Dual-arm compound robot based on automatic task planning

Abstract

The application discloses a double-arm composite robot based on automatic task planning and relates to the technical field of service robots, which comprises a double-layer structure support table, a robot moving mechanism, double mechanical arms for performing linen grabbing and arranging tasks and an external perception component for environment detection. A nine-axis gyroscope module for collecting robot posture data and an industrial control host are arranged in the support table, and the industrial control host is electrically connected with an automatic task planning module. The automatic task planning module comprises a task priority scheduling module, a task identification module, an execution parameter analysis module and an operation real-time verification submodule. The priority scheduling module fuses the task emergency value, the environment complexity value and the like by weighting to obtain a comprehensive priority to lock the current task. The task identification module splits the task into a standardized operation sequence by a tree decomposition method. The execution parameter analysis module generates motion, force control and double-arm cooperation parameters. The operation real-time verification module verifies the operation by multi-sensor fusion and triggers abnormal processing.

Description

Technical Field

[0001] This invention relates to the field of service robot technology, and in particular to a dual-arm composite robot based on automatic task planning. Background Technology

[0002] In the field of linen robot technology, existing automated linen robots have significant shortcomings in environmental perception and autonomous movement, making it difficult to meet the needs of precise and stable automated linen handling in scenarios such as hotels.

[0003] Existing robots suffer from weak environmental perception capabilities, often relying on single-type sensing components such as a single LiDAR or a single vision camera. This fails to achieve omnidirectional detection of the surrounding environment and is prone to misjudging the location of obstacles, people, or linens due to blind spots. Furthermore, the data collaboration efficiency between sensing components and control units, such as industrial control systems, is low. The lack of real-time integration and processing capabilities for multi-source sensing data prevents fine-tuning of position through visual image comparison. Consequently, after arriving at a designated room or linen handling area, the robot struggles to accurately calibrate its position, impacting the efficiency of subsequent linen grasping and sorting tasks. Moreover, the absence of adaptive posture adjustment components prevents real-time correction of posture deviations caused by uneven ground or load changes during movement. This leads to swaying and deviations during movement, affecting overall operational stability and potentially indirectly reducing the accuracy of subsequent linen handling. These shortcomings fail to meet the core requirements of automated linen handling tasks for precise navigation and stable movement. Therefore, these problems need to be addressed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a dual-arm composite robot based on automatic task planning.

[0005] To solve the above-mentioned technical problems, the present invention provides a dual-arm composite robot based on automatic task planning, comprising: a support platform with a double-layer structure, a robot movement mechanism, a first manipulator and a second manipulator for performing linen grasping and sorting tasks, and an external sensing component for environmental detection.

[0006] A nine-axis gyroscope module for collecting robot posture data is fixedly installed on the top surface inside the support platform; an industrial control host is fixedly installed inside the support platform, and the industrial control host is electrically connected to an automatic task planning module.

[0007] The automatic task planning module includes a task priority scheduling module, a task identification module, an execution parameter parsing module, and an operation real-time verification module;

[0008] The priority scheduling module is used to perform priority analysis on the tasks to be executed in the task list, so as to obtain the comprehensive priority value of the tasks to be executed, select the task with the largest comprehensive priority value as the current task to be executed, and lock it; wherein the priority analysis includes:

[0009] The task urgency value is calculated based on the task setting time, the environmental complexity value is calculated based on the environmental point cloud data and visual image data acquired by the external sensing components, the task distance value is calculated based on the robot's current position and the task position, and the robot's attitude stability impact value is calculated based on the real-time attitude data collected by the nine-axis gyroscope module. The task urgency, environmental complexity coefficient, task distance value and attitude stability impact value are weighted and fused to obtain the comprehensive priority value of each task to be executed.

[0010] The task identification module is used to identify the type of the currently executed task and break it down into a standardized operation sequence using a tree-like decomposition method.

[0011] The execution parameter parsing module is used to generate motion control parameters and force control parameters corresponding to the operation based on the operation requirements and environmental data.

[0012] The real-time operation verification module is used to verify whether any operation in the standardized operation sequence is completed normally through multi-sensor fusion and to trigger an exception handling mechanism.

[0013] Preferably, the priority scheduling module performs priority analysis on the tasks to be executed in the task list as follows:

[0014] Retrieve the configured task queue, identify the latest deadline for any pending task in the task queue, read the system alarm clock to obtain the current time, and subtract the current time from the latest deadline to obtain the remaining task duration ST1; identify the duration between the publication and deadline of the pending task and record it as the task cycle duration ST2; calculate the task urgency value ST by combining the remaining task duration and the task cycle duration, with the formula ST = 1 - ST1 / ST2; if ST ≥ 1, it indicates that the task has timed out;

[0015] The system identifies the task area corresponding to any task to be executed, obtains the actual number of obstacles within the task area, and presets the maximum number of obstacles per unit area based on the size of the task area. The actual number of obstacles is divided by the preset maximum number of obstacles per unit area to obtain the obstacle quantity value. The system identifies the volume of all obstacles within the task area and calculates the sum of the volumes as the obstacle volume value. The system presets the spatial volume of the task area based on the size of the task area, and divides the obstacle volume value by the preset spatial volume of the task area to obtain the obstacle space density value. The system then obtains a visual image of the linen in the task area, identifies the actual texture entropy of the linen visual image, and, based on the preset upper limit of the texture entropy of the messy linen image in the task area, divides the actual texture entropy by the upper limit of the texture entropy of the messy linen image to obtain the obstacle messiness value. The obstacle quantity value, obstacle space density value, and obstacle messiness value are weighted by preset weights to obtain the environmental complexity value HJ of the task to be executed.

[0016] Obtain the robot's current position, identify the target position of the task to be performed, calculate the three-dimensional Euclidean distance d1 between the current position and the target position, preset the maximum movement distance d2 that the robot can cover in a single task, and calculate the task-robot distance value JG based on the three-dimensional Euclidean distance and the maximum movement distance, with the formula JG=1-d1 / d2; if d1≤0, it means that the robot has reached the target position, and JG=1 at this time;

[0017] The robot's real-time pitch angle, real-time roll angle, and standard attitude angle for stable movement are then collected by the nine-axis gyroscope module. The absolute values ​​of the differences between the real-time pitch angle, real-time roll angle, and standard attitude angle are used to obtain the pitch angle deviation and roll angle deviation, respectively. The pitch angle deviation and roll angle deviation are then weighted with preset weights to obtain the current attitude value. The attitude value at any moment within the preset deviation time domain is obtained, and the attitude mean and attitude standard deviation are calculated using the mean and standard deviation formulas. The attitude value at the current moment, the attitude mean, and the attitude standard deviation are then weighted with preset weights to obtain the attitude stability impact value.

[0018] The overall priority value is obtained by weighting the mission urgency value, environmental complexity value, mission distance value, and attitude stability impact value.

[0019] Preferably, the specific implementation of the task identification and operation decomposition module includes:

[0020] Based on the semantic parsing model, the type of the currently executing task is matched, and the corresponding preset template is invoked;

[0021] It is split using a tree-based analysis method:

[0022] Taking the currently executing task as the root node, it is decomposed into three levels: stage layer, action layer, and unit layer. The stage layer is divided into preparation operation, core operation, and closing operation. The preparation operation includes robotic arm positioning and sensor calibration. The core operation includes linen positioning, grasping, sorting, and placement. The closing operation includes robotic arm reset and data upload. The action layer is used to further break down the core operation into executable actions. The unit layer breaks down each action into the smallest operation unit with a single objective and a time consumption of less than the set action duration.

[0023] Each operation unit is assigned a unique ID, which indicates the execution subject, associated sensors, and target accuracy requirements. The execution subject can be the first robotic arm, the second robotic arm, or a combination of both robotic arms.

[0024] Preferably, the parameter generation logic of the execution parameter parsing module is as follows:

[0025] Motion control parameters: Based on the DH parameter model, inverse kinematics is solved, and combined with the trajectory data output by the path dynamic cost adjustment module, joint angular displacement, motion velocity and acceleration are generated;

[0026] Force control parameters: generated using a PSO-optimized PID algorithm, with gripping force parameters adapted to the fabric material and assembly or placement force thresholds adapted to the target pose accuracy.

[0027] Dual-arm coordination parameters: The timing of actions is optimized using the DDPG algorithm, and the time difference of parallel operations is constrained to ≤200ms, and the connection gap of serial operations is constrained to ≤500ms.

[0028] Preferably, the verification logic of the real-time operation verification module includes:

[0029] The robot arm's pose, end-effector contact force, operation time, and linen visual image data are acquired in real time through external sensing components, and joint angle deviation, sensor calibration error, grasping pose deviation, placement pose deviation, linen wrinkle rate, and operation time deviation are calculated.

[0030] During the preparation phase, the joint angle deviation after the robotic arm returns to its original position must be ≤ ±0.1 rad and the sensor calibration error must be ≤ ±1%.

[0031] During the core operation phase, the following conditions must be met: the linen grasping posture deviation < 0.5mm, the linen wrinkle rate ≤ 5% when folded, the placement posture deviation ≤ ± 5mm, and the force feedback value within the preset range.

[0032] During the final operation phase, the robotic arm must be reset to its initial position and the task data upload must be 100% complete.

[0033] During anomaly handling, the corresponding deviation value is selected according to the operation type. The preset operation deviation threshold is divided into two levels. When the deviation value is less than the first threshold, the parameter is fine-tuned and retried. When the deviation value is greater than or equal to the first threshold and less than the second threshold, the path is replanned. When the deviation value is greater than or equal to the second threshold, an emergency shutdown is initiated and an audible and visual alarm is triggered. At the same time, the abnormal data is uploaded to the cloud.

[0034] Preferably, the front end of the support platform is equipped with a mounting platform with a forward tilt angle. A drive groove is formed in the middle of the top surface of the mounting platform. A servo motor is installed in the drive groove. The output shaft of the servo motor passes through the top surface of the mounting platform and is fixed to a fixed platform. A second lidar is installed in the fixed platform. The servo motor drives the second lidar to perform rotational scanning to obtain scanning information. Based on the scanning information, a three-dimensional point cloud map of the environment is generated.

[0035] Preferably, the mounting platform has a detection cavity, and the external sensing component is installed inside the detection cavity. The external sensing component includes a first positioning plate horizontally installed above the detection cavity and a second positioning plate below the cavity. An RGBD depth camera is longitudinally installed at the rear end of the first positioning plate. An observation slot for the RGBD depth camera is provided at the rear end of the mounting platform, and a first lidar is installed at the center of the front end of the mounting platform. A first vision camera and a second vision camera are horizontally fixedly installed on both sides of the top surface of the second positioning plate, and the first vision camera and the second vision camera face opposite directions. An operation screen is fixedly installed on the inclined surface at the front end of the mounting platform by bolts.

[0036] Preferably, multiple equidistant batteries are longitudinally installed along the length of the support platform at the center of the bottom surface, with adjacent batteries connected in series via copper busbars; aluminum alloy heat sinks are attached to the front and rear faces of the batteries using thermally conductive adhesive; a charging slot is provided in the center of the bottom surface of the support platform, and a wireless charging coil is installed between the charging slots; a horizontally facing charging platform is longitudinally installed at the rear end of the bottom surface of the support platform; a transverse sliding groove is provided above the rear end face of the support platform, and an L-shaped placement plate is slidably connected within the sliding groove; self-locking electric push cylinders are installed on both sides of the top surface of the support platform, and the output shaft of the electric push cylinder is fixedly connected to the placement plate.

[0037] Preferably, the placement plate has horizontally extending grooves on both sides, and an extending plate is slidably connected in the extending groove. Multiple springs are fixedly connected at equal intervals inside the extending groove, and the other end of each spring is fixedly connected to the extending plate. The extending plate has rounded corners at its front and rear ends. The first and second robotic arms are both equipped with a placement platform. A detection box is installed on the top surface of the placement platform, and a third vision camera is installed inside the detection box.

[0038] An automated task execution method based on the above-mentioned dual-arm composite robot includes the following steps:

[0039] S1, Task reception and priority filtering: The industrial control host receives the queue of tasks to be executed, and the task priority scheduling module calculates the comprehensive priority value of each task and locks the task with the largest comprehensive priority value as the current task to be executed.

[0040] S2, Task Identification and Operation Decomposition: The task identification and operation decomposition module matches the type of the currently executed task, generates a standardized operation sequence through a tree-like decomposition method, and simultaneously acquires environmental data collected by external sensing components;

[0041] S3, Operation Parameter Analysis: The execution parameter analysis module generates motion control parameters, force control parameters, and dual-arm coordination parameters based on the operation requirements and environmental data corresponding to the current task being executed;

[0042] S4, Operation Execution and Real-time Verification: The execution module drives the dual arms and moving mechanism to perform operations according to the control parameters. The real-time operation verification module verifies the operation results through multi-sensor fusion. If the result is normal, the next operation is executed. If the result is abnormal, the exception handling mechanism is triggered.

[0043] S5, Task Completion and Data Recording: After all operations are completed, the robotic arm resets, the data storage module records all task information, including the overall priority value, operation sequence and verification results, and updates the status of the currently executing task to completed.

[0044] Compared with related technologies, the dual-arm composite robot based on automatic task planning provided by this invention has the following beneficial effects:

[0045] 1. This invention utilizes an RGBD depth camera, dual LiDARs, a first vision camera, and a second vision camera in conjunction with an industrial control host to process data in real time. The second LiDAR, driven by a servo motor, scans 360° to form a cloud map. Combined with image comparison from the dual vision cameras, it achieves omnidirectional environmental detection and position fine-tuning. A stepper motor drives the active wheel in conjunction with the driven wheel, and a nine-axis gyroscope module adaptively adjusts the robot's posture to ensure smooth movement. Simultaneously, when external personnel or objects enter the robot's trajectory, the LiDAR feedback data triggers an audible and visual alarm and an emergency stop for the robot. Its advantages include high navigation accuracy, stable movement, and strong safety protection, meeting the needs of robots automatically traveling to designated locations in scenarios such as hotels.

[0046] 2. This invention achieves efficient linen sorting and flexible carrying through the cooperation of dual robotic arms, a load-bearing adjustment structure, and a visual positioning component. The first and second robotic arms work in conjunction with a third vision camera to initially locate the linen position using the first and second vision cameras, and then the third vision camera identifies the edges and corners of the linen for precise grasping and sorting. An electric push cylinder drives the placement plate to extend, and the telescopic plate unfolds under the action of a spring to increase the load-bearing space. The rounded corners of the telescopic plate prevent bumps and knocks. The advantages are high linen sorting accuracy and flexible adjustment of the load-bearing space, which can adapt to the sorting and transportation needs of different quantities of linens and improve linen processing efficiency.

[0047] 3. This invention integrates multi-source external sensing components, a nine-axis gyroscope module, and an automatic task planning module that includes task priority scheduling, task identification, execution parameter analysis, and real-time operation verification. Combined with dual robotic arms and a moving mechanism, it can eliminate sensing blind spots, prioritize responses to highly urgent tasks, detect operational anomalies in real time, and ensure smooth movement, effectively improving the accuracy and reliability of linen handling. Attached Figure Description

[0048] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0049] Figure 1 This is a schematic diagram of the overall structure proposed in this invention;

[0050] Figure 2 This is a half-sectional schematic diagram of the overall structure proposed in this invention;

[0051] Figure 3 This is a partial sectional view of the overall structure proposed in this invention;

[0052] Figure 4 The present invention proposes Figure 2 Enlarged diagram of part A in the middle;

[0053] Figure 5 The present invention proposes Figure 3 Enlarged diagram of section B;

[0054] Figure 6 The present invention proposes Figure 3 Enlarged diagram of section C;

[0055] Figure 7 This is a block diagram illustrating the principle of the automatic task planning module proposed in this invention.

[0056] The components in the diagram are numbered as follows: 1. Support platform; 2. Driven wheel; 3. Stepper motor; 4. Drive wheel; 5. First robotic arm; 6. Second robotic arm; 7. Mounting platform; 8. First LiDAR; 9. Operation screen; 10. Second positioning plate; 11. First positioning plate; 12. First vision camera; 13. Second vision camera; 14. Industrial control host; 15. RGBD depth camera; 16. Servo motor; 17. Fixing platform; 18. Second LiDAR; 19. Charging slot; 20. Wireless charging coil; 21. Battery; 22. Heat sink; 23. Nine-axis gyroscope module; 24. Charging platform; 25. Electric push cylinder; 26. Placement plate; 27. Telescopic plate; 28. Telescopic slot; 29. ​​Spring; 30. Placement platform; 31. Third vision camera. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0058] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “group,” “class,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0059] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0060] Please refer to the following: Figures 1-7 A dual-arm composite robot based on automatic task planning includes a support platform 1 with a double-layer structure, a robot movement mechanism, a first robotic arm 5 and a second robotic arm 6 for performing linen grasping and sorting tasks, and an external sensing component for environmental detection;

[0061] The robot's moving mechanism includes: a support platform 1 for mounting stepper motors 3 via external welding or snap-fit ​​components, and a limit platform for limiting driven wheels 2 connected by rotating rods; driven wheels 2 are mounted on both sides of the front end of the support platform 1, with a rotating rod fixed between the two driven wheels 2, allowing the driven wheels 2 to move the entire support platform 1 in conjunction with the drive wheel 4; a limit platform is installed on the inner bottom surface of the support platform 1 to limit the rotation of the rotating rod; stepper motors 3 are mounted on both sides of the rear end of the inner bottom surface of the support platform 1, allowing the stepper motors 3 to connect to the drive wheel 4 via a connection process and drive the drive wheel 4 to rotate; the output shaft of the stepper motor 3 is fixedly connected to the drive wheel 4, allowing the drive wheel 4 to drive the support platform 1 to move;

[0062] A nine-axis gyroscope module 23 for collecting robot posture data is fixedly installed on the top surface of the support platform 1. The nine-axis gyroscope module 23 facilitates the detection of the posture of the entire product and, in conjunction with the program inside the industrial control host 14, adaptively adjusts the product posture. The model of the nine-axis gyroscope module 23 is BWT901BLECL. The industrial control host 14 is fixedly installed inside the support platform 1 and is electrically connected to an automatic task planning module.

[0063] The automatic task planning module includes a task priority scheduling module, a task identification module, an execution parameter parsing module, and an operation real-time verification module;

[0064] The priority scheduling module is used to perform priority analysis on the tasks to be executed in the task list, to obtain a comprehensive priority value for the tasks to be executed, and to select the task with the highest comprehensive priority value as the current task to be executed and lock it; the priority analysis includes:

[0065] The task urgency value is calculated based on the task setting time, the environmental complexity value is calculated based on the environmental point cloud data and visual image data obtained from the external sensing components, the task distance value is calculated based on the robot's current position and the task position, and the robot's attitude stability impact value is calculated based on the real-time attitude data collected by the nine-axis gyroscope module 23. The task urgency, environmental complexity coefficient, task distance value and attitude stability impact value are weighted and fused to obtain the comprehensive priority value of each task to be executed.

[0066] The task identification module is used to identify the type of the currently executing task and break it down into a standardized sequence of operations using a tree-like decomposition method;

[0067] The execution parameter parsing module is used to generate motion control parameters and force control parameters corresponding to the operation based on the operation requirements and environmental data;

[0068] The real-time operation verification module is used to verify whether any operation in the standardized operation sequence is completed normally through multi-sensor fusion and to trigger the exception handling mechanism.

[0069] Specifically, the priority scheduling module performs priority analysis on the tasks to be executed in the task list as follows:

[0070] Retrieve the configured task queue, identify the latest deadline for any pending task in the task queue, read the system alarm clock to obtain the current time, and subtract the current time from the latest deadline to obtain the remaining task duration ST1; identify the duration between the publication and deadline of the pending task and record it as the task cycle duration ST2; calculate the task urgency value ST by combining the remaining task duration and the task cycle duration, with the formula ST = 1 - ST1 / ST2; if ST ≥ 1, it indicates that the task has timed out;

[0071] The system identifies the task area corresponding to any task to be executed, obtains the actual number of obstacles within the task area, and presets the maximum number of obstacles per unit area based on the size of the task area. The actual number of obstacles is divided by the preset maximum number of obstacles per unit area to obtain the obstacle quantity value. The system identifies the volume of all obstacles within the task area and calculates the sum of the volumes as the obstacle volume value. The system presets the spatial volume of the task area based on the size of the task area, and divides the obstacle volume value by the preset spatial volume of the task area to obtain the obstacle space density value. The system then obtains a visual image of the linen in the task area, identifies the actual texture entropy of the linen visual image, and, based on the preset upper limit of the texture entropy of the messy linen image in the task area, divides the actual texture entropy by the upper limit of the texture entropy of the messy linen image to obtain the obstacle messiness value. The obstacle quantity value, obstacle space density value, and obstacle messiness value are weighted by preset weights to obtain the environmental complexity value HJ of the task to be executed.

[0072] Obtain the robot's current position, identify the target position of the task to be performed, calculate the three-dimensional Euclidean distance d1 between the current position and the target position, preset the maximum movement distance d2 that the robot can cover in a single task, and calculate the task-robot distance value JG based on the three-dimensional Euclidean distance and the maximum movement distance, with the formula JG=1-d1 / d2; if d1≤0, it means that the robot has reached the target position, and JG=1 at this time;

[0073] The robot's real-time pitch angle, real-time roll angle, and standard attitude angle for stable movement are then collected by the nine-axis gyroscope module 23. The absolute values ​​of the differences between the real-time pitch angle, real-time roll angle, and standard attitude angle are used to obtain the pitch angle deviation and roll angle deviation, respectively. The pitch angle deviation and roll angle deviation are then weighted with their preset weights to obtain the current attitude value. The attitude value at any moment within the preset deviation time domain is obtained, and the attitude mean and attitude standard deviation are calculated using the mean and standard deviation formulas. The attitude value at the current moment, the attitude mean, and the attitude standard deviation are then weighted with preset weights to obtain the attitude stability influence value ZT.

[0074] The overall priority value HZ is calculated by weighting the mission urgency value ST, environmental complexity value HJ, mission distance value JG, and attitude stability impact value ZT. The formula is HZ=z1×ST+z2×HJ+z3×JG+z4×ZT, where z1, z2, z3, and z4 represent the weights corresponding to the mission urgency value, environmental complexity value, mission distance value, and attitude stability impact value, respectively.

[0075] Specifically, the implementation of the task identification and operation decomposition module includes:

[0076] Based on the semantic parsing model, the type of the current task is matched, such as linen grabbing, linen sorting or linen transportation, and the corresponding preset template is called.

[0077] It is split using a tree-based analysis method:

[0078] Taking the currently executing task as the root node, it is decomposed into three levels: stage layer, action layer, and unit layer. The stage layer is divided into preparation operation, core operation, and closing operation. The preparation operation includes robotic arm positioning and sensor calibration. The core operation includes linen positioning, grasping, sorting, and placement. The closing operation includes robotic arm reset and data upload. The action layer is used to further break down the core operation into executable actions. The unit layer breaks down each action into the smallest operation unit with a single objective and a time consumption of less than the set action duration.

[0079] Each operation unit is assigned a unique ID, which indicates the execution subject, associated sensors, and target accuracy requirements. The execution subject includes the first robotic arm 5, the second robotic arm 6, or a combination of both robotic arms.

[0080] Specifically, the parameter generation logic of the execution parameter parsing module is as follows:

[0081] Motion control parameters: Inverse kinematics are solved based on the DH parameter model. For the i-th joint of the robotic arm, the system of equations is solved. Obtain joint angular displacement Combined with the trajectory data output by the path dynamic cost adjustment module, the motion speed is generated. and acceleration ;in , This represents the constraint threshold corresponding to the kinematics of the robotic arm.

[0082] Force control parameters: generated using a PSO-optimized PID algorithm, controlling the output. In the formula, Force control deviation, i.e., the difference between the target force and the actual force. , , The PID parameters are determined by the PSO algorithm in the search space. Internal optimization yielded, among which These represent the value range of the corresponding PID parameters, the gripping force parameter adapts to the fabric material, and the assembly or placement force threshold adapts to the target pose accuracy.

[0083] Dual-arm coordination parameters: The motion timing is optimized using the DDPG algorithm, with the objective function being... And constrain the time difference of parallel operations The connection gap of serial operation ,in , To preset the collaborative timing threshold, k represents the index of the parallel operation, used to traverse all actions executed in parallel, m represents the total number of parallel operations, i.e., the total number of actions executed simultaneously by both arms, l represents the index of the serial operation, used to traverse all actions executed serially, and n represents the total number of serial operations, i.e., the total number of actions executed sequentially by both arms. The purpose of this formula is to optimize the action timing of the two arms collaboration by minimizing the sum of the time difference of the parallel operations and the connection gap of the serial operations, so as to ensure the synchronization and smoothness of the two-arm operations.

[0084] Specifically, the verification logic of the real-time verification module includes:

[0085] Real-time data on robotic arm pose, end-effector contact force, operation time, and visual images of linens are acquired using external sensing components. Various operational deviations are calculated based on preset standard parameters, as follows:

[0086] For the robotic arm's return and reset operations, obtain the actual angle of the i-th joint. Compared with standard attitude angle absolute value The maximum value of all joint deviations is taken as the final joint angle deviation. ;

[0087] The sensor's calibration value is then obtained and recorded. The standard calibration value corresponding to the preset sensor is then used to calculate the sensor's calibration value. Compared with standard calibration value The relative deviation between them yields the calibration error. The formula is The calibration values ​​of the sensors are calculated according to the type of sensor. The calculation logic is as follows: vision sensors are calibrated using the Zhang Zhengyou calibration method, force / torque sensors are calibrated using the standard force / torque loading calibration method, and distance measuring sensors are calibrated using the standard distance comparison calibration method (all of which are existing calibration technologies).

[0088] Based on the base coordinate system, obtain the actual grasping pose of the end effector. Grasping pose of the target The translation deviation is the three-dimensional Euclidean distance. The rotational deviation is the sum of the absolute values ​​of the Euler angle differences. Calculate the grasping position deviation based on three-dimensional Eulerian distance and rotational deviation ,in , Preset weights;

[0089] Similarly, obtain the actual placement pose of the end effector. Positioning with the target The method for calculating the grasping pose deviation is used to prevent pose deviation. ;

[0090] Segmenting the area of ​​wrinkled regions using visual images Total area of ​​linens Calculate the wrinkle rate of linens The formula is ;

[0091] Identify actual operation time Compared with standard time The relative deviation is used to calculate the operation time deviation, and the formula is as follows: ;

[0092] The verification at each stage of operation must meet the corresponding deviation threshold requirements:

[0093] When preparing for operation, it is necessary to ensure that the robotic arm returns to its original position. ;

[0094] During core operations, the force feedback value must be within a preset range when the linen is grasped. When folding When placing ;

[0095] During the final operation, the robotic arm must be reset to its initial position and the task data upload must be 100% complete.

[0096] When handling exceptions, select the corresponding deviation value based on the operation type. ,in It can be Any value; the preset operation deviation threshold is denoted as... , ,but:

[0097] when < If necessary, perform parameter fine-tuning and retry;

[0098] when When necessary, perform path replanning;

[0099] when In case of emergency, the system will shut down immediately and trigger an audible and visual alarm, while simultaneously uploading abnormal data to the cloud.

[0100] Specifically, a mounting platform 7 with a forward-tilting angle is installed at the front end of the support platform 1. A drive groove is opened in the middle of the top surface of the mounting platform 7. A servo motor 16 is installed in the drive groove. The output shaft of the servo motor 16 passes through the top surface of the mounting platform 7 and is fixed to the fixed platform 17. A second lidar 18 is installed in the fixed platform 17. The servo motor 16 drives the second lidar 18 to perform a 360° rotation scan to obtain scanning information. Based on the scanning information, a three-dimensional point cloud map of the environment is generated, and the three-dimensional point cloud map of the environment is fused with the depth data collected by the RGBD depth camera 15.

[0101] Specifically, the mounting platform 7 has a detection cavity 32, and the external sensing components are installed inside the detection cavity 32. The external sensing components include a first positioning plate 11 horizontally installed above the detection cavity 32 and a second positioning plate 10 below the cavity. An RGBD depth camera 15 is vertically installed at the rear end of the first positioning plate 11. An observation slot for the RGBD depth camera 15 is opened at the rear end of the mounting platform 7, and a first lidar 8 is installed at the center of the front end of the mounting platform 7. A first vision camera 12 and a second vision camera 13 are horizontally fixed on both sides of the top surface of the second positioning plate 10, respectively. The first vision camera 12 and the second vision camera 13 face opposite directions and respectively cover the visual detection range of the left and right sides of the support platform 1. An operation screen 9 is fixedly installed on the inclined surface at the front end of the mounting platform 7 by bolts. The operation screen 9 is electrically connected to the industrial control host 14 and is used for human-machine interaction and task status display.

[0102] Specifically, multiple equidistantly arranged batteries 21 are longitudinally installed along the length of the support platform 1 in the middle of the inner bottom surface, and adjacent batteries 21 are connected in series by copper busbars; aluminum alloy heat sinks 22 are attached to the front and rear faces of the batteries 21 with thermally conductive adhesive to assist in heat dissipation of the batteries 21; a charging slot 19 is opened in the middle of the bottom surface of the support platform 1, and wireless charging coils 20 are installed between the charging slots 19; a horizontally rearward charging platform 24 is installed longitudinally at the rear end of the bottom surface of the support platform 1, which facilitates charging of the batteries 21 in conjunction with an external charging base; a horizontal sliding groove is opened above the rear end face of the support platform 1, and an L-shaped placement plate 26 is slidably connected in the sliding groove; and self-locking electric push cylinders 25 are installed on both sides of the inner top surface of the support platform 1, with the output shaft of the electric push cylinder 25 fixedly connected to the placement plate 26, which facilitates the provision of driving force for the placement plate 26 in the sliding groove.

[0103] Specifically, both sides of the placement plate 26 are provided with horizontal telescopic grooves 28, and telescopic plates 27 are slidably connected in the telescopic grooves 28. The telescopic plates 27 facilitate the expansion space of the placement plate 26. Multiple springs 29 are fixedly connected at equal intervals inside the telescopic grooves 28. The other end of the springs 29 is fixedly connected to the telescopic plates 27, and the springs 29 facilitate the provision of driving force for the expansion of the telescopic plates 27. The front and rear ends of the telescopic plates 27 are rounded. The first robotic arm 5 and the second robotic arm 6 are both equipped with placement platforms 30. The top surface of the placement platform 30 is equipped with a detection box, and a third vision camera 31 is installed in the detection box. The third vision camera 31 facilitates the first robotic arm 5 and the second robotic arm 6 to provide a precise grasping position when arranging linens. The model of the third vision camera 31 is MV-CA013-A0GM.

[0104] Based on the above, this invention proposes an automated task execution method for a dual-arm composite robot, comprising the following steps:

[0105] S1, Task reception and priority filtering: The industrial control host 14 receives the queue of tasks to be executed, and the task priority scheduling module calculates the comprehensive priority value of each task and locks the task with the largest comprehensive priority value as the current task to be executed.

[0106] S2, Task Identification and Operation Decomposition: The task identification and operation decomposition module matches the type of the currently executed task, generates a standardized operation sequence through a tree-like decomposition method, and simultaneously acquires environmental data collected by external sensing components;

[0107] S3, Operation Parameter Analysis: The execution parameter analysis module generates motion control parameters, force control parameters, and dual-arm coordination parameters based on the operation requirements and environmental data corresponding to the current task being executed;

[0108] S4, Operation Execution and Real-time Verification: The execution module drives the dual arms and moving mechanism to perform operations according to the control parameters. The real-time operation verification module verifies the operation results through multi-sensor fusion. If the result is normal, the next operation is executed. If the result is abnormal, the exception handling mechanism is triggered.

[0109] S5, Task Completion and Data Recording: After all operations are completed, the robotic arm resets, the data storage module records all task information, including the overall priority value, operation sequence and verification results, and updates the status of the currently executing task to completed.

[0110] The working principle of this invention revolves around the entire process of standby initialization, task reception and scheduling, mobile positioning, linen processing, and return charging. It combines hardware component collaboration with closed-loop control of the automatic task planning module to achieve automated and precise execution of linen grabbing and sorting tasks. The specific process includes the following:

[0111] Standby and initialization detection phase

[0112] Standby state preparation: The entire robot is placed on the external supporting charging base, and the battery 21 is charged by the charging platform 24 or the wireless charging coil 20; at this time, the aluminum alloy heat sink 22 at the front and rear ends of the battery 21 continues to dissipate heat to ensure the stable operation of the energy system.

[0113] Component Function Testing: The user initiates a component testing command via the operation screen 9 on the inclined surface of the front of the mounting platform 7. The industrial control host 14 drives the external sensing components, including the RGBD depth camera 15, the first LiDAR 8, the second LiDAR 18, the first vision camera 12, the second vision camera 13, the third vision camera 31, and the nine-axis gyroscope module 23, to complete the self-test.

[0114] External sensing components: RGBD depth camera 15 acquires images of the robot's back end through the observation slot, first LiDAR 8 scans the front-end environment, second LiDAR 18 rotates under the drive of servo motor 16 to test the 360° scanning function, and three vision cameras acquire images to verify the image clarity.

[0115] Nine-axis gyroscope module 23: Collects the robot's initial posture data and compares it with the preset stable movement standard posture angle to verify whether the posture detection accuracy meets the requirements;

[0116] Standby data transmission: After the self-test is passed, the robot enters the standby state. The first vision camera 12 and the second vision camera 13 on the second positioning plate 10 remain running, collecting images of the surrounding environment in real time. After being processed by the industrial control host 14, the images are transmitted to the cloud to provide data support for subsequent remote command reception and environmental pre-judgment.

[0117] Task reception and priority scheduling phase

[0118] Command reception and parsing: When applied to scenarios such as hotels, users send linen processing commands to the cloud via external mobile devices. The cloud forwards the commands to the industrial control host 14. The industrial control host 14 parses the commands and extracts key task information, including task type, target location, deadline, and linen type.

[0119] Task priority calculation: The task priority scheduling module in the automatic task planning module is activated, and calculates the comprehensive priority value of the tasks to be executed according to preset logic:

[0120] Calculate the task urgency value: Calculate the urgency value by the ratio of the remaining task time (latest deadline minus the current time) to the task cycle duration (total duration from task release to deadline). Subtract the ratio of the remaining task time to the task cycle duration from 1 to obtain the task urgency value. If the result is greater than or equal to 1, it means the task has timed out.

[0121] Calculate the environmental complexity value: Combine the environmental point cloud and visual data collected by the external sensing components to calculate the obstacle quantity value (the actual number of obstacles in the task area divided by the preset maximum number of obstacles per unit area), obstacle density value (the total volume of obstacles in the task area divided by the preset spatial volume of the task area), and obstacle clutter value (the actual texture entropy of the linen visual image divided by the preset upper limit of the texture entropy of the cluttered linen image). Then, multiply these three values ​​by preset weights and add them together to obtain the environmental complexity value.

[0122] Calculate the task distance value: By combining the three-dimensional Euclidean distance between the robot's current position and the task target position with the preset maximum movement distance of the robot in a single task, subtract the proportion of the three-dimensional Euclidean distance to the maximum movement distance from 1 to obtain the task distance value; if the three-dimensional Euclidean distance is less than or equal to 0, it means that the robot has reached the target position, and the task distance value is set to 1.

[0123] Calculate the attitude stability impact value: Based on the real-time pitch angle and real-time roll angle collected by the nine-axis gyroscope module 23, compare them with the standard attitude angles for smooth robot movement to obtain the pitch angle deviation and roll angle deviation. Multiply the two by preset weights and add them to obtain the current attitude value. Then, obtain the attitude values ​​at each time in the preset deviation time domain, calculate the mean and standard deviation of the attitude values ​​in the time domain, and finally multiply the current attitude value, the attitude mean, and the attitude standard deviation by preset weights and add them to obtain the attitude stability impact value.

[0124] Calculate the overall priority value: Multiply the task urgency value, environmental complexity value, task distance value, and attitude stability impact value by their respective preset weights and then sum them to obtain the overall priority value; lock the task with the highest overall priority value as the current task to be executed.

[0125] Movement and target location positioning stage

[0126] Motion drive and attitude adjustment: The industrial control host 14 sends a start command to the stepper motor 3 at the rear end of the inner bottom surface of the support platform 1. The output shaft of the stepper motor 3 drives the drive wheel 4 to rotate. The drive wheel 4, together with the driven wheel 2 at the front end (connected by a rotating rod, and the limit table restricts the circumferential rotation of the rotating rod), drives the robot to move. During the movement, the nine-axis gyroscope module 23 collects the robot's pitch angle and roll angle in real time, compares them with the standard attitude angle and calculates the deviation. The industrial control host 14 adjusts the speed and direction of the stepper motor 3 in real time according to the deviation to ensure that the robot moves smoothly.

[0127] Environmental safety detection: During the movement, the first lidar 8 on the top surface of the mounting platform 7 scans the environment in front of the robot, and the second lidar 18 performs a 360° scan and generates a three-dimensional point cloud map of the environment as the servo motor 16 drives the fixed platform 17 to rotate. This point cloud map is fused with the depth data collected by the RGBD depth camera 15 on the first positioning plate 11, and the industrial control host 14 analyzes the distribution of obstacles in the environment to determine whether there are obstacles that affect the movement. If an obstacle is detected, the industrial control host 14 adjusts the movement path in real time to avoid the obstacle.

[0128] Target position fine-tuning: After the robot arrives at the designated task area (such as a hotel room) via the active wheel 4 and the driven wheel 2, the first vision camera 12 and the second vision camera 13 (facing opposite directions, respectively covering the left and right visual ranges of the robot) collect images of the area and compare them with the "standard room environment data" pre-stored in the industrial control host 14 to calculate the deviation between the robot's current position and the ideal working position; the industrial control host 14 drives the stepper motor 3 to fine-tune the robot's position to ensure that the working range of the first robotic arm 5 and the second robotic arm 6 can accurately cover the area where the linens are located.

[0129] Linen positioning and sorting execution phase

[0130] (a) Precise positioning of linens

[0131] Preliminary positioning: The first vision camera 12 and the second vision camera 13 acquire images of the area where the linen is located. Combined with the depth data acquired by the RGBD depth camera 15, the approximate location and distribution range of the linen are preliminarily determined, and the data is transmitted to the industrial control host 14.

[0132] Precise positioning: The industrial control host 14 drives the first robotic arm 5 and the second robotic arm 6 to move above the linen. The third vision camera 31 in the detection box on the robotic arm end plate 30 is activated to collect detailed images of the linen and identify features such as the edges and wrinkles of the linen. If the third vision camera 31 does not clearly capture the edges and corners of the linen, the industrial control host 14 drives the robotic arm to slightly flip the linen until the coordinates of the edges and corners of the linen can be accurately located.

[0133] (II) Organizing Parameter Generation and Execution

[0134] Parameter parsing: The execution parameter parsing module in the automatic task planning module is activated, generating three types of core control parameters based on linen positioning data and environmental data:

[0135] Motion control parameters: Based on the DH parameter model, inverse kinematics is solved, and combined with the trajectory data output by the path dynamic cost adjustment module, the angular displacement, motion velocity (not exceeding the maximum velocity of the robot arm kinematic constraints) and acceleration (not exceeding the maximum acceleration of the robot arm kinematic constraints) of each joint of the robot arm are generated.

[0136] Force control parameters: generated using a proportional-integral-differential algorithm optimized by particle swarm optimization, with appropriate gripping force adapted to the fabric material (such as cotton, synthetic fiber) and appropriate placement force threshold adapted to the precision requirements of the placement location (such as metal storage rack, plastic storage rack).

[0137] Dual-arm coordination parameters: The timing of actions is optimized through a deep deterministic strategy gradient algorithm to constrain the time difference between parallel operations of the two arms (such as the left arm grasping the linen and the right arm positioning the linen) to no more than 200 milliseconds, and to constrain the connection gap between serial operations of the two arms (such as grasping the linen and then folding it) to no more than 500 milliseconds, so as to avoid interference between the two arm movements.

[0138] Linen sorting execution: The industrial control host 14 converts the control parameters into drive commands and sends them to the first robotic arm 5 and the second robotic arm 6 to execute the linen sorting action.

[0139] Grasping action: The two arms grasp the edges of the linen with appropriate grasping force according to the preset coordinated timing. The third vision camera 31 verifies the grasping posture deviation in real time to ensure that the deviation is less than 0.5 mm.

[0140] The linen is unfolded and smoothed out with both arms working together. The first visual camera 12 and the second visual camera 13 detect the wrinkle rate of the linen in real time to ensure that the wrinkle rate does not exceed 5%.

[0141] Loading preparation: Simultaneously start the self-locking electric push cylinders 25 on both sides of the inner top surface of the support platform 1. The output shaft of the electric push cylinder 25 pushes the L-shaped placement plate 26 to extend along the transverse sliding groove. After the placement plate 26 is fully extended from the support platform 1, the telescopic plates 27 in the telescopic grooves 28 on both sides unfold under the elastic force of the spring 29, expanding the load-bearing space (the front and rear ends of the telescopic plates 27 are designed with rounded corners to avoid collision with linens or surrounding objects).

[0142] (III) Real-time Operation Verification and Anomaly Handling

[0143] The real-time operation verification module monitors the entire linen handling process, collecting data and verifying operation results through external sensing components.

[0144] Grasping verification: if the grasping force fed back by the force sensor is within the preset range and the grasping pose deviation detected by the third vision camera 31 is less than 0.5 mm, the grasping operation is judged to be normal.

[0145] Verification: The wrinkle rate of the linen detected by the visual camera does not exceed 5%, indicating that the finishing operation is normal;

[0146] Placement verification: If the deviation of the placement position of the linen detected by the visual camera does not exceed ±5 mm, and the linen does not fall or shift, the placement operation is judged to be normal.

[0147] If an operation deviates (e.g., a grasping pose deviation of 1.2 mm), it will be handled according to preset thresholds: if the deviation is less than 1 mm, the force control or motion parameters will be fine-tuned and the operation will be retried; if the deviation is greater than or equal to 1 mm but less than 3 mm, the path replanning will be triggered and the operation will be re-executed; if the deviation is greater than or equal to 3 mm, the machine will be immediately shut down and an audible and visual alarm will be triggered. At the same time, the abnormal data (including real-time visual images) will be uploaded to the cloud and external personnel will be requested to take over.

[0148] Mission wrap-up and return recharging phase

[0149] Task completion: After the linens are sorted, the first robotic arm 5 places the sorted linens one by one onto the unfolded placement plate 26; after all the linens are placed, the industrial control host 14 drives the electric push cylinder 25 to retract the placement plate 26, and the telescopic plate 27 is compressed by the side wall of the support platform 1, which compresses the spring 29 and finally retracts into the telescopic groove 28; at the same time, the first robotic arm 5 and the second robotic arm 6 are reset to their initial positions, the task data storage module records all task information including the comprehensive priority value, operation sequence, and verification results of each operation, and updates the status of the currently executed task to "completed".

[0150] Return and charging: The industrial control host 14 combines the 360° scanning data of the second lidar 18 and the depth data of the RGBD depth camera 15 to plan the return path and drive the stepper motor 3 to drive the robot back to the external charging base. After arriving at the charging base, it first connects to the charging base through the wireless charging coil 20 in the charging slot 19 for charging. If the wireless connection fails more than 3 times, it switches to the charging station 24 for wired charging until the battery 21 is restored to the preset threshold. The robot then enters standby mode again and waits for the next task instruction.

[0151] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0152] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A dual-arm composite robot based on automatic task planning, characterized in that, It includes a support platform (1) with a double-layer structure, a robot movement mechanism, a first robotic arm (5) and a second robotic arm (6) for performing linen grasping and sorting tasks, and an external sensing component for environmental detection. A nine-axis gyroscope module (23) for collecting robot posture data is fixedly installed on the top surface of the support platform (1); an industrial control host (14) is fixedly installed inside the support platform (1), and the industrial control host (14) is electrically connected to an automatic task planning module. The automatic task planning module includes a task priority scheduling module, a task identification module, an execution parameter parsing module, and an operation real-time verification module; The priority scheduling module is used to perform priority analysis on the tasks to be executed in the task list, so as to obtain the comprehensive priority value of the tasks to be executed, select the task with the largest comprehensive priority value as the current task to be executed, and lock it. Priority analysis includes: The task urgency value is calculated based on the task setting time, the environmental complexity value is calculated based on the environmental point cloud data and visual image data obtained by the external sensing components, the task distance value is calculated based on the robot's current position and the task position, and the robot's attitude stability impact value is calculated based on the real-time attitude data collected by the nine-axis gyroscope module (23). The task urgency, environmental complexity coefficient, task distance value and attitude stability impact value are weighted and fused to obtain the comprehensive priority value of each task to be executed. Specifically, the task urgency value ST is calculated by combining the remaining task duration ST1 and the task cycle duration ST2, with the formula ST=1-ST1 / ST2; the environmental complexity value HJ of the task to be executed is obtained by weighting the obstacle quantity value, obstacle density value, and obstacle disorder value with preset weights respectively. Obtain the robot's current position, identify the target position of the task to be performed, calculate the three-dimensional Euclidean distance d1 between the current position and the target position, preset the maximum movement distance d2 that the robot can cover in a single task, and calculate the task-robot distance value JG based on the three-dimensional Euclidean distance and the maximum movement distance, with the formula JG=1-d1 / d2; if d1≤0, it means that the robot has reached the target position, and JG=1 at this time; Then, the robot's real-time pitch angle, real-time roll angle, and standard attitude angle for smooth robot movement are collected by the nine-axis gyroscope module (23). The absolute values ​​of the differences between the real-time pitch angle, real-time roll angle, and standard attitude angle are used to obtain the pitch angle deviation and roll angle deviation. The pitch angle deviation and roll angle deviation are weighted and calculated with their preset weights to obtain the current attitude value. The attitude value at any moment within the preset deviation time domain is obtained. The attitude value within the preset deviation time domain is calculated using the mean and standard deviation formulas to obtain the attitude mean and attitude standard deviation. The attitude value, attitude mean, and attitude standard deviation at the current moment are weighted and calculated with the preset weights to obtain the attitude stability influence value ZT. The overall priority value HZ is calculated by weighting the mission urgency value ST, environmental complexity value HJ, mission distance value JG, and attitude stability impact value ZT. The formula is HZ=z1×ST+z2×HJ+z3×JG+z4×ZT, where z1, z2, z3, and z4 represent the weights corresponding to the mission urgency value, environmental complexity value, mission distance value, and attitude stability impact value, respectively. The task identification module is used to identify the type of the currently executed task and break it down into a standardized operation sequence using a tree-like decomposition method. The execution parameter parsing module is used to generate motion control parameters and force control parameters corresponding to the operation based on the operation requirements and environmental data. The real-time operation verification module is used to verify whether any operation in the standardized operation sequence is completed normally through multi-sensor fusion and to trigger an exception handling mechanism.

2. The dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, The priority scheduling module performs priority analysis on the tasks to be executed in the task list as follows: Get the configured task queue, identify the latest deadline for any task to be executed in the task queue, read the system alarm clock to get the current time, and subtract the current time from the latest deadline to get the remaining task duration ST1; identify the duration between the publication and the deadline of the task to be executed and record it as the task cycle duration ST2. If ST≥1, it means the task has timed out; Identify the task area corresponding to any task to be executed, obtain the actual number of obstacles in the task area, and preset the maximum number of obstacles per unit area based on the size of the task area. Divide the actual number of obstacles by the preset maximum number of obstacles per unit area to obtain the number of obstacles. Identify the volume of all obstacles within the task area and calculate the sum of their volumes as the obstacle volume value. Preset the spatial volume of the task area based on its size, and divide the obstacle volume value by the preset spatial volume of the task area to obtain the obstacle density value. Then, acquire the visual image of the linen in the task area, identify the actual texture entropy of the linen visual image, and based on the preset upper limit of the texture entropy of the messy linen image in the task area, divide the actual texture entropy by the upper limit of the texture entropy of the messy linen image to obtain the obstacle messiness value.

3. A dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, The specific implementation of the task identification and operation decomposition module includes: Based on the semantic parsing model, the type of the currently executing task is matched, and the corresponding preset template is invoked; It is split using a tree-based analysis method: Taking the currently executing task as the root node, it is decomposed into three levels: stage layer, action layer, and unit layer. The stage layer is divided into preparation operation, core operation, and closing operation. The preparation operation includes robotic arm positioning and sensor calibration. The core operation includes linen positioning, grasping, sorting, and placement. The closing operation includes robotic arm reset and data upload. The action layer is used to further break down the core operation into executable actions. The unit layer breaks down each action into the smallest operation unit with a single objective and a time consumption of less than the set action duration. Each operation unit is assigned a unique ID, which marks the execution subject, associated sensor and target accuracy requirements of the operation. The execution subject includes the first robot (5), the second robot (6) or the collaboration of two robot arms.

4. A dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, The parameter generation logic of the execution parameter parsing module is as follows: Motion control parameters: Based on the DH parameter model, inverse kinematics is solved, and combined with the trajectory data output by the path dynamic cost adjustment module, joint angular displacement, motion velocity and acceleration are generated; Force control parameters: generated using a PSO-optimized PID algorithm, with gripping force parameters adapted to the fabric material and assembly or placement force thresholds adapted to the target pose accuracy. Dual-arm coordination parameters: The timing of actions is optimized using the DDPG algorithm, and the time difference of parallel operations is constrained to ≤200ms, and the connection gap of serial operations is constrained to ≤500ms.

5. A dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, The verification logic of the real-time operation verification module includes: The robot arm's pose, end-effector contact force, operation time, and linen visual image data are acquired in real time through external sensing components, and joint angle deviation, sensor calibration error, grasping pose deviation, placement pose deviation, linen wrinkle rate, and operation time deviation are calculated. During the preparation phase, the joint angle deviation after the robotic arm returns to its original position must be ≤ ±0.1 rad and the sensor calibration error must be ≤ ±1%. During the core operation phase, the following conditions must be met: the linen grasping posture deviation < 0.5mm, the linen wrinkle rate ≤ 5% when folded, the placement posture deviation ≤ ± 5mm, and the force feedback value within the preset range. During the final operation phase, the robotic arm must be reset to its initial position and the task data upload must be 100% complete. During anomaly handling, the corresponding deviation value is selected according to the operation type. The preset operation deviation threshold is divided into two levels. When the deviation value is less than the first threshold, the parameter is fine-tuned and retried. When the deviation value is greater than or equal to the first threshold and less than the second threshold, the path is replanned. When the deviation value is greater than or equal to the second threshold, an emergency shutdown is initiated and an audible and visual alarm is triggered. At the same time, the abnormal data is uploaded to the cloud.

6. A dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, The support platform (1) has a mounting platform (7) with a forward tilt at its front end. A drive groove is provided in the middle of the top surface of the mounting platform (7). A servo motor (16) is installed in the drive groove. The output shaft of the servo motor (16) passes through the top surface of the mounting platform (7) and is fixed to a fixed platform (17). A second laser radar (18) is installed in the fixed platform (17). The servo motor (16) drives the second laser radar (18) to perform rotational scanning to obtain scanning information. Based on the scanning information, a three-dimensional point cloud map of the environment is generated.

7. A dual-arm composite robot based on automatic task planning according to claim 6, characterized in that, The mounting platform (7) has a detection cavity (32) inside, and the external sensing component is installed inside the detection cavity (32). The external sensing component includes a first positioning plate (11) horizontally installed above the detection cavity (32) and a second positioning plate (10) below the cavity. An RGBD depth camera (15) is installed longitudinally at the rear end of the first positioning plate (11). An observation slot for the RGBD depth camera (15) is opened at the rear end of the mounting platform (7), and a first lidar (8) is installed at the middle of the front end of the mounting platform (7). A first vision camera (12) and a second vision camera (13) are horizontally fixed on both sides of the top surface of the second positioning plate (10). The first vision camera (12) and the second vision camera (13) face opposite directions. An operation screen (9) is fixedly installed on the inclined surface at the front end of the mounting platform (7) by bolts.

8. A dual-arm composite robot based on automatic task planning according to claim 1, characterized in that, Multiple equidistant batteries (21) are longitudinally installed in the middle of the bottom surface of the support platform (1) along the length of the support platform (1), and adjacent batteries (21) are connected in series by copper busbars; the front and rear surfaces of the batteries (21) are both attached with heat dissipation platforms (22) made of aluminum alloy by thermal conductive adhesive; a charging slot (19) is opened in the middle of the bottom surface of the support platform (1), and a wireless charging coil (20) is installed between the charging slots (19); a horizontal charging platform (24) is installed longitudinally at the rear end of the bottom surface of the support platform (1); a horizontal sliding groove is opened above the rear end surface of the support platform (1), and an L-shaped placement plate (26) is slidably connected in the sliding groove; and self-locking electric push cylinders (25) are installed on both sides of the top surface of the support platform (1), and the output shaft of the electric push cylinder (25) is fixedly connected to the placement plate (26).

9. A dual-arm composite robot based on automatic task planning according to claim 8, characterized in that, The placement plate (26) has horizontally opened telescopic grooves (28) on both sides. A telescopic plate (27) is slidably connected in the telescopic groove (28). Multiple springs (29) are fixedly connected at equal intervals in the inner side of the telescopic groove (28). The other end of the spring (29) is fixedly connected to the telescopic plate (27). The front and rear ends of the telescopic plate (27) are rounded. The first robotic arm (5) and the second robotic arm (6) are both equipped with a placement platform (30). A detection box is installed on the top surface of the placement platform (30). A third vision camera (31) is installed in the detection box.

10. An automated task execution method based on the dual-arm composite robot according to any one of claims 1-9, characterized in that, Includes the following steps: S1, Task reception and priority filtering: The industrial control host (14) receives the queue of tasks to be executed, and the task priority scheduling module calculates the comprehensive priority value of each task and locks the task with the largest comprehensive priority value as the current task to be executed. S2, Task Identification and Operation Decomposition: The task identification and operation decomposition module matches the type of the currently executed task, generates a standardized operation sequence through a tree-like decomposition method, and simultaneously acquires environmental data collected by external sensing components; S3, Operation Parameter Analysis: The execution parameter analysis module generates motion control parameters, force control parameters, and dual-arm coordination parameters based on the operation requirements and environmental data corresponding to the current task being executed; S4, Operation Execution and Real-time Verification: The execution module drives the dual arms and moving mechanism to perform operations according to the control parameters. The real-time operation verification module verifies the operation results through multi-sensor fusion. If the result is normal, the next operation is executed. If the result is abnormal, the exception handling mechanism is triggered. S5, Task Completion and Data Recording: After all operations are completed, the robotic arm resets, the data storage module records all task information, including the overall priority value, operation sequence and verification results, and updates the status of the currently executing task to completed.

Images