Safe operation control method for multifunctional double-arm robot
By constructing a multimodal environmental perception system using global safety radar, obstacle avoidance radar, and binocular vision, and combining it with a YOLO network based on the CBAM attention mechanism, the safety response problem of dual-arm robots in complex environments was solved, achieving high-precision recognition and rapid safety decision-making, thus improving operational safety and efficiency.
Patent Information
- Application Number
- CN202511623826.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-07
- Publication Date
- 2026-02-24
AI Technical Summary
Existing dual-arm robot safety solutions rely on passive protection measures, which cannot quickly respond to safety risks in complex environments, resulting in decreased recognition accuracy and affecting safety and efficiency.
A multimodal environmental perception system is constructed using all-domain safety radar, obstacle avoidance radar, binocular vision and sound sensors. Combined with point cloud data processing and a YOLO network with CBAM attention mechanism, it can accurately identify people and objects and perform hierarchical safety response through an integrated control module.
It improves the recognition accuracy of static objects, dynamic objects, and human bodies in complex environments, enables rapid and safe response, and enhances safety and operational continuity.
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot control technology, specifically to a method for safe operation control of a multifunctional dual-arm robot. Background Technology
[0002] With the rapid development of industrial automation and intelligent manufacturing, dual-arm robots are increasingly widely used in complex operation scenarios such as assembly, handling, and inspection. Dual-arm robots have stronger collaborative operation capabilities and can simulate human arms to complete more delicate and complex tasks. However, the cross-functionality and complexity of their work space also bring higher safety risks.
[0003] In existing technologies, traditional dual-arm robot safety solutions mainly rely on passive protection measures such as physical fences, light grids, and emergency stop buttons. This greatly limits the robot's operational flexibility and the efficiency of human-robot collaboration, and has obvious limitations. In environments with changes in lighting, object occlusion, or complex backgrounds, the recognition accuracy will drop significantly. For irregularly shaped objects, non-standard human postures, or fast-moving targets, it is easy to make misjudgments or omissions, and it is impossible to provide safety response measures quickly, which affects the overall safety and work efficiency.
[0004] Therefore, a multifunctional dual-arm robot safety operation control method is needed to ensure the absolute safety of personnel and equipment while improving the robot's autonomous operation capability and work efficiency. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a multifunctional dual-arm robot safe operation control method, which solves the problem that existing devices cannot achieve rapid and safe response.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a multifunctional dual-arm robot safe operation control method, comprising an integrated control module, a head information acquisition module, an execution module, an AGV mobile platform, and sensing components. The integrated control module processes environmental information and generates decision commands. The execution module includes a head output module and a dual-arm robot. The AGV mobile platform has a touch strip at its bottom. The sensing components include a global safety radar, an obstacle avoidance radar, and binocular vision and sound sensors. The control method includes the following steps: S1: When the dual-arm robot is started, the system triggers a 10-second audible and visual alarm. After the alarm ends, it enters a safety status assessment. S2: Collect point cloud data of the work area through the full-domain safety radar and obstacle avoidance radar, and combine it with the preset theoretical model to identify objects in the area and determine whether there are human bodies or objects outside the model. S201: Preprocessing to remove outliers from point cloud data; S202: The height and density of the point cloud map are extracted using the projection method, and combined with the reflection intensity to form a three-channel image; S203: Based on a three-channel image training recognition model, it relies on the size, outline and point cloud density of the detected object to determine whether there is an identifiable human body or object and to distinguish them. S204: Embed the CBAM attention mechanism in the YOLO backbone network to focus on the features of human limbs and head; S3: Based on the identification results of step S2 and the distance between the human body or object and the center of the work area, execute the corresponding safety control; S301: If only a stationary object is detected, and the distance between the human body or object and the center of the work area is >5000mm, the integrated control module controls the dual-arm robot to operate normally; S302: If a human or object is detected to enter within 5000mm of the center of the work area and its movement trajectory will cross the work space, the integrated control module controls the dual-arm robot to decelerate until the human or object leaves the warning area or enters the braking area. S303: If a human or object is detected to enter within 2000mm of the center of the work area, the integrated control module controls the dual-arm robot to immediately stop working and restart after clearing the obstacle; S4: When the contact strip at the bottom of the AGV mobile platform detects a collision signal, the integrated control module immediately controls the dual-arm robot to sound an alarm and stop moving; S5: The integrated control module drives the dual-arm robot to move, and adjusts the motion parameters in real time based on the environmental recognition results of step S2.
[0007] Preferably, in step S1, the audible and visual alarm includes a high-frequency buzzer and a red LED warning light, wherein the buzzer alarm volume is not less than 85 decibels and the red LED warning light flashes at a frequency of 2-3 Hz.
[0008] Preferably, in step S2, the collected point cloud data is preprocessed, and reflective noise is injected to simulate metal or plastic objects, and translational theory is used to simulate the change in radar viewpoint caused by the movement of a robotic arm. The binocular vision generates three-dimensional coordinates and point cloud data by combining parallax with the principle of triangulation.
[0009] Preferably, the all-domain safety radar and the binocular vision are located at the top of the dual-arm robot. The head information acquisition module includes a vision acquisition unit, a voice acquisition unit, and a tactile acquisition unit. The voice acquisition unit is implemented through a sound sensor, and the tactile acquisition unit is acquired through the force sensor of the dual-arm robot arm. The obstacle avoidance radar adopts a high-precision infrared photoelectric sensor.
[0010] Preferably, the integrated control module processes the environmental data collected by the sensing components and generates decision commands through motion control algorithms, logic control algorithms, intelligent optimization algorithms, multimodal data fusion, and edge computing capabilities. Beneficial effects
[0011] This invention provides a multifunctional dual-arm robot safety operation control method with the following advantages: By integrating a global safety radar, obstacle avoidance radar, binocular vision, and sound sensors to construct a multimodal environmental perception system, and using point cloud height, density, and reflection intensity to synthesize three-channel images for model training, the accuracy of recognizing static objects, dynamic objects, and human bodies in complex environments is effectively improved. By embedding a YOLO network with a CBAM attention mechanism, the system can accurately identify key parts of a person even under occlusion. The system implements graded safety responses based on the position and intention of the person in the work area. When the person approaches the safety boundary, the robot decelerates; if it enters a high-risk area, it immediately stops operating and automatically resumes operation after the person leaves, achieving intelligent start-stop within the safety range. By injecting data augmentation techniques such as reflective noise, the model's generalization ability in real industrial environments is improved. Finally, by integrating motion control, intelligent algorithms, and edge computing into an integrated control module, real-time processing of multi-source data and rapid safety decision-making are achieved, effectively improving safety and operational continuity. Detailed Implementation
[0012] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0013] This invention provides a technical solution: a multifunctional dual-arm robot safe operation control method, comprising an integrated control module, a head information acquisition module, an execution module, an AGV mobile platform, and sensing components. The integrated control module is used to process environmental information and generate decision commands. The execution module includes a head output module and a dual-arm robot. The AGV mobile platform has a touch edge strip at its bottom. The sensing components include a global safety radar, an obstacle avoidance radar, and binocular vision and sound sensors. The control method includes the following steps: S1: When the dual-arm robot is started, the system triggers a 10-second audible and visual alarm. After the alarm ends, it enters a safety status assessment. S2: Collect point cloud data of the work area through the full-domain safety radar and obstacle avoidance radar, and combine it with the preset theoretical model to identify objects in the area and determine whether there are human bodies or objects outside the model. S201: Preprocessing to remove outliers from point cloud data; S202: The height and density of the point cloud map are extracted using the projection method, and combined with the reflection intensity to form a three-channel image; S203: Based on a three-channel image training recognition model, it relies on the size, outline and point cloud density of the detected object to determine whether there is an identifiable human body or object and to distinguish them. S204: Embed the CBAM attention mechanism in the YOLO backbone network to focus on the features of human limbs and head; S3: Based on the identification results of step S2 and the distance between the human body or object and the center of the work area, execute the corresponding safety control; S301: If only a stationary object is detected, and the distance between the human body or object and the center of the work area is >5000mm, the integrated control module controls the dual-arm robot to operate normally; S302: If a human or object is detected to enter within 5000mm of the center of the work area and its movement trajectory will cross the work space, the integrated control module controls the dual-arm robot to decelerate until the human or object leaves the warning area or enters the braking area; S303: If a human or object is detected to enter within 2000mm of the center of the work area, the integrated control module controls the dual-arm robot to stop working immediately and restart after clearing the obstacle; S4: When the contact strip at the bottom of the AGV mobile platform detects a collision signal, the integrated control module immediately controls the dual-arm robot to sound an alarm and stop moving; S5: The integrated control module drives the dual-arm robot to move, adjusting motion parameters in real time based on the environmental recognition results from step S2. In step S1, the audible and visual alarm includes a high-frequency buzzer and a red LED warning light. The buzzer alarm volume is not less than 85 decibels, and the red LED warning light flashes at a frequency of 2-3 Hz. In step S2, the collected point cloud data is preprocessed, injecting reflective noise to simulate metal or plastic objects and performing translation processing to simulate changes in radar perspective caused by the movement of the robotic arm. The binocular vision generates three-dimensional coordinates and point clouds through parallax combined with triangulation principles. Data; the global safety radar and the binocular vision are located at the top of the dual-arm robot. The head information acquisition module includes a vision acquisition unit, a voice acquisition unit, and a tactile acquisition unit. The voice acquisition unit is implemented through a sound sensor, and the tactile acquisition unit is acquired through the force sensor of the dual-arm robot arm. The obstacle avoidance radar adopts a high-precision infrared photoelectric sensor. The integrated control module processes the environmental data collected by the sensing components and generates decision commands through motion control algorithms, logic control algorithms, intelligent optimization algorithms, multimodal data fusion, and edge computing capabilities.
[0014] Those skilled in the art can connect the components in this case sequentially. The specific connection and operation sequence should refer to the working principle described below. The detailed connection methods are well-known technologies in the field. The working principle and process are mainly described below.
[0015] Example: First, pressing the start button triggers a 10-second audible and visual alarm. After the alarm ends, the system enters a safety status judgment loop. The full-domain safety radar and obstacle avoidance radar continuously scan the robot's surrounding environment and generate real-time point cloud data. Simultaneously, the binocular vision camera acquires images and generates supplementary 3D point cloud data through parallax calculation. The integrated control module preprocesses the original point cloud data, filtering out abnormal discrete points caused by dust or electronic noise. The preprocessed point cloud data is converted into a three-channel image: the first channel is the height information of the point cloud, the second channel is the density distribution of the point cloud on the horizontal plane, and the third channel is the reflection intensity of the point cloud, to facilitate subsequent neural network processing. The three-channel image is input into a pre-trained improved YOLO recognition model. Because this model embeds the CBAM attention mechanism in the backbone network, the system can accurately identify whether there are human bodies or objects outside the preset model in the working area and distinguish and locate them.
[0016] When the system detects a stationary object 5.5 meters away, and no personnel or other moving objects enter the 5000mm range, the integrated control module determines that it is in a safe state and controls the dual-arm robot to perform its work tasks at a normal speed.
[0017] When the system detects a worker approaching from the side and about to cross the workspace of the dual-arm robot, and the worker has entered the warning zone 4500mm from the center of the work area, the integrated control module immediately generates a decision command to reduce the movement speed of the dual-arm robot to 30% of its normal speed, and continuously monitors the worker's movements until the worker leaves the warning zone, at which point the dual-arm robot resumes full-speed operation.
[0018] When a worker accidentally enters the braking zone 1500mm from the center of the work area, the system instantly detects the intrusion. The integrated control module immediately issues an emergency stop command, causing all movements of the dual-arm robot to stop immediately. At the same time, the system issues an alarm signal. Only after the worker has eliminated the situation and manually restarted the process can the dual-arm robot resume normal operation.
[0019] When the bottom edge of the AGV platform collides with a low obstacle that is not detected by radar or vision during the movement of the AGV platform, the edge immediately generates a collision signal. After receiving the signal, the integrated control module immediately controls the AGV platform and the dual-arm robot to cut off power and brake, quickly realizing emergency protection.
[0020] Throughout the entire operation, the aforementioned environmental perception and safety decision-making process is continuously cyclical, achieving real-time dynamic safety assurance. The integrated control module comprehensively adopts motion control algorithms and intelligent optimization algorithms to ensure that the movement of the dual-arm robot is both smooth and efficient, and can respond to environmental changes at any time, thereby achieving a balance between safety and operational efficiency.
[0021] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0022] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for safe operation control of a multifunctional dual-arm robot, characterized in that, The system includes an integrated control module, a head information acquisition module, an execution module, an AGV mobile platform, and sensing components. The integrated control module processes environmental information and generates decision commands. The execution module includes a head output module and a dual-arm robot. The AGV mobile platform has a touch edge strip at its bottom. The sensing components include a global safety radar, an obstacle avoidance radar, and binocular vision and sound sensors. The control method includes the following steps: S1: When the dual-arm robot is started, the system triggers a 10-second audible and visual alarm. After the alarm ends, it enters a safety status assessment. S2: Collect point cloud data of the work area through the full-domain safety radar and obstacle avoidance radar, and combine it with the preset theoretical model to identify objects in the area and determine whether there are human bodies or objects outside the model. S201: Preprocessing to remove outliers from point cloud data; S202: The height and density of the point cloud map are extracted using the projection method, and combined with the reflection intensity to form a three-channel image; S203: Based on a three-channel image training recognition model, it relies on the size, outline and point cloud density of the detected object to determine whether there is an identifiable human body or object and to distinguish them. S204: Embedding the CBAM attention mechanism in the YOLO backbone network to focus on the features of human limbs and head; S3: Based on the identification results of step S2 and the distance between the human body or object and the center of the work area, execute the corresponding safety control; S301: If only a stationary object is detected, and the distance between the human body or object and the center of the work area is >5000mm, the integrated control module controls the dual-arm robot to operate normally; S302: If a human or object is detected to enter within 5000mm of the center of the work area and its movement trajectory will cross the work space, the integrated control module controls the dual-arm robot to decelerate until the human or object leaves the warning area or enters the braking area. S303: If a human or object is detected to enter within 2000mm of the center of the work area, the integrated control module controls the dual-arm robot to immediately stop working and restart after clearing the obstacle; S4: When the contact strip at the bottom of the AGV mobile platform detects a collision signal, the integrated control module immediately controls the dual-arm robot to sound an alarm and stop moving; S5: The integrated control module drives the dual-arm robot to move, and adjusts the motion parameters in real time based on the environmental recognition results of step S2.
2. The method for safe operation control of a multifunctional dual-arm robot according to claim 1, characterized in that, In step S1, the audible and visual alarm includes a high-frequency buzzer and a red LED warning light. The buzzer alarm volume is not less than 85 decibels, and the red LED warning light flashes at a frequency of 2-3 Hz.
3. The method for safe operation control of a multifunctional dual-arm robot according to claim 1, characterized in that, In step S2, the collected point cloud data is preprocessed by injecting reflective noise to simulate metal or plastic objects and translation processing to simulate the change in radar viewpoint caused by the movement of a robotic arm. The binocular vision generates three-dimensional coordinates and point cloud data by combining parallax with the principle of triangulation.
4. The method for safe operation control of a multifunctional dual-arm robot according to claim 1, characterized in that, The all-domain safety radar and the binocular vision are located at the top of the dual-arm robot. The head information acquisition module includes a vision acquisition unit, a voice acquisition unit, and a tactile acquisition unit. The voice acquisition unit is implemented through a sound sensor, and the tactile acquisition unit is acquired through the force sensor of the dual-arm robot arm. The obstacle avoidance radar adopts a high-precision infrared photoelectric sensor.
5. The method for safe operation control of a multifunctional dual-arm robot according to claim 1, characterized in that, The integrated control module processes the environmental data collected by the sensing components and generates decision commands through motion control algorithms, logic control algorithms, intelligent optimization algorithms, multimodal data fusion, and edge computing capabilities.