the capacitance sensors that meet the impending contact condition.

[0022] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described non-contact warning methods or the robotic arm anti-collision methods.

[0023] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the non-contact early warning method or the robotic arm anti-collision method described above.

[0024] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the non-contact early warning method or the robotic arm anti-collision method described above.

[0025] The non-contact early warning method, robotic arm collision avoidance method, and system provided in this application determine at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by various capacitance sensors during the target robot's task execution. Each capacitance sensor is installed on the target robot. For each capacitance sensor, when the capacitance characteristics corresponding to that sensor meet the impending contact condition, an obstacle warning is triggered. The impending contact condition indicates that the target robot is about to come into contact with an obstacle. This application detects nearby obstacles through minute changes in the capacitance field, achieving proactive early warning before collision. It can trigger warnings before the robot and / or robotic arm make physical contact with an obstacle, improving safety. It is unaffected by lighting, occlusion, or object material, making it suitable for complex scenarios and highly adaptable to various environments. It has low cost and low computational overhead, requiring no complex vision systems or deep learning models, making deployment simple and suitable for large-scale applications. It also boasts strong real-time performance and fast capacitance sensing response, meeting the obstacle avoidance requirements of highly dynamic environments. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating the non-contact early warning method provided in this application.

[0028] Figure 2 This is a flowchart illustrating the anti-collision method for robotic arms provided in this application.

[0029] Figure 3 This is a schematic diagram of the anti-collision system for the robotic arm provided in this application.

[0030] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] First, a brief explanation of the relevant content involved in this application will be given.

[0033] One related technology provides a human-machine collaborative control method for robotic arms based on tactile sensing. This method primarily generates tactile images by collecting pressure distribution data when the robotic arm contacts obstacles, and then adjusts its motion trajectory in real time by combining this data with visual data. By collecting tactile information from the robotic arm's contact with the external environment, this method compensates for the limitations of machine vision in practical applications, which is easily affected by environmental conditions, and improves the efficiency of robotic arms in performing complex tasks.

[0034] Related technology two provides a machine vision-based obstacle avoidance control method for multi-degree-of-freedom robotic arms. This method acquires 360-degree panoramic image data around the robotic arm through a multi-camera system, and combines this data with 2D and 3D information from RGB (Red, Green, Blue) and depth cameras to construct an environment model. A Kalman filter is then used to predict the trajectory of moving obstacles, and an improved fast expanding random tree algorithm is employed for path planning, ensuring the robotic arm maintains efficient and accurate obstacle avoidance capabilities in dynamic environments. This method enhances the dynamic adaptability and safe operation performance of robotic arms in complex environments.

[0035] Related technology 3 provides a collision detection method for robotic arms based on multi-view object detection. First, it acquires global three-view observation images and inputs them into a pre-trained network model, outputting obstacle IDs and their bounding box information in each view. Then, based on the robotic arm's kinematics, it obtains the coordinates of each joint and converts them into pixel coordinates according to camera calibration data. Finally, based on a collision detection algorithm, it sequentially detects collisions between the robot's upper arm, forearm, and end effector and obstacles. Although this method wastes some of the robotic arm's workspace, it greatly simplifies the spatial arrangement of the robotic arm and obstacles, facilitating collision determination and detection, and improving obstacle avoidance reliability.

[0036] However, the relevant technologies rely on physical contact to trigger collisions, making it impossible to provide proactive warnings before a collision occurs. Furthermore, the tactile sensors can only respond after a collision has occurred, which is insufficient to meet the real-time obstacle avoidance requirements in highly dynamic environments.

[0037] Although the second related technology improves the dynamic obstacle avoidance capability of the robotic arm, it relies on a complex vision system, has a large computational cost, and is easily affected by factors such as changes in lighting and occlusion, resulting in a significant decrease in performance in low light or high reflectivity environments.

[0038] The third related technology requires pre-trained network models and relies on accurate camera calibration, resulting in high deployment and maintenance costs. Furthermore, the limited field of view of multi-view systems can lead to blind spots, affecting the comprehensiveness of detection.

[0039] To address the aforementioned issues, this application provides a non-contact early warning method, a robotic arm collision avoidance method, and a system.

[0040] It should be noted that the non-contact early warning method, robotic arm anti-collision method and system provided in this application can be applied to scenarios requiring human-machine interaction, such as industrial robotic arms, collaborative robots, and humanoid robots.

[0041] The following is combined with Figures 1-4 This application describes a non-contact early warning method, a robotic arm collision avoidance method, and a system.

[0042] Figure 1 This is a flowchart illustrating the non-contact early warning method provided in this application, such as... Figure 1 As shown, the method includes steps 101 and 102.

[0043] Step 101: During the task execution of the target robot, based on the target capacitance values ​​collected by each capacitance sensor, at least two dimensions of capacitance characteristics are determined, and each capacitance sensor is installed on the target robot.

[0044] Specifically, a target robot refers to a robot with a robotic arm. The task performed by a target robot can be a grasping and placing task, that is, the task of grasping or placing objects.

[0045] In practical applications, at least two capacitive sensors are installed on the target robot. Since the target robot performs grasping and releasing tasks, the main action is taken by the robotic arm, which is the primary point of contact with obstacles. To reduce the number of capacitive sensors and lower the cost and power consumption of the target robot, the capacitive sensors can be mainly installed on the robotic arm for non-contact early warning.

[0046] It should be noted that the capacitive sensors described may be arranged in a non-uniform grid, such as a hexagonal honeycomb structure.

[0047] The density of capacitive sensors in a first region of the target robot's surface is greater than the density of capacitive sensors in a second region of the target robot's surface. The first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold. That is, in regions with high curvature, such as near joints, the density of capacitive sensors is increased, while in flat regions, the density is appropriately reduced to improve coverage while decreasing redundant nodes.

[0048] Each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the target robot's shell.

[0049] Specifically, a multi-layer flexible circuit board can be used to embed the capacitive sensor into the inner layer of the target robot's shell, such as in the forearm shell of a robotic arm. Furthermore, an insulating decorative layer ≤0.5mm thick, such as matte paint or composite material, can be applied to the surface to maintain a consistent appearance of the target robot. The PCB (Printed Circuit Board) of the capacitive sensor is mounted inside the shell, with the capacitive sensor positioned on both sides of the PCB. The capacitive sensor interacts with the target robot's main control module via the RS485 communication protocol.

[0050] Specifically, during the task execution of the target robot, real-time sensor data, i.e., the target capacitance value, is acquired from each capacitance sensor. Further, multi-dimensional capacitance feature extraction is performed on the target capacitance value acquired by each capacitance sensor to obtain at least two dimensions of capacitance features corresponding to each sensor, such as at least two of the following dimensions: peak value, gradient value, and consistency value.

[0051] Step 102: For each of the capacitive sensors, if the capacitive characteristics corresponding to the capacitive sensor meet the imminent contact condition, an obstacle warning is triggered, wherein the imminent contact condition indicates that the target robot is about to contact an obstacle.

[0052] Specifically, the imminent contact condition can be characterized by any part of the target robot being about to come into contact with an obstacle, such as the robotic arm of the target robot being about to come into contact with an obstacle.

[0053] In practical applications, if at least one of the capacitive sensors installed on the target robot is a target capacitive sensor, it indicates that the target robot is about to come into contact with an obstacle, and an obstacle warning needs to be triggered. Specifically, the capacitance characteristics of each target capacitive sensor must meet the conditions for impending contact.

[0054] This application provides a non-contact early warning method that detects nearby obstacles by detecting minute changes in the capacitance field, enabling proactive early warning before collisions. This method can trigger warnings before physical contact between the robot and / or robotic arm and the obstacle, improving safety. It is unaffected by lighting, occlusion, or object material, making it suitable for complex scenarios and highly adaptable to various environments. It is low-cost and computationally inefficient, requiring no complex vision systems or deep learning models, making deployment simple and suitable for large-scale applications. Furthermore, it offers strong real-time performance with fast capacitance sensing response, meeting obstacle avoidance requirements in highly dynamic environments.

[0055] Optionally, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; The method for determining capacitance characteristics in at least two dimensions based on the target capacitance values ​​collected by each capacitance sensor includes: For each of the capacitive sensors, local extrema calculation is performed on the target capacitance value collected by the capacitive sensor to obtain the peak value corresponding to the capacitive sensor; By selecting a fixed time window, gradient feature statistics are performed on the target capacitance value collected by the capacitance sensor to obtain the gradient feature corresponding to the capacitance sensor. Based on the target capacitance value collected by the capacitance sensor and the target capacitance value collected by the adjacent capacitance sensors, a sign consistency judgment is performed to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor.

[0056] Specifically, the target capacitance value C(t) = [C1(t), ..., C2(t)] can be synchronously collected by multiple (at least two) capacitive sensors arranged on the target robot. n [(t)], and the change in the target capacitance value ΔC(t) = [ΔC1(t), ..., ΔC n[t], establish a three-dimensional feature vector, including amplitude features (peak value), gradient features (dΔC / dt), and the changing trends of adjacent capacitive sensors. Where t represents time, C represents capacitance value, n represents the number of capacitive sensors, C(t) is the set of all target capacitance values ​​at time t, C1(t) is the first target capacitance value at time t, and C... n Let ΔC(t) be the nth target capacitance value at time t, ΔC(t) be the set of changes in all target capacitance values ​​at time t, and ΔC1(t) be the change in the first target capacitance value at time t. n (t) represents the change in the nth target capacitance value at time t.

[0057] In practical applications, for each capacitive sensor, the peak value, gradient features, and the changing trends of adjacent capacitive sensors can be extracted according to the following process.

[0058] Peak value calculation: For the capacitance data (target capacitance value) collected by the capacitance sensor, peak value calculation is performed, including: calculating the local extrema of the target capacitance value to obtain the corresponding peak value of the capacitance sensor, such as the minimum value.

[0059] Gradient feature calculation: Since the capacitance data (target capacitance value) of the capacitance sensor changes drastically when it approaches an object (obstacle), a fixed time window w can be selected to statistically analyze the gradient feature of the target capacitance value.

[0060] Determining the trend of adjacent sensors: Since the area where capacitive sensors are arranged and the proximity of obstacles to the sensors both affect the values ​​sensed by the sensors (target capacitance value), when there are no abnormalities in the capacitive sensors and no external "noise" interference, the trend of adjacent capacitive sensors should be consistent when approaching an obstacle. Detecting the consistency of trends between adjacent sensors can improve the accuracy of obstacle detection. Considering computational complexity, the consistency of the derivative signs can be checked between the target capacitance values ​​collected by the capacitive sensors and those collected by adjacent capacitive sensors to determine whether the trends of adjacent sensors are consistent.

[0061] It should be noted that before performing peak value calculation and / or gradient feature calculation, the capacitance data (target capacitance value) collected by the capacitance sensor can be filtered, such as by median filtering or first-order filtering. Then, peak value calculation and / or gradient feature calculation can be performed on the filtered capacitance data (target capacitance value).

[0062] In this embodiment, capacitance characteristics are determined from three dimensions: peak value, gradient characteristics, and the changing trend of adjacent capacitance sensors. This improves the reliability and accuracy of capacitance characteristics and is beneficial to improving the reliability and accuracy of non-contact early warning.

[0063] Optionally, the step of performing a sign consistency judgment based on the target capacitance value collected by the capacitance sensor and the target capacitance values ​​collected by adjacent capacitance sensors to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor includes: The target capacitance value collected by the capacitance sensor is calculated by differential value to obtain a first differential value, and the target capacitance value collected by the adjacent capacitance sensor is calculated by differential value to obtain a second differential value; Perform a sign consistency check on the first difference value and the second difference value; If the signs are the same, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensors are determined to be in the same direction. If the symbols are inconsistent, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensor are determined to be opposite.

[0064] Specifically, the steps for judging the changing trend of adjacent sensors can be as follows: (1) Calculate the first difference value of the capacitive sensor and the second difference value of the adjacent capacitive sensor: ΔC i (t) = C i (t) - C i (t-1), where t represents time, C i (2) Symbol extraction: Sign(ΔC1(t)) × Sign(ΔC2(t)), where C1 represents the current capacitor sensor and C2 represents the adjacent capacitor sensor; if the result of symbol extraction is 1, it indicates that the trend is in the same direction, and if the result of symbol extraction is -1, it indicates that the trend is in the opposite direction.

[0065] In this embodiment of the application, the sign consistency judgment is performed by using the difference value, which can improve the efficiency and accuracy of the changing trend of adjacent capacitive sensors.

[0066] Optionally, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; Before triggering an obstacle warning when the capacitance characteristics corresponding to each capacitance sensor meet the impending contact condition, the method further includes: The consistency ratio of the changing trends of adjacent capacitive sensors corresponding to the capacitive sensor is determined to be in the same direction; Determine whether the peak value, gradient feature, and consistency ratio corresponding to the capacitive sensor meet the upcoming contact conditions.

[0067] Specifically, based on the changing trends of adjacent capacitive sensors, the consistency ratio can be calculated: assuming that the number of times adjacent capacitive sensor changing trends are in the same direction is N within T instances, the consistency ratio R = N / T can be calculated. 100%. Furthermore, the peak value, gradient characteristics, and consistency ratio of the capacitive sensor are compared with the upcoming contact conditions to determine whether the conditions are met. This ensures the reliability of the judgment.

[0068] Optionally, the peak value is a minimum value; The conditions for impending contact are that the minimum value corresponding to the capacitive sensor is less than a specified peak threshold, the gradient feature corresponding to the capacitive sensor is greater than a specified gradient feature threshold, and the consistency ratio corresponding to the capacitive sensor is greater than a specified ratio threshold. The specified peak threshold is either a general peak threshold or a dedicated peak threshold corresponding to the capacitive sensor; The specified gradient feature threshold is either a general gradient feature threshold or a dedicated gradient feature threshold corresponding to the capacitive sensor. The specified ratio threshold is either a general ratio threshold or a specific ratio threshold corresponding to the capacitance sensor.

[0069] Specifically, for any given capacitive sensor, if there exists a C of the capacitive sensor... i (t) <T c ΔC i (t)>T ΔC And R i >R t If this occurs, an obstacle warning will be triggered. Among them, C... i The minimum value of the i-th capacitive sensor is represented by t, which represents time, and ΔC. i R represents the gradient characteristics of the i-th capacitive sensor. i T represents the consistency ratio of the i-th capacitive sensor. c To specify the peak threshold, T ΔC To specify the gradient feature threshold, R t Specify the percentage threshold.

[0070] Optionally, different capacitive sensors may use the same peak threshold, i.e., the specified peak threshold may be a general peak threshold; different capacitive sensors may also use their own corresponding peak thresholds, i.e., the specified peak threshold may be a dedicated peak threshold.

[0071] Different capacitive sensors can use the same gradient feature threshold, that is, the specified gradient feature threshold can be a general gradient feature threshold; different capacitive sensors can also use their own corresponding gradient feature thresholds, that is, the specified gradient feature threshold can be a special gradient feature threshold.

[0072] Different capacitive sensors can use the same proportional threshold, meaning the specified proportional threshold can be a general proportional threshold; different capacitive sensors can also use their own corresponding proportional thresholds, meaning the specified proportional threshold can be a dedicated proportional threshold.

[0073] Optionally, the process of determining the dedicated peak threshold corresponding to the capacitive sensor includes: Local extremum calculations are performed on the target data corresponding to the capacitance sensor to obtain multiple candidate peak values ​​corresponding to the capacitance sensor. The target data is the historical capacitance value collected by the capacitance sensor, or the target data includes the historical capacitance value collected by the capacitance sensor and the target capacitance value. Select the n target peaks with the smallest values ​​from the plurality of candidate peaks, where n is a positive integer; Based on the n target peak values, a dedicated peak threshold corresponding to the capacitive sensor is determined.

[0074] Specifically, for each capacitance sensor, the target data (historical capacitance value, or historical capacitance value and target capacitance) collected by the capacitance sensor is filtered; local minima are calculated on the filtered target data to obtain multiple (at least two) candidate peaks; the multiple candidate peaks are sorted in ascending order; the top n smallest target peaks are determined, and their average value is taken as the dedicated peak threshold T. c .

[0075] In addition, after sorting multiple candidate peaks in ascending order, peaks that are close in size can be filtered out, and then the top n smallest target peaks can be determined, with the average value taken as the dedicated peak threshold T. c .

[0076] In this embodiment, the dedicated peak threshold is determined based on the target data collected by each capacitive sensor, which can improve the reliability and accuracy of the dedicated peak threshold.

[0077] Optionally, the process of determining the dedicated gradient feature threshold corresponding to the capacitive sensor includes: Gradient feature statistics are performed on the target data corresponding to the capacitive sensor to obtain multiple candidate gradient features corresponding to the capacitive sensor. Local extremum calculations are performed on the multiple candidate gradient features to obtain multiple candidate peak values ​​corresponding to the capacitive sensor; Select the m target gradient peaks with the smallest values ​​from the plurality of candidate peaks, where m is a positive integer; Based on the m target gradient peaks, the dedicated gradient features corresponding to the capacitive sensor are determined.

[0078] Specifically, for each capacitance sensor, the target data (historical capacitance value, or historical capacitance value and target capacitance) collected by the capacitance sensor is filtered; a fixed time window w is selected, and gradient feature statistics are performed on the filtered target data to obtain multiple (at least two) candidate gradient features. Then, local minima are calculated on the multiple candidate gradient features to obtain multiple (at least two) alternative peak values; then, the multiple alternative peak values ​​are sorted in ascending order; the first m smallest target peak values ​​are determined, and the average value is taken as the dedicated gradient feature threshold T. ΔC .

[0079] Furthermore, after sorting multiple candidate peaks in ascending order, peaks that are close in size can be filtered out, and then the top m smallest target peaks can be determined, with their average value used as the dedicated gradient feature threshold T. ΔC .

[0080] In this embodiment, the dedicated gradient feature threshold is determined based on the target data collected by each capacitive sensor, which can improve the reliability and accuracy of the dedicated gradient feature threshold.

[0081] The following describes the anti-collision method for robotic arms provided in this application. The anti-collision method for robotic arms described below can be referred to in correspondence with the non-contact early warning method described above.

[0082] Figure 2 This is a flowchart illustrating the collision avoidance method for the robotic arm provided in this application, as shown below. Figure 2 As shown, the method includes steps 201 and 202.

[0083] Step 201: During the process of the target robot's robotic arm performing the task, based on the target capacitance values ​​collected by each capacitance sensor, at least two dimensions of capacitance characteristics are determined, and each capacitance sensor is installed on the robotic arm.

[0084] Specifically, the task performed by the robotic arm can be a grasping and placing task, that is, the task of grasping or placing objects.

[0085] In practical applications, at least two capacitive sensors are installed on the robotic arm of the target robot to prevent collisions.

[0086] It should be noted that the capacitive sensors described may be arranged in a non-uniform grid, such as a hexagonal honeycomb structure.

[0087] The density of capacitive sensors in a first region of the robotic arm's surface is greater than that in a second region of the robotic arm's surface. The first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold. That is, in regions with high curvature, such as near joints, the density of capacitive sensors is increased, while in flat regions, the density is appropriately reduced, thereby improving coverage while reducing redundant nodes.

[0088] Each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the robotic arm housing.

[0089] Specifically, a multi-layer flexible circuit board can be used to embed the capacitive sensor into the inner layer of the robotic arm's housing, such as in the forearm housing attachment. Furthermore, an insulating decorative layer of ≤0.5mm, such as matte paint or composite material, can be applied to the surface to maintain a consistent appearance of the target robot. The PCB (Printed Circuit Board) of the capacitive sensor is mounted inside the housing, with the capacitive sensor positioned on both sides of the PCB. The capacitive sensor interacts with the target robot's main control module via the RS485 communication protocol.

[0090] Specifically, during the robotic arm's task execution, real-time sensor data, i.e., the target capacitance value, is acquired from each capacitance sensor. Further, multi-dimensional capacitance feature extraction is performed on the target capacitance value acquired by each capacitance sensor, resulting in at least two dimensions of capacitance features corresponding to each sensor, such as at least two of the following: peak value dimension, gradient dimension, and consistency dimension.

[0091] Step 202: For each of the capacitive sensors, based on the capacitance characteristics and impending contact conditions corresponding to each capacitive sensor, perform anti-collision control on the robotic arm, wherein the impending contact conditions indicate that the robotic arm is about to come into contact with an obstacle.

[0092] Specifically, the imminent contact condition can be characterized by any part of the robotic arm being about to come into contact with an obstacle, such as the robotic arm's fingers being about to come into contact with an obstacle.

[0093] In practical applications, based on the impending contact conditions and the capacitance characteristics of each capacitive sensor on the robotic arm, it is possible to monitor whether the robotic arm is about to contact an obstacle, and to implement anti-collision control when the robotic arm is about to contact the obstacle.

[0094] This application provides a collision avoidance method for robotic arms, which detects nearby obstacles by detecting minute changes in the capacitance field, enabling proactive early warning before collision. This method can trigger a warning before the robot and / or robotic arm makes physical contact with an obstacle, improving safety. It is unaffected by lighting, occlusion, or object material, making it suitable for complex scenarios and highly adaptable to various environments. It is low-cost and computationally inefficient, requiring no complex vision systems or deep learning models, making it easy to deploy and suitable for large-scale applications. Furthermore, it offers strong real-time performance with fast capacitive sensing response, meeting the obstacle avoidance requirements of highly dynamic environments.

[0095] Optionally, the step of performing anti-collision control on the robotic arm based on the capacitance characteristics corresponding to each of the capacitance sensors and the impending contact conditions includes: When the capacitance characteristics corresponding to the capacitance sensors meet the conditions for imminent contact, the robotic arm is controlled to avoid obstacles.

[0096] In practical applications, if at least one target capacitance sensor is present among all the capacitance sensors installed on the robotic arm, it indicates that the robotic arm is about to come into contact with an obstacle. In this case, collision avoidance control is required, such as adjusting the robotic arm's posture. If no target capacitance sensor is present, the robotic arm continues to perform its task until it is completed. The capacitance characteristics of the target capacitance sensors must meet the conditions for impending contact.

[0097] In this way, by judging whether the capacitance characteristics meet the conditions for imminent contact, the state of the robotic arm can be quickly and accurately determined, and anti-collision control can be performed on the robotic arm when it is about to contact the obstacle, thereby improving the anti-collision efficiency.

[0098] Optionally, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; The method for determining capacitance characteristics in at least two dimensions based on the target capacitance values ​​collected by each capacitance sensor includes: For each of the capacitive sensors, local extrema calculation is performed on the target capacitance value collected by the capacitive sensor to obtain the peak value corresponding to the capacitive sensor; By selecting a fixed time window, gradient feature statistics are performed on the target capacitance value collected by the capacitance sensor to obtain the gradient feature corresponding to the capacitance sensor. Based on the target capacitance value collected by the capacitance sensor and the target capacitance value collected by the adjacent capacitance sensors, a sign consistency judgment is performed to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor.

[0099] Optionally, the step of performing a sign consistency judgment based on the target capacitance value collected by the capacitance sensor and the target capacitance values ​​collected by adjacent capacitance sensors to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor includes: The target capacitance value collected by the capacitance sensor is calculated by differential value to obtain a first differential value, and the target capacitance value collected by the adjacent capacitance sensor is calculated by differential value to obtain a second differential value; Perform a sign consistency check on the first difference value and the second difference value; If the signs are the same, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensors are determined to be in the same direction. If the symbols are inconsistent, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensor are determined to be opposite.

[0100] Optionally, the at least two dimensions of capacitance characteristics include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; Before triggering an obstacle warning when the capacitance characteristics corresponding to each capacitance sensor meet the impending contact condition, the method further includes: The consistency ratio of the changing trends of adjacent capacitive sensors corresponding to the capacitive sensor is determined to be in the same direction; Determine whether the peak value, gradient feature, and consistency ratio corresponding to the capacitive sensor meet the upcoming contact conditions.

[0101] Optionally, the peak value is a minimum value; The conditions for impending contact are that the minimum value corresponding to the capacitive sensor is less than a specified peak threshold, the gradient feature corresponding to the capacitive sensor is greater than a specified gradient feature threshold, and the consistency ratio corresponding to the capacitive sensor is greater than a specified ratio threshold. The specified peak threshold is either a general peak threshold or a dedicated peak threshold corresponding to the capacitive sensor; The specified gradient feature threshold is either a general gradient feature threshold or a dedicated gradient feature threshold corresponding to the capacitive sensor. The specified ratio threshold is either a general ratio threshold or a specific ratio threshold corresponding to the capacitance sensor.

[0102] Optionally, the process of determining the dedicated peak threshold corresponding to the capacitive sensor includes: Local extremum calculations are performed on the target data corresponding to the capacitance sensor to obtain multiple candidate peak values ​​corresponding to the capacitance sensor. The target data is the historical capacitance value collected by the capacitance sensor, or the target data includes the historical capacitance value collected by the capacitance sensor and the target capacitance value. Select the n target peaks with the smallest values ​​from the plurality of candidate peaks, where n is a positive integer; Based on the n target peak values, a dedicated peak threshold corresponding to the capacitive sensor is determined.

[0103] Optionally, the process of determining the dedicated gradient feature threshold corresponding to the capacitive sensor includes: Gradient feature statistics are performed on the target data corresponding to the capacitive sensor to obtain multiple candidate gradient features corresponding to the capacitive sensor. Local extremum calculations are performed on the multiple candidate gradient features to obtain multiple candidate peak values ​​corresponding to the capacitive sensor; Select the m target gradient peaks with the smallest values ​​from the plurality of candidate peaks, where m is a positive integer; Based on the m target gradient peaks, the dedicated gradient features corresponding to the capacitive sensor are determined.

[0104] Optionally, before determining at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor during the task performed by the robotic arm of the target robot, the method further includes: Receive the task to be executed; The task is broken down into multiple sub-tasks; Based on the multiple sub-tasks, the robotic arm is controlled to perform the task.

[0105] Specifically, the task to be executed can be obtained by the human-machine communication module. The human-machine communication module can obtain natural language task instructions issued by the user through voice, text input, or button operation to instruct the robotic arm to perform a task. These natural language task instructions carry the task to be executed.

[0106] Furthermore, the human-machine communication module can send the task to be executed to the instruction parsing module. The instruction parsing module can, based on a language processing model, such as the Qianwen edge-side big model, break down the task to be executed (natural language task) received from the human-machine communication module into steps that the machine can execute, i.e., at least two sub-tasks. For example, for an object grasping task, it will be broken down into recognizing the specific object, calculating the object's pose, and controlling the robotic arm's motion.

[0107] It should be noted that the task to be executed can be converted into natural language by the human-machine communication module and / or instruction parsing module.

[0108] After obtaining multiple subtasks, the robotic arm can be controlled to perform tasks by executing each subtask step by step.

[0109] In this embodiment of the application, by breaking down a task into sub-tasks that the machine can execute, the task execution efficiency and success rate can be improved.

[0110] Optionally, the multiple sub-tasks include object recognition, object pose calculation, and robotic arm motion control. The step of controlling the robotic arm to perform the task based on the multiple sub-tasks includes: Based on the object recognition task, the target object corresponding to the task in the task scene is identified and segmented to obtain a segmented image; Based on the object pose calculation task, the coordinate information and depth information of the segmented image are processed to obtain the pose information of the target object. Based on the robotic arm motion control task, the pose information is analyzed to obtain the motion angles and / or positions of each joint on the robotic arm; Based on the motion angles and / or positions of each joint, the robotic arm is controlled to perform the task.

[0111] In practical applications, the object detection module primarily performs object recognition tasks. It can use visual segmentation models, such as SAM2 (Segment Anything Model 2), to identify and segment target objects in the task scene, obtaining segmented images and their coordinate information. The coordinate information of the segmented images is then transmitted to the object pose calculation module. The images of the task scene can be acquired by imaging devices mounted on the robot, and the target objects are those to be grasped or placed.

[0112] The object pose calculation module primarily performs the task of calculating the object pose. It combines image segmentation information (the coordinate information of the segmented image) with corresponding depth information to calculate the grasping and releasing pose of the target object, and then transmits this pose to the robotic arm motion control module. The GraspNet (high-efficiency convolutional neural network) method can be used to calculate the grasping and releasing pose, while the depth information can be obtained from a depth camera mounted on the robot.

[0113] The robotic arm motion control module primarily performs the robotic arm motion control tasks. Using object pose information (the grasping and releasing pose of the target object), the module calculates the motion angles and / or positions of each joint of the robotic arm based on inverse kinematics, and controls the robotic arm to perform tasks based on these motion angles and / or positions.

[0114] It should be noted that during the execution of a task, when the robotic arm approaches an obstacle or an object, the main control module will analyze and judge the capacitance data (target capacitance value) collected by the tactile sensing module (including at least two capacitive sensors) in real time. If it senses that it is approaching an obstacle, it will send the judgment result to the robotic arm control module to prevent the robotic arm from colliding.

[0115] In this embodiment, parameters for controlling the robotic arm are obtained by executing each sub-task step by step, thereby controlling the robotic arm to perform tasks. This not only helps to prevent collisions with the robotic arm to a certain extent, but also improves the efficiency and success rate of task execution.

[0116] Optionally, the method further includes: When the capacitance characteristics corresponding to the capacitance sensor meet the conditions for imminent contact, an obstacle warning is broadcast.

[0117] It should be noted that after the main control module sends the judgment result (whether the robotic arm is about to contact the obstacle) to the robotic arm control module, the robotic arm control module can generate the robotic arm's status information based on the judgment result. For example, if none of the capacitance characteristics corresponding to each of the capacitance sensors meet the impending contact condition, i.e., the judgment result is that the robotic arm is not about to contact the obstacle, the robotic arm's status information is determined to be normal; if some of the capacitance characteristics corresponding to the capacitance sensors meet the impending contact condition, i.e., the judgment result is that the robotic arm is about to contact the obstacle, the robotic arm's status information is determined to be about to contact.

[0118] Furthermore, the robotic arm control module sends the robotic arm's status information to the human-machine communication module. The human-machine communication module receives the robotic arm's status information returned from the robotic arm control module. If the status information indicates that the robotic arm is about to come into contact with an obstacle (i.e., the status information indicates imminent contact), then an obstacle warning is issued.

[0119] In this embodiment, the robot arm status is used to issue early warnings, which can alert the user that the robot arm is in an abnormal situation, thus improving the user-friendliness.

[0120] The robotic arm collision avoidance system provided in this application is described below. The robotic arm collision avoidance system described below can be referred to in correspondence with the non-contact early warning method and robotic arm collision avoidance method described above.

[0121] Figure 3 This is a structural schematic diagram of the robotic arm anti-collision system provided in this application, as shown below. Figure 3 As shown, the system includes: The tactile sensing module 301 is configured to collect the target capacitance value based on each capacitance sensor during the task performed by the robotic arm of the target robot. The main control module 302 is configured to determine at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor, wherein each capacitance sensor is mounted on the robotic arm; for each capacitance sensor, the robotic arm is subjected to anti-collision control according to the capacitance characteristics corresponding to the capacitance sensor and the impending contact condition, wherein the impending contact condition indicates that the robotic arm is about to come into contact with an obstacle.

[0122] In practical applications, a tactile sensing module is installed on the robotic arm of the target robot. The tactile sensing module includes at least two capacitive sensors for collision avoidance of the robotic arm.

[0123] It should be noted that the capacitive sensors described may be arranged in a non-uniform grid, such as a hexagonal honeycomb structure.

[0124] The density of capacitive sensors in a first region of the robotic arm's surface is greater than that in a second region of the robotic arm's surface. The first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold. That is, in regions with high curvature, such as near joints, the density of capacitive sensors is increased, while in flat regions, the density is appropriately reduced, thereby improving coverage while reducing redundant nodes.

[0125] Each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the robotic arm housing.

[0126] Specifically, a multi-layer flexible circuit board can be used to embed the capacitive sensor into the inner layer of the robotic arm's housing, such as in the forearm housing attachment. Furthermore, an insulating decorative layer of ≤0.5mm, such as matte paint or composite material, can be applied to the surface to maintain a consistent appearance of the target robot. The PCB (Printed Circuit Board) of the capacitive sensor is mounted inside the housing, with the capacitive sensor positioned on both sides of the PCB. The capacitive sensor interacts with the target robot's main control module via the RS485 communication protocol.

[0127] Specifically, during the robotic arm's task execution, the tactile sensing module collects sensor data, i.e., the target capacitance value, in real time through various capacitive sensors. Then, the collected target capacitance value is transmitted to the main control module.

[0128] Furthermore, the main control module performs multi-dimensional capacitance feature extraction on the target capacitance values ​​collected by each capacitance sensor to obtain at least two dimensions of capacitance features corresponding to each capacitance sensor, such as at least two dimensions of peak value, gradient value, and consistency value.

[0129] Specifically, the imminent contact condition can be characterized by any part of the robotic arm being about to come into contact with an obstacle, such as the robotic arm's fingers being about to come into contact with an obstacle.

[0130] In practical applications, the main control module can monitor whether the robotic arm is about to come into contact with an obstacle based on the impending contact conditions and the capacitance characteristics of each capacitive sensor on the robotic arm. When the robotic arm is about to come into contact with an obstacle, it can perform anti-collision control on the robotic arm.

[0131] This application provides a robotic arm collision avoidance system that detects nearby obstacles by detecting minute changes in the capacitance field, enabling proactive warnings before collisions. The system can trigger warnings before the robot and / or robotic arm makes physical contact with an obstacle, improving safety. It is unaffected by lighting, occlusion, or object material, making it suitable for complex scenarios and highly adaptable to various environments. It is low-cost and computationally inefficient, requiring no complex vision systems or deep learning models, making deployment simple and suitable for large-scale applications. Furthermore, it offers strong real-time performance with fast capacitive sensing response, meeting obstacle avoidance requirements in highly dynamic environments.

[0132] Optionally, the system further includes a human-machine communication module, an instruction parsing module, an object detection module, an object pose calculation module, and a robotic arm control module; The human-machine communication module 303 is configured to receive the task to be executed and send the task to the instruction parsing module; The instruction parsing module 304 is configured to split the task into an object recognition task, an object pose calculation task, and a robotic arm motion control task, and then distribute the tasks. The object detection module 305 is configured to identify and segment the target object corresponding to the task in the task scene based on the object recognition task, and obtain a segmented image; and send the coordinate information of the segmented image to the object pose calculation module. The object pose calculation module 306 is configured to process the coordinate information and depth information based on the object pose calculation task to obtain the pose information of the target object; and send the pose information to the robotic arm control module. The robotic arm control module 307 is configured to parse the pose information based on the robotic arm motion control task to obtain the motion angles and / or positions of each joint on the robotic arm; and to control the robotic arm to perform the task based on the motion angles and / or positions of each joint.

[0133] Specifically, the task to be executed can be obtained by the human-machine communication module. The human-machine communication module can obtain natural language task instructions issued by the user through voice, text input, or button operation to instruct the robotic arm to perform a task. These natural language task instructions carry the task to be executed.

[0134] Furthermore, the human-machine communication module can send the task to be executed to the instruction parsing module. The instruction parsing module can, based on a language processing model, such as the Qianwen edge-side big model, decompose the task to be executed (natural language task) received from the human-machine communication module into three steps that the machine can execute, namely, object recognition task, object pose calculation task, and robotic arm motion control task.

[0135] The object detection module primarily performs object recognition tasks. It can use visual segmentation models, such as SAM2 (Segment Anything Model 2), to identify and segment target objects in the task scene, obtaining segmented images and their coordinate information. The coordinate information of the segmented images is then transmitted to the object pose calculation module. The images of the task scene can be acquired by imaging devices mounted on the robot, and the target objects are those to be grasped or placed.

[0136] The object pose calculation module primarily performs the task of calculating the object pose. It combines image segmentation information (the coordinate information of the segmented image) with corresponding depth information to calculate the grasping and releasing pose of the target object, and then transmits this pose to the robotic arm motion control module. The GraspNet (high-efficiency convolutional neural network) method can be used to calculate the grasping and releasing pose, while the depth information can be obtained from a depth camera mounted on the robot.

[0137] The robotic arm motion control module primarily performs the robotic arm motion control tasks. Using object pose information (the grasping and releasing pose of the target object), the module calculates the motion angles and / or positions of each joint of the robotic arm based on inverse kinematics, and controls the robotic arm to perform tasks based on these motion angles and / or positions.

[0138] It should be noted that during the execution of a task, when the robotic arm approaches an obstacle or an object, the main control module will analyze and judge the capacitance data (target capacitance value) collected by the tactile sensing module (including at least two capacitive sensors) in real time. If it senses that it is approaching an obstacle, it will send the judgment result to the robotic arm control module to prevent the robotic arm from colliding.

[0139] In this embodiment, parameters for controlling the robotic arm are obtained by executing each sub-task step by step, thereby controlling the robotic arm to perform tasks. This not only helps to prevent collisions with the robotic arm to a certain extent, but also improves the efficiency and success rate of task execution.

[0140] Optionally, the robotic arm control module 307 is further configured to determine the state information of the robotic arm based on the capacitance characteristics corresponding to each of the capacitance sensors and the impending contact condition; and send the state information to the human-machine communication module. The human-machine communication module 303 is also configured to broadcast an obstacle warning when the status information indicates that contact is imminent.

[0141] In practical applications, after the main control module sends the judgment result (whether the robotic arm is about to contact the obstacle) to the robotic arm control module, the robotic arm control module can generate the robotic arm's status information based on the judgment result. For example, if none of the capacitance characteristics corresponding to each of the capacitance sensors meet the impending contact condition, i.e., the judgment result indicates that the robotic arm is not about to contact the obstacle, the robotic arm's status information is determined to be normal. If some of the capacitance characteristics corresponding to the capacitance sensors meet the impending contact condition, i.e., the judgment result indicates that the robotic arm is about to contact the obstacle, the robotic arm's status information is determined to be about to contact the obstacle.

[0142] Furthermore, the robotic arm control module sends the robotic arm's status information to the human-machine communication module. The human-machine communication module receives the robotic arm's status information returned from the robotic arm control module. If the status information indicates that the robotic arm is about to come into contact with an obstacle (i.e., the status information indicates imminent contact), then an obstacle warning is issued.

[0143] In this embodiment, the robot arm status is used to issue early warnings, which can alert the user that the robot arm is in an abnormal situation, thus improving the user-friendliness.

[0144] The non-contact early warning method, robotic arm collision avoidance method and system provided in this application are as follows: Based on capacitive tactile sensors, this invention monitors and analyzes the target capacitance value of the sensor in real time during grasping and releasing tasks. When the target capacitance value meets the non-contact collision detection conditions (approaching contact conditions), the robot and / or robotic arm's movement speed can be reduced or instantaneously braked, and a voice alert is given to the user. This capacitive sensor-based pre-collision protection, by monitoring minute changes in the capacitance field as an object approaches, can trigger an early warning within 0-50mm of the obstacle's distance from the robotic arm surface. Compared to traditional tactile sensor solutions that require physical contact to respond, this reduces the safety response time by 300-1000ms. The non-contact capacitive sensing early warning mechanism is suitable for humanoid robot robotic arm movement scenarios. Therefore, this application has a response time advantage by advancing the detection point from when a collision occurs to before the collision. At a humanoid robotic arm movement speed of 0.1m / s, the braking distance can be shortened from 50mm to 15mm. This significantly reduces the probability of physical collisions with the robot, fundamentally solving the safety hazards of "post-collision processing" in contact-based solutions.

[0145] By employing a concealed, non-uniformly distributed capacitive sensor layout, the sensors are positioned inside the robot and / or robotic arm, without affecting the robot's or arm's aesthetic design. An adaptive sensor placement strategy, based on the curved surface characteristics of the robot and / or robotic arm, utilizes a dense hexagonal arrangement in high-risk areas such as joints and a sparse grid in low-curvature areas, achieving an effective coverage rate of over 85%. By embedding the sensors within the robotic arm's outer shell (1-2 mm from the surface) and covering them with an ultra-thin (≤0.5 mm) insulating decorative layer, both industrial design integrity and the sensitivity of the capacitive field to external interference are maintained. By extracting multi-dimensional capacitance features and combining peak values, gradient features, and the changing trends of adjacent capacitance sensors, the robustness of the collision avoidance detection mechanism is improved. A three-level detection model based on spatiotemporal feature fusion is proposed: the first level detects abnormal regions using the difference matrix between adjacent sensors; the second level confirms dynamic threats by combining capacitance gradients (dC / dt > 10 pF / ms); and the third level reduces the false alarm rate through multi-sensor collaborative verification. This algorithm maintains a high detection rate while reducing the false trigger rate, significantly outperforming single-threshold detection methods.

[0146] The non-contact early warning method, robotic arm collision avoidance method, and system provided in this application employ capacitive tactile sensors, offering a highly advantageous collision avoidance solution for robots and / or robotic arms: It eliminates the need for complex sensor systems or advanced algorithms, significantly reducing costs and computational overhead; the capacitive field detection mechanism can sensitively detect minute changes before the robot and / or robotic arm makes physical contact with obstacles, thereby triggering early warnings and greatly improving the safety of robotic arms in complex environments; this application is unaffected by factors such as lighting, occlusion, or object material, exhibiting strong environmental adaptability and meeting the real-time obstacle avoidance requirements in highly dynamic environments; furthermore, the efficient, low-cost, and highly robust solution provided in this application not only effectively improves the safety and reliability of robots and / or robotic arms but also provides a solid technical guarantee for the widespread application of robots and / or robotic arms in complex scenarios.

[0147] Specifically, the non-contact early warning method, robotic arm collision avoidance method and system provided in this application have the following advantages: Response time advantage. Compared to related technology one, this application advances the detection node from when a collision occurs to before the collision, allowing the robot and / or robotic arm to shorten the braking distance from 50mm to 15mm at a movement speed of 0.1m / s. This significantly reduces the probability of physical collisions with the robot, fundamentally solving the safety hazard of the "collision first, then processing" approach in contact-based solutions.

[0148] Environmental adaptability advantage. Compared to related technology 2, this application significantly improves the detection stability under different lighting conditions and object materials (including transparent glass, mirrored metal, etc.). Experimental data shows that the solution provided by this application has a detection success rate close to that in strong light / dark light environments and normal lighting environments, while the performance fluctuates by 30-60% in similar scenarios. Furthermore, it is unaffected by visual obstruction and can detect approaching obstacles within blind spots.

[0149] Cost and reliability advantages. Compared with related technology three, the hardware cost of the capacitive sensor involved in this application is far lower than that of the multi-view camera system. It requires less resources for sensor data analysis and computation, does not require GPU (Graphics Processing Unit) inference, has low system response latency, and low maintenance costs, requiring no periodic calibration. Based on the multi-dimensional feature fusion analysis of the capacitive sensor, the reliability of collision avoidance detection is effectively guaranteed.

[0150] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application, such as... Figure 4As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440. The processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions from the memory 430 to execute non-contact warning methods or robotic arm collision avoidance methods.

[0151] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0152] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the non-contact early warning method or the robotic arm anti-collision method provided by the above methods.

[0153] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the non-contact early warning method or the robotic arm collision avoidance method provided by the above methods.

[0154] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0155] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A non-contact early warning method, characterized in that, include: During the task execution process of the target robot, at least two dimensions of capacitance characteristics are determined based on the target capacitance values ​​collected by each capacitance sensor, and each capacitance sensor is installed on the target robot. For each of the capacitive sensors, if the capacitive characteristics corresponding to the capacitive sensor meet the impending contact condition, an obstacle warning is triggered, wherein the impending contact condition indicates that the target robot is about to come into contact with an obstacle.

2. The non-contact early warning method according to claim 1, characterized in that, The capacitance characteristics in at least two dimensions include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; The method for determining capacitance characteristics in at least two dimensions based on the target capacitance values ​​collected by each capacitance sensor includes: For each of the capacitive sensors, local extrema calculation is performed on the target capacitance value collected by the capacitive sensor to obtain the peak value corresponding to the capacitive sensor; By selecting a fixed time window, gradient feature statistics are performed on the target capacitance value collected by the capacitance sensor to obtain the gradient feature corresponding to the capacitance sensor. Based on the target capacitance value collected by the capacitance sensor and the target capacitance value collected by the adjacent capacitance sensors, a sign consistency judgment is performed to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor.

3. The non-contact early warning method according to claim 2, characterized in that, The step of performing a sign consistency judgment based on the target capacitance value collected by the capacitance sensor and the target capacitance values ​​collected by adjacent capacitance sensors to obtain the change trend of the adjacent capacitance sensors corresponding to the capacitance sensor includes: The target capacitance value collected by the capacitance sensor is calculated by differential value to obtain a first differential value, and the target capacitance value collected by the adjacent capacitance sensor is calculated by differential value to obtain a second differential value; Perform a sign consistency check on the first difference value and the second difference value; If the signs are the same, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensors are determined to be in the same direction. If the symbols are inconsistent, then the changing trends of the adjacent capacitive sensors corresponding to the capacitive sensor are determined to be opposite.

4. The non-contact early warning method according to claim 1, characterized in that, The capacitance characteristics in at least two dimensions include: peak value, gradient characteristics, and the variation trend of adjacent capacitance sensors; Before triggering an obstacle warning when the capacitance characteristics corresponding to each capacitance sensor meet the impending contact condition, the method further includes: The consistency ratio of the changing trends of adjacent capacitive sensors corresponding to the capacitive sensor is determined to be in the same direction; Determine whether the peak value, gradient feature, and consistency ratio corresponding to the capacitive sensor meet the upcoming contact conditions.

5. The non-contact early warning method according to claim 4, characterized in that, The peak value is the minimum value; The conditions for impending contact are that the minimum value corresponding to the capacitive sensor is less than a specified peak threshold, the gradient feature corresponding to the capacitive sensor is greater than a specified gradient feature threshold, and the consistency ratio corresponding to the capacitive sensor is greater than a specified ratio threshold. The specified peak threshold is either a general peak threshold or a dedicated peak threshold corresponding to the capacitive sensor; The specified gradient feature threshold is either a general gradient feature threshold or a dedicated gradient feature threshold corresponding to the capacitive sensor. The specified ratio threshold is either a general ratio threshold or a specific ratio threshold corresponding to the capacitance sensor.

6. The non-contact early warning method according to claim 4, characterized in that, The process of determining the dedicated peak threshold corresponding to the capacitive sensor includes: Local extremum calculations are performed on the target data corresponding to the capacitance sensor to obtain multiple candidate peak values ​​corresponding to the capacitance sensor. The target data is the historical capacitance value collected by the capacitance sensor, or the target data includes the historical capacitance value collected by the capacitance sensor and the target capacitance value. Select the n target peaks with the smallest values ​​from the plurality of candidate peaks, where n is a positive integer; Based on the n target peak values, a dedicated peak threshold corresponding to the capacitive sensor is determined.

7. The non-contact early warning method according to claim 4, characterized in that, The process of determining the dedicated gradient feature threshold corresponding to the capacitive sensor includes: Gradient feature statistics are performed on the target data corresponding to the capacitive sensor to obtain multiple candidate gradient features corresponding to the capacitive sensor. Local extremum calculations are performed on the multiple candidate gradient features to obtain multiple candidate peak values ​​corresponding to the capacitive sensor; Select the m target gradient peaks with the smallest values ​​from the plurality of candidate peaks, where m is a positive integer; Based on the m target gradient peaks, the dedicated gradient features corresponding to the capacitive sensor are determined.

8. The non-contact early warning method according to any one of claims 1-7, characterized in that, Each of the capacitive sensors is arranged in a non-uniform grid. And / or, the capacitive sensor density in the first region of the target robot surface is greater than the capacitive sensor density in the second region of the target robot surface, wherein the first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold; And / or, each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the target robot's shell.

9. A collision avoidance method for a robotic arm, characterized in that, include: During the task performed by the robotic arm of the target robot, at least two dimensions of capacitance characteristics are determined based on the target capacitance values ​​collected by each capacitance sensor, and each capacitance sensor is installed on the robotic arm. For each of the capacitive sensors, collision avoidance control is performed on the robotic arm based on the capacitive characteristics and impending contact conditions corresponding to the capacitive sensors, wherein the impending contact conditions indicate that the robotic arm is about to come into contact with an obstacle.

10. The anti-collision method for a robotic arm according to claim 9, characterized in that, The step of performing anti-collision control on the robotic arm based on the capacitance characteristics corresponding to the capacitance sensors and the impending contact conditions includes: When the capacitance characteristics corresponding to the capacitance sensors meet the conditions for imminent contact, the robotic arm is controlled to avoid obstacles.

11. The anti-collision method for a robotic arm according to claim 9, characterized in that, Before determining at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor during the task performed by the robotic arm of the target robot, the process further includes: Receive the task to be executed; The task is broken down into multiple sub-tasks; Based on the multiple sub-tasks, the robotic arm is controlled to perform the task.

12. The anti-collision method for a robotic arm according to claim 11, characterized in that, The multiple sub-tasks include object recognition task, object pose calculation task, and robotic arm motion control task; The step of controlling the robotic arm to perform the task based on the multiple sub-tasks includes: Based on the object recognition task, the target object corresponding to the task in the task scene is identified and segmented to obtain a segmented image; Based on the object pose calculation task, the coordinate information and depth information of the segmented image are processed to obtain the pose information of the target object. Based on the robotic arm motion control task, the pose information is analyzed to obtain the motion angles and / or positions of each joint on the robotic arm; Based on the motion angles and / or positions of each joint, the robotic arm is controlled to perform the task.

13. The robotic arm collision avoidance method according to any one of claims 9-12, characterized in that, The method further includes: When the capacitance characteristics corresponding to the capacitance sensor meet the conditions for imminent contact, an obstacle warning is broadcast.

14. The robotic arm collision avoidance method according to any one of claims 9-12, characterized in that, Each of the capacitive sensors is arranged in a non-uniform grid. And / or, the density of capacitive sensors in the first region of the robotic arm surface is greater than the density of capacitive sensors in the second region of the robotic arm surface, wherein the first region is a region with curvature greater than or equal to a curvature threshold, and the second region is a region with curvature less than the curvature threshold; And / or, each of the capacitive sensors is mounted on a flexible circuit board, which is embedded in the inner layer of the robotic arm housing.

15. A collision avoidance system for a robotic arm, characterized in that, include: The tactile sensing module is configured to collect the target capacitance value based on various capacitance sensors during the task performed by the robotic arm of the target robot. The main control module is configured to determine at least two dimensions of capacitance characteristics based on the target capacitance values ​​collected by each capacitance sensor, wherein each capacitance sensor is mounted on the robotic arm. For each of the capacitive sensors, collision avoidance control is performed on the robotic arm based on the capacitive characteristics and impending contact conditions corresponding to the capacitive sensors, wherein the impending contact conditions indicate that the robotic arm is about to come into contact with an obstacle.

16. The robotic arm anti-collision system according to claim 15, characterized in that, The system also includes a human-machine communication module, an instruction parsing module, an object detection module, an object pose calculation module, and a robotic arm control module; The human-machine communication module is configured to receive the task to be executed and send the task to the instruction parsing module; The instruction parsing module is configured to split the task into an object recognition task, an object pose calculation task, and a robotic arm motion control task, and then distribute the tasks. The object detection module is configured to identify and segment the target object corresponding to the task in the task scene based on the object recognition task, and obtain a segmented image; The coordinate information of the segmented image is sent to the object pose calculation module; The object pose calculation module is configured to process the coordinate information and depth information based on the object pose calculation task to obtain the pose information of the target object. The pose information is sent to the robotic arm control module; The robotic arm control module is configured to parse the pose information based on the robotic arm motion control task to obtain the motion angles and / or positions of each joint on the robotic arm; and to control the robotic arm to perform the task based on the motion angles and / or positions of each joint.

17. The robotic arm anti-collision system according to claim 16, characterized in that, The robotic arm control module is further configured to determine the state information of the robotic arm based on the capacitance characteristics corresponding to each of the capacitance sensors and the impending contact condition; and to send the state information to the human-machine communication module. The human-machine communication module is also configured to broadcast an obstacle warning when the status information indicates that contact is imminent.

18. The robotic arm anti-collision system according to claim 17, characterized in that, The robotic arm control module is further configured to determine that the state information of the robotic arm is normal when none of the capacitance features corresponding to each of the capacitance sensors meet the impending contact condition; and to determine that the state information of the robotic arm is about to make contact when there are capacitance features corresponding to each of the capacitance sensors that meet the impending contact condition.

19. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the non-contact early warning method as described in any one of claims 1 to 8, or the robotic arm anti-collision method as described in any one of claims 9 to 14.

20. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the non-contact early warning method as described in any one of claims 1 to 8, or the robotic arm collision avoidance method as described in any one of claims 9 to 14.

21. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the non-contact early warning method as described in any one of claims 1 to 8, or the robotic arm collision avoidance method as described in any one of claims 9 to 14.

Images