Intelligent robot handling control method

Abstract

This invention relates to the field of robotics, and more particularly to a method for controlling the handling of an intelligent robot. The method includes: constructing an indoor map; sending instructions to an AI agent via a smart terminal, the instructions containing handling events of the intelligent robot; the AI ​​agent analyzing the instructions and identifying the handling events; obtaining the current coordinates of the intelligent robot; controlling the intelligent robot to move to the object-grabbing position based on the object's pick-up position; using the ROI ratio of the object's image as a feedback signal; and controlling the intelligent robot to approach the object using an image approach motion algorithm to grasp the object; formulating a second navigation path based on the object's handling position and placement position; and controlling the intelligent robot to move to the object's placement position based on the second navigation path and a constructed handling reward function to place the object. This invention improves the efficiency, safety, and generalization ability of the intelligent robot during handling.

Description

Technical Field

[0001] This invention relates to the field of robotics, and more particularly to a method for controlling the handling of intelligent robots. Background Technology

[0002] With the rapid development of intelligent manufacturing technology, industrial intelligent robots have been widely used in material handling, assembly operations, and other scenarios. In material handling scenarios, robotic arms need to have the ability to perceive complex environments, flexible motion planning capabilities, and the ability to quickly adapt to new tasks.

[0003] Currently, the control of intelligent robots in material handling scenarios in intelligent manufacturing mainly adopts the following two technical approaches: (I) Robotic arm control based on teach-in programming: This is currently the most widely used robotic arm control method in the industrial field. Its technical principle is as follows: the operator pre-sets the robotic arm's motion trajectory and action sequence through manual teaching, and the robotic arm repeatedly executes specific tasks according to the preset program. At the motion control level, a PID (proportional-integral-derivative) controller is typically used to adjust the motor parameters in real time to achieve trajectory tracking. However, this approach has significant technical drawbacks: once the environment changes (such as the position of the transported object shifting or the appearance of obstacles), the robotic arm cannot autonomously adjust its motion strategy, resulting in poor flexibility and insufficient adaptability, making it difficult to meet the needs of modern flexible manufacturing systems for rapid changeover and multi-variety, small-batch production.

[0004] (II) End-to-End Control Based on Deep Reinforcement Learning: To address the lack of flexibility in teach-and-write programming, researchers have recently proposed end-to-end control schemes based on deep reinforcement learning. This scheme utilizes a camera mounted on the wrist of a robotic arm to collect multi-angle image information, extracts visual features through deep learning models such as convolutional neural networks, and directly maps these features to control commands for the robotic arm (such as joint angles or end effector poses). To reduce the cost of trial-and-error learning in real physical environments, training is typically performed tens of thousands or even millions of times in a virtual simulation environment before the trained strategy is transferred and applied to the real world. Although this scheme theoretically possesses the potential for autonomous learning and environmental adaptation, it has revealed the following technical shortcomings in practical industrial applications: (1) The sample utilization efficiency is extremely low. Deep reinforcement learning requires intelligent robots to learn complex handling skills through massive trial and error (usually requiring millions of steps of exploration) by constantly interacting with the environment. Performing such large-scale exploration on physical intelligent robots is not only time-consuming, but also causes serious equipment wear and even damage, resulting in high economic costs.

[0005] (2) Poor safety. The exploration process of reinforcement learning is random and uncontrollable. Intelligent robots may perform dangerous actions that exceed safety constraints during the learning process, posing a threat to the safety of their own equipment, the surrounding environment and the operators.

[0006] (3) Limited generalization ability. Existing deep reinforcement learning-based strategies often overfit to specific training scenarios (such as fixed object placement, specific lighting conditions, and single type of transported objects). When the actual application scenario changes slightly (such as changes in lighting, differences in object shape, and background interference), the performance of the strategy drops sharply and it is difficult to adapt to the complex and ever-changing environment of industrial sites.

[0007] (4) The reward function is poorly designed: existing deep reinforcement learning-based strategies lack a refined reward design, resulting in slow convergence speed. Summary of the Invention

[0008] The technical problem to be solved by this invention is: in order to solve the problems of low efficiency, poor safety and limited generalization ability of existing deep reinforcement learning-based policy control intelligent robots in handling, this invention provides an intelligent robot handling control method that improves the efficiency, safety and generalization ability of intelligent robots in handling.

[0009] The technical solution adopted by this invention to solve its technical problem is: an intelligent robot handling control method, wherein the intelligent robot includes: An edge computing device, wherein an AI agent is installed on the edge computing device; The transportation mechanism includes a base and drive components located at the four corners of the base. The drive components consist of drive wheels and servo motors. The servo motors are used to control the rotational speed and direction of the drive wheels. The four drive wheels are used to drive the base to move in any direction. A robotic arm is mounted on the transport mechanism, and the robotic arm is equipped with multiple joint motors; Multiple sensing modules are used for sensing data acquisition, including a vision sensor, a tactile sensor, and an IMU inertial measurement unit; The method includes the following steps: S1, the intelligent robot is driven to move indoors based on SLAM mapping instructions, an indoor map is constructed using images collected by the visual sensor, and the indoor map is deployed in the AI ​​intelligent body; S2, Send an instruction to the AI ​​agent through a smart terminal. The instruction contains the handling event of the smart robot. The AI ​​agent analyzes the instruction and identifies the handling event of the smart robot. The instructions include natural language, information or programming language, and the transport events include the object to be transported, the position of the object to be picked up, and the position of the object to be placed. S3, obtain the current position coordinates of the intelligent robot, control the intelligent robot to move to the position of the object to be transported based on the current position coordinates and the position of the object to be transported, and use the ROI ratio of the image of the object to be transported as a feedback signal, combined with the image approach motion algorithm to control the intelligent robot to approach the object to be transported. S4, the AI ​​agent controls the robotic arm's movements by combining the constructed grasping collaborative reward function to grasp the object to be transported; S5, the AI ​​agent formulates a second navigation path based on the object's transport location and the object's placement location, and controls the intelligent robot to move to the object's placement location based on the second navigation path and the constructed transport reward function; S6, after the intelligent robot arrives at the location where the object to be transported is placed, the AI ​​agent controls the movement of the robotic arm by combining the constructed placement cooperative reward function to place the object to be transported.

[0010] Furthermore, specifically, step S1 includes the following steps: S11, start the intelligent robot, enter the standby state, and connect the power; S12, Initialize the SLAM algorithm module; S13, SLAM closed-loop detection mechanism is activated, and environmental map construction begins; S14, based on SLAM mapping instructions, control the intelligent robot to move in the indoor environment, and collect images in real time through a visual sensor; S15, the image is input into the target detection network for instance segmentation to obtain target features, which include: object name, object pixel coordinates and object ROI; S16, compare the object's ROI with the set threshold. If the ROI of the object is less than the threshold, return to step S14 to obtain the next frame image collected in real time by the visual sensor. If the ROI of the object is greater than the threshold, proceed to the next step; S17, determine whether the object is a known object based on the object name; If the object is known, then assign the corresponding semantic category label to the object; If it is an unknown object, the unknown object is stored as an unknown tag ID, and the unknown object is semantically annotated and stored frame by frame; S18, Update the global point cloud map based on the labeled object semantic information, and construct a 3D environment model containing semantic information; S19. Detect whether the intelligent robot has returned to a previously visited location area based on the SLAM closed-loop detection mechanism: If the judgment result is negative and a closed loop has not been formed, return to step S14 and control the intelligent robot to move and collect images of the new area. If the judgment result is yes, the indoor map construction is completed, and the indoor map is deployed in the AI ​​intelligent body.

[0011] Furthermore, specifically, step S3 includes the following steps: S31, by using SLAM positioning technology combined with an IMU inertial measurement unit to measure the angular velocity and acceleration of the intelligent robot's movement, the current position coordinates of the intelligent robot are obtained; S32, determine the navigation task type based on the current position coordinates and the object transport position. The navigation task type includes moving the intelligent robot to the pick-up position and / or transporting the object. Calculate the target visible area scattering in a straight line from the transported object to surrounding obstacles. The vector calculation formula for the target visible area is:

[0012] The center position of the object in the map coordinate system is ( x 0, y 0), the set of obstacle boundary points is represented as ; S33, Based on the current location coordinates and the navigation task type, determine whether the intelligent robot is within the visible area of ​​the target; If the target is within the visible area, proceed to the next step; If the target is not within the visible area, the first navigation path is determined based on the current location coordinates and the position of the transported object. The intelligent robot is then controlled to move to the visible area of ​​the target based on the first navigation path and the path tracking reward function before proceeding to the next step. S34, Let the ROI of the object to be moved in the image at the current moment be... R t The ROI of the object to be moved in the image at the previous moment was . R t 1. Construct an image approach motion strategy, expressed as:

[0013] in, α As the weight for the rate of change of ROI, β Weights are the absolute values ​​of ROI.c Directional deviation penalty weight, target The orientation angle of the object to be moved. current The current orientation angle of the intelligent robot; S35, determine whether the ROI of the transported object is increasing; If so, the intelligent robot will approach the object based on the image-based motion approach strategy and then proceed to the next step. If not, proceed to the next step; S36 controls the PWM values ​​of the four servo motors to make the forward direction of the intelligent robot close to the vector angle of the target visible area of ​​the object to be transported, thus completing the navigation of the intelligent robot.

[0014] Furthermore, specifically, step S4 includes the following steps: S41: Start the intelligent robot grasping task, enter the grasping operation state, and obtain the width dimension Wobject of the object to be moved; S42, compare the width dimension Wobject of the object to be transported with the maximum gripping width Whand of a single-arm dexterous hand to determine the applicable gripping strategy; S43, compare the width Wobject of the object to be moved with the maximum gripping width Whand of a single-arm dexterous hand to determine the applicable gripping strategy: If Wobject≤Whand, then the object to be transported is suitable for single-arm grasping. The robot arm is controlled to grasp the object according to the single-arm grasping strategy. If Wobject > Whand, then the object to be transported is suitable for dual-arm grasping. The robotic arm is controlled to grasp the object according to the dual-arm grasping strategy.

[0015] Furthermore, specifically, the formula for the transport reward function is as follows: rtransport = x 1 | x ¨ object | x 2 | y ¨ object | x 3 z ¨ object x 4 slip in, x ¨ object、y ¨ object、z ¨ object These represent the accelerations of the object to be transported relative to the robotic arm in three directions. slip For sliding detection signal, x 1= x 2= x 3=1.0 is the acceleration weighting coefficient for the offset of the object to be moved. x 4 is the weighting coefficient for the sliding of the object to be moved.

[0016] Furthermore, specifically, the formula for the placement collaborative reward function is as follows: rplace = g 1 1 aligned g 2 | vdescent vtarget | g 3 Fimpac Among them, 1 aligned To align the indicators, vdescent This represents the actual rate of descent. vtarget For the target descent speed, Fimpact To withstand the impact force, g 1 = 10.0 g 2 = 10.0, g 3 = 10.0.

[0017] Furthermore, specifically, grasping the object with one arm includes the following steps: A1, Calculate the three-dimensional spatial distance between the end effector of the robotic arm and the gripping point of the object to be transported. d t The calculation formula is:

[0018] in, This refers to the current position of the end effector of the robotic arm. The center position of the object to be transported;

[0019] A2, Calculate the degree of attitude alignment between the end effector posture of the robotic arm and the grasping posture of the object to be transported. r orientation The calculation formula is:

[0020] in, To enable dexterous hands to grasp the normal vector, The surface normal vector of the object; A3, based on the three-dimensional spatial distance d t Based on the degree of alignment with the stated posture, calculate the 3D distance reward. r grasp The calculation formula is:

[0021] in, d max For the maximum effective distance, µ 1 represents the reward weight based on proximity. µ 2 represents the reward weight for motion smoothness. µ 3 represents the reward weight for motion smoothness. v max This represents the maximum speed at which the arm moves. A4, based on 3D distance reward r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object to be transported; A5. Monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported. If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; A6 sends a clamping command to the dexterous hand of the robotic arm, controlling the fingers to close in order to grasp the object to be transported; The A7 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

[0022] Furthermore, specifically, grasping the object with both arms includes the following steps: B1, calculate the three-dimensional spatial distance between the end of each robotic arm and the gripping point of the object to be transported, which are respectively the first three-dimensional spatial distances. d L Second and third-dimensional spatial distance d R ; B2. Calculate the degree of posture alignment between the end effector posture of each robotic arm and the grasping posture of the object to be transported; B3, based on the first three-dimensional spatial distance d L Calculate the first 3D distance reward based on the corresponding pose alignment. ; According to the second three-dimensional spatial distance d L The second 3D distance reward is calculated based on the corresponding pose alignment. ; B4. Calculate the symmetry reward of the two robotic arms' positions relative to the center of the object to be transported. The calculation formula is:

[0023] in, ɛ To prevent division by zero of small constants; B5, Calculate the safe clearance between the ends of the two robotic arms and the surface of the object to be transported. The calculation formula is:

[0024] in, R ma for d arms = d safe The maximum reward value at that time, d arms The distance between the ends of the arms. d safe For safe distance threshold, k 1 and k 2 represents the penalty coefficients for being too close and too far, respectively; B6, based on the first 3D distance reward Second 3D distance bonus 1. Rewards based on the symmetry of the arm positions and safety gaps Calculate the dual-arm cooperative reward function r grasp The calculation formula is:

[0025] in, ; B7, Based on a Dual-Arm Collaborative Reward Function r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object; B8, monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported; If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; B9 sends a clamping command to the dexterous hand of the robotic arm, controlling the fingers to close in order to grasp the object to be transported; B10 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

[0026] Furthermore, specifically, the expression for the lift reward function is:

[0027] in, hThe current height of the object to be moved. h target The relative lifting height between the object being lifted and the robotic arm. sh For a high degree of tolerance, dslip、dtol for dslip exist dtol There is no penalty within the tolerance range; beyond the tolerance range, the penalty increases according to a quadratic function. dmax Maximum permissible sliding distance r 1 represents a high reward coefficient. r 2 represents the sliding penalty coefficient.

[0028] Furthermore, specifically, constructing the path tracking reward function includes: Let the first navigation path be { p 0, p 1, ..., pm},in pi =( xi , yes The current coordinates of the intelligent robot are: pcurrent =( xc , yc The nearest path point is pnearest The target point is pgoal Then path tracking reward rpath The formula is as follows: rpath = l 1 deviation l 2 | therobot thepath |+ l 3 Δ goal in, deviation This is the lateral offset distance. therobot The current orientation of the intelligent robot, thepath Δ represents the tangent direction of the path. goal The amount of distance reduction towards the target point ,l 1 represents the lateral offset penalty coefficient. l 2 represents the penalty coefficient for directional deviation. l The target is 3, which is close to the reward coefficient.

[0029] The beneficial effects of this invention are: (1) Based on SLAM mapping, the robot can achieve accurate perception and autonomous localization of the global environment, obtain complete environmental information in the handling scene in real time, accurately lock the robot's own pose, avoid positioning deviation and path deviation during the handling process, greatly improve the environmental adaptability of handling operations, and avoid safety hazards caused by collisions and deviations from the source, thus comprehensively improving the safety of handling operations.

[0030] (2) By integrating path planning and motor control through AI intelligent agents, the technical barriers of independent control modules and poor coordination in traditional technology are broken, realizing the integrated management and control of the entire process of robot handling path planning, instruction issuance and motor execution, simplifying system control logic, improving instruction transmission and execution efficiency, ensuring the continuity and accuracy of robot handling actions, and further optimizing the smoothness of handling operations.

[0031] (3) The constructed reward function can accurately guide the robot to optimize its handling behavior and select the optimal handling strategy. Attached Figure Description

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0033] Figure 1 This is a schematic diagram of the robot structure according to Embodiment 1 of the present invention.

[0034] Figure 2 This is a schematic diagram of the method flow of Embodiment 1 of the present invention.

[0035] Figure 3 This is a schematic diagram of the indoor map construction process according to Embodiment 1 of the present invention.

[0036] Figure 4 This is a schematic diagram of the robot grasping process according to Embodiment 1 of the present invention.

[0037] Figure 5 This is a schematic diagram of the robot handling process according to Embodiment 1 of the present invention.

[0038] Figure 6 This is a schematic diagram of the hardware structure of a computer device according to Embodiment 2 of the present invention.

[0039] In the diagram, 10 is a computer device; 1002 is a processor; 1004 is a memory; and 1006 is a transmission device. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0041] Example 1: like Figure 1As shown in the figure, this application provides an intelligent robot handling control method, the intelligent robot including: Edge computing devices, which are equipped with AI agents; The transportation mechanism includes a base and drive components located at the four corners of the base. Each drive component consists of drive wheels and servo motors. The servo motors control the rotational speed and direction of the drive wheels, and the four drive wheels drive the base to move in any direction. The drive wheels can be Mecanum wheels. The robotic arm is mounted on the transport mechanism and has multiple joint motors. Multiple sensing modules are used for sensing data acquisition. The sensing modules include a vision sensor, a tactile sensor, and an IMU (inertial measurement unit). like Figure 2 As shown, the method includes the following steps: S1, based on SLAM mapping instructions, drives an intelligent robot to move indoors, builds an indoor map through images collected by a visual sensor, and deploys the indoor map in the AI ​​intelligent body; S2, send instructions to the AI ​​agent through the smart terminal. The instructions include the intelligent robot's handling events. The AI ​​agent analyzes the instructions and identifies the intelligent robot's handling events. The instructions include natural language, information or programming language, and the transport events include the object to be transported, the location where the object is picked up, and the location where the object is placed. S3, obtain the current position coordinates of the intelligent robot, control the intelligent robot to move to the position of the object to be transported based on the current position coordinates and the position of the object to be transported, and use the ROI ratio of the image of the object to be transported as a feedback signal, combined with the image approach motion algorithm to control the intelligent robot to approach the object to be transported. S4, the AI ​​agent controls the robotic arm's movements by combining the constructed grasping collaborative reward function to grasp the object to be transported; S5, the AI ​​agent formulates a second navigation path based on the object's transport location and the object's placement location, and controls the intelligent robot to move to the object's placement location based on the second navigation path and the constructed transport reward function; S6: After the intelligent robot arrives at the location where the object to be transported is placed, the AI ​​agent controls the movement of the robotic arm by combining the constructed placement cooperative reward function to place the object to be transported.

[0042] In this embodiment, as Figure 3 As shown, step S1 specifically includes the following steps: S11, start the intelligent robot, enter the standby state, and connect the power; S12, Initialize the SLAM (Simultaneous Localization and Mapping) algorithm module; S13, SLAM closed-loop detection mechanism is activated, and environmental map construction begins; S14, based on SLAM mapping instructions, controls the intelligent robot to move in the indoor environment and collects images in real time through a visual sensor; S15, input the image into the object detection network for instance segmentation to obtain object features, which include: object name, object pixel coordinates and object ROI; S16, compare the object's ROI with the set threshold, the threshold value range is 10-30; If the object's ROI is less than the threshold, return to step S14 to obtain the next frame image captured in real time by the visual sensor. If the object's ROI is greater than the threshold, proceed to the next step; S17, determine whether an object is a known object based on its name; If the object is known, then assign the corresponding semantic category label to the object (such as table, chair, lamp). If it is an unknown object, the unknown object is stored as an unknown tag ID, and the unknown object is semantically annotated and stored frame by frame. S18, Update the global point cloud map based on the labeled object semantic information, and construct a 3D environment model containing semantic information; S19, Detect whether the intelligent robot has returned to a previously visited location area based on the SLAM closed-loop detection mechanism: If the judgment result is negative and a closed loop has not been formed, return to step S14 and control the intelligent robot to move and collect images of the new area. If the judgment result is yes, the indoor map construction is completed and the indoor map is deployed in the AI ​​intelligent body.

[0043] It should be noted that in this embodiment, the Large Language Model (LLM) is used to perform frame-by-frame semantic annotation and storage of unknown objects. Specifically, this includes: performing intelligent question-and-answer annotation on unknown objects, generating candidate semantic descriptions of unknown objects through multimodal reasoning; and in order to improve accuracy, the candidate semantic annotations are also manually reviewed and confirmed to ensure the accuracy of the annotations and correct erroneous identification results.

[0044] Furthermore, the newly labeled data (especially the correct labels of unknown objects) after manual review are added to the training set to perform incremental learning on the object detection network, thereby improving the model's ability to recognize new object categories. In this embodiment, the object detection network is the Mask R-CNN instance segmentation model.

[0045] In this embodiment, as Figure 4 As shown, step S3 specifically includes the following steps: S31, by using SLAM positioning technology combined with an IMU inertial measurement unit to measure the angular velocity and acceleration of the intelligent robot's movement, the current position coordinates of the intelligent robot are obtained; S32, determine the navigation task type based on the current position coordinates and the object's transport position. The navigation task type includes a task of moving the intelligent robot to a pick-up position and / or a task of transporting an object. Calculate the target visible area scattering a straight line from the transported object to surrounding obstacles. The vector calculation formula for the target visible area is:

[0046] The center position of the object in the map coordinate system is ( x 0, y 0), the set of obstacle boundary points is represented as ; S33, based on the current location coordinates and navigation task type, determine whether the intelligent robot is within the target's visible area; If the target is within the visible area, proceed to the next step; If the target is not within the visible area, a first navigation path is determined based on the current location coordinates and the position of the object being transported. The robot is then controlled to move to the visible area of ​​the target based on the first navigation path and the path tracking reward function before proceeding to the next step. It should be noted that, based on the global map, the first navigation path is calculated from the current location coordinates and the location of the object to be transported, generating a path point sequence. Within a period T (e.g., 5 seconds), a PWM sequence for the four-wheel motors is output. Based on the planned path, the PWM (Pulse Width Modulation) control sequence for the four-wheel drive motors is calculated to drive the intelligent robot along the path, controlling its movement to the target visible area. During movement, a path tracking reward is calculated. Based on the deviation of the intelligent robot's actual trajectory from the planned path, a path tracking reward value is calculated to evaluate and optimize navigation performance. When the intelligent robot deviates from the planned path, path correction is performed based on the path tracking reward value to ensure the intelligent robot moves to the target visible area. Furthermore, the path tracking reward function is constructed as follows: Let the first navigation path be { p 0, p 1, ..., pm},in pi =( xi , yes The current coordinates of the intelligent robot are: pcurrent =( xc , yc The nearest path point is pnearest The target point is pgoal Then path tracking reward rpath The formula is as follows: rpath = l 1 deviation l 2 | therobot thepath |+ l 3 Δ goal in, deviation The lateral offset distance is calculated using the following formula: deviation =∥ pcurren pnearest ∥, therobot The current orientation of the intelligent robot, thepath Δ represents the tangent direction of the path. goal The amount of distance reduction towards the target point ,l 1 represents the lateral offset penalty coefficient, set to 1.0. l 2 represents the directional deviation penalty coefficient, set to 0.5. l The target proximity reward coefficient is set to 3, which is set to 2.0.

[0047] S34, Let the ROI of the object to be moved in the image at the current moment be... R t The ROI of the object to be moved in the image at the previous moment was . R t 1. Construct an image approach motion strategy, expressed as:

[0048] in, α This is the weight for the rate of change of ROI, with a default value of 0.8. β This is the absolute value weight of the ROI, with a default value of 0.2. c The penalty weight for directional deviation is 0.3 by default. target The orientation angle of the object to be moved. current The current orientation angle of the intelligent robot; S35, determine whether the ROI of the transported object is increasing; If so, the intelligent robot will approach the object based on the image-based motion approach strategy and then proceed to the next step. If not, proceed to the next step; S36 controls the PWM values ​​of the four servo motors, making the forward direction of the intelligent robot close to the vector angle of the visible area of ​​the target object to be transported, thus completing the navigation of the intelligent robot.

[0049] In this embodiment, as Figure 5 As shown, step S4 specifically includes the following steps: S41: Start the intelligent robot grasping task, enter the grasping operation state, and obtain the width dimension Wobject of the object to be moved; S42, compare the width dimension Wobject of the object to be moved with the maximum gripping width Whand of the single-arm dexterous hand to determine the appropriate gripping strategy; S43, compare the width Wobject of the object to be moved with the maximum gripping width Whand of a single-arm dexterous hand to determine the applicable gripping strategy: If Wobject≤Whand, then the object to be transported is suitable for single-arm grasping. The robot arm is controlled to grasp the object according to the single-arm grasping strategy. If Wobject > Whand, then the object to be transported is suitable for dual-arm grasping. The robotic arm is controlled to grasp the object according to the dual-arm grasping strategy.

[0050] Furthermore, grasping an object with one arm specifically includes the following steps: A1, Calculate the three-dimensional spatial distance between the end effector of the robotic arm and the gripping point of the object to be transported. d t The calculation formula is:

[0051] in, This refers to the current position of the end effector of the robotic arm. The center position of the object to be transported; A2, Calculate the degree of attitude alignment between the end effector posture of the robotic arm and the grasping posture of the object to be transported. r orientation The calculation formula is:

[0052] in, To enable dexterous hands to grasp the normal vector, The surface normal vector of the object; A3, based on three-dimensional spatial distance d t Alignment with pose to calculate 3D distance reward r grasp The calculation formula is:

[0053] in, d max For the maximum effective distance, µ 1 represents the reward weight for proximity, set to 2.0. µ 2 represents the motion smoothness reward weight, set to 1.0. µ 3 represents the motion smoothness reward weight, set to 0.5. v max The maximum arm movement speed is used; if the arm movement speed exceeds the maximum arm movement speed, the 3D distance bonus is reduced. A4, based on 3D distance reward r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object to be transported; A5, monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported. If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; A6 sends clamping commands to the robotic arm's dexterous hand, controlling the fingers to close and grasp the object to be moved; The A7 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

[0054] Furthermore, grasping an object with both arms specifically includes the following steps: B1, calculate the three-dimensional spatial distance between the end effector of each robotic arm and the gripping point of the object to be transported, which are the first three-dimensional spatial distances. d L Second and third-dimensional spatial distance d R ; B2. Calculate the degree of posture alignment between the end effector posture of each robotic arm and the grasping posture of the object to be transported; B3, based on the first three-dimensional spatial distance d L Calculate the first 3D distance reward based on the corresponding pose alignment. ; According to the second three-dimensional spatial distance d L The second 3D distance reward is calculated based on the corresponding pose alignment. ; B4. Calculate the symmetry reward of the two robotic arms' positions relative to the center of the object to be transported. The calculation formula is:

[0055] in, ɛ To prevent division by zero of small constants; B5, Calculate the safe clearance between the ends of the two robotic arms and the surface of the object to be transported. The calculation formula is:

[0056] in, d arms The distance between the ends of the arms. d safe For safe distance threshold, R max >0 is d arms= d safe The maximum reward value at that time, which defaults to 5. k 1 and k 2 represents the penalty coefficients for being too close and too far, respectively. k 1=2, k 1=1, k 1> k 2. To make the punishment more severe when too close; B6, based on the first 3D distance reward Second 3D distance bonus 1. Rewards based on the symmetry of the arm positions and safety gaps Calculate the dual-arm cooperative reward function rgrasp The calculation formula is:

[0057] in, ; B7, Based on a Dual-Arm Collaborative Reward Function r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object; B8, monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported; If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; B9 sends a clamping command to the robotic arm's dexterous hand, controlling the fingers to close and grasp the object to be moved; B10 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

[0058] In this embodiment, during single-arm or dual-arm grasping, the output of the dexterous hand drive motor is dynamically adjusted based on the deviation between the tactile feedback and the target force threshold, thereby increasing or decreasing the current clamping force. F This ensures that the actual clamping force falls within the safe threshold range. During the grasping process, based on the object's current height change, posture stability, and energy efficiency, a comprehensive reward value for the lifting process is calculated using a lifting reward function. The robotic arm then generates and outputs a PWM control sequence for the arm joint motors based on this comprehensive reward value, driving one or both arms to lift the object off the support surface. To ensure the stability of the grasped object, tactile sensors monitor the relative sliding between the object and the finger contact surfaces. If movement is detected, the gripping force of the dexterous hand is dynamically increased to enhance friction and prevent the object from slipping.

[0059] Furthermore, the expression for the reward function is:

[0060] in, h The current height of the object to be moved, when h=h target At that time, the maximum reward is 1. h target The relative lifting height between the object being lifted and the robot arm is set to 10cm by default. sh For high tolerance, adjust according to task precision; default is 0.05m. dslip、dtol for dslip exist dtol There is no penalty within the tolerance range; beyond the tolerance range, the penalty increases according to a quadratic function. dmax Maximum allowable sliding distance, default 0.02m. r 1 represents the high reward coefficient, with a default value of 5.0. r 2 is the sliding penalty coefficient, which defaults to 2.0.

[0061] In this embodiment, in step S5, a second navigation path is determined based on the object's transport location and placement location. The intelligent robot is then controlled to move to the object's placement location based on this second navigation path, using the same method as in step S3. For brevity, this will not be repeated here. When the intelligent robot transports an object, it uses a transport reward function to ensure a smooth transport process. The transport reward function is expressed as follows: rtransport = x 1 | x ¨ object | x 2 | y ¨ object | x 3 z ¨ object x 4 slip in, x ¨ object、y ¨ object、z ¨ object These represent the accelerations of the object to be transported relative to the robotic arm in three directions. slip For sliding detection signal, x 1= x 2= x 3=1.0 is the acceleration weighting coefficient for the offset of the object to be moved. x 4 is the weighting coefficient for the sliding of the object to be moved.

[0062] It should be noted that the acceleration of the object to be moved relative to the robotic arm in three directions is obtained through a visual sensor or an IMU inertial measurement unit.

[0063] In this embodiment, step S6 specifically includes the following steps: S61 uses a vision system to precisely align the object to be placed with the target position.

[0064] S62 controls the robotic arm to drive the dexterous hand, causing the object to slowly descend along a preset path and approach the target position.

[0065] S63: Real-time detection of the impact force during object placement; determination of whether the impact force is less than a preset impact force threshold: If the impact force is greater than or equal to the threshold, pause the descent operation and return to step S61; if the impact force is less than the threshold, proceed to the next step. S64, after confirming the object is placed stably, the dexterous hand releases its gripping force on the object based on the placement cooperative reward function, fully opening its fingers to complete the object's placement. Furthermore, the placement cooperative reward function is expressed as follows: rplace = g 1 1 aligned g 2 | vdescent vtarget | g 3 Fimpac Among them, 1 aligned This is the alignment indicator (1 for alignment, 0 for disalignment). vdescent This represents the actual rate of descent. vtarget For the target descent speed, Fimpact To withstand the impact force, g1 = 10.0 g 2 = 10.0, g 3 = 10.0.

[0066] S65 controls the robotic arm to lift the dexterous hand back to the preset safe standby position, waiting for the next task.

[0067] In summary, this invention achieves precise global environmental perception and autonomous localization based on SLAM mapping. It can acquire complete environmental information within the handling scenario in real time, accurately lock the robot's own pose, and avoid problems such as positioning deviation and path deviation during the handling process, significantly improving the environmental adaptability of handling operations. Simultaneously, it avoids safety hazards caused by collisions and offsets from the source, comprehensively improving the safety of handling operations. Through the unified integration of path planning and motor control by an AI agent, it breaks down the technical barriers of independent and poorly coordinated control modules in traditional technologies. It achieves integrated management of the entire process of robot handling path planning, command issuance, and motor execution, simplifying system control logic, improving command transmission and execution efficiency, ensuring the continuity and accuracy of robot handling actions, and further optimizing the smoothness of handling operations. Finally, the constructed reward function can accurately guide the robot to optimize handling behavior and select the optimal handling strategy.

[0068] Example 2: This application provides a computer device including a processor and a memory. The memory stores at least one instruction or at least one program, which is loaded and executed by the processor to implement an intelligent robot handling control method as provided in the above method embodiments.

[0069] Figure 6 This diagram illustrates a hardware structure of a device for implementing an intelligent robot handling control method provided in an embodiment of this application. The device can participate in or include the apparatus or system provided in the embodiment of this application. Figure 6 As shown, the computer device 10 may include one or more processors 1002 (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 1004 for storing data, and a transmission device 1006 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 6 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer device 10 may also include... Figure 6 The more or fewer components shown, or having the same Figure 6 The different configurations shown.

[0070] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuit may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer device 10 (or mobile device). As involved in the embodiments of this application, the data processing circuit serves as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0071] The memory 1004 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the intelligent robot handling control method in this embodiment of the application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 1004, thereby implementing the aforementioned method. The memory 1004 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 1004 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer device 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0072] The transmission device 1006 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer device 10. In one example, the transmission device 1006 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 1006 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0073] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer device 10 (or mobile device).

[0074] Example 3: This application also provides a computer-readable storage medium, which can be disposed in a server to store at least one instruction or at least one program related to implementing an intelligent robot handling control method in the method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the intelligent robot handling control method provided in the above method embodiment.

[0075] Optionally, in this embodiment, the storage medium may be located at at least one of the multiple network servers in a computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0076] Example 4: This invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform an intelligent robot handling control method provided in the various optional embodiments described above.

[0077] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device, equipment, and storage medium embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0079] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0080] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for controlling the handling of intelligent robots, characterized in that, The intelligent robot includes: An edge computing device, wherein an AI agent is installed on the edge computing device; The transportation mechanism includes a base and drive components located at the four corners of the base. The drive components consist of drive wheels and servo motors. The servo motors are used to control the rotational speed and direction of the drive wheels. The four drive wheels are used to drive the base to move in any direction. A robotic arm is mounted on the transport mechanism, and the robotic arm is equipped with multiple joint motors; Multiple sensing modules are used for sensing data acquisition, including a vision sensor, a tactile sensor, and an IMU inertial measurement unit; The method includes the following steps: S1, the intelligent robot is driven to move indoors based on SLAM mapping instructions, an indoor map is constructed using images collected by the visual sensor, and the indoor map is deployed in the AI ​​intelligent body; S2, Send an instruction to the AI ​​agent through a smart terminal. The instruction contains the handling event of the smart robot. The AI ​​agent analyzes the instruction and identifies the handling event of the smart robot. The instructions include natural language, information or programming language, and the transport events include the object to be transported, the position of the object to be picked up, and the position of the object to be placed. S3, obtain the current position coordinates of the intelligent robot, control the intelligent robot to move to the position of the object to be transported based on the current position coordinates and the position of the object to be transported, and use the ROI ratio of the image of the object to be transported as a feedback signal, combined with the image approach motion algorithm to control the intelligent robot to approach the object to be transported. S4, the AI ​​agent controls the robotic arm's movements by combining the constructed grasping collaborative reward function to grasp the object to be transported; S5, the AI ​​agent formulates a second navigation path based on the object's transport location and the object's placement location, and controls the intelligent robot to move to the object's placement location based on the second navigation path and the constructed transport reward function; S6, after the intelligent robot arrives at the location where the object to be transported is placed, the AI ​​agent controls the movement of the robotic arm by combining the constructed placement cooperative reward function to place the object to be transported.

2. The intelligent robot handling control method according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11, start the intelligent robot, enter the standby state, and connect the power; S12, Initialize the SLAM algorithm module; S13, SLAM closed-loop detection mechanism is activated, and environmental map construction begins; S14, based on SLAM mapping instructions, control the intelligent robot to move in the indoor environment, and collect images in real time through a visual sensor; S15, the image is input into the target detection network for instance segmentation to obtain target features, which include: object name, object pixel coordinates and object ROI; S16, compare the object ROI with the set threshold; If the ROI of the object is less than the threshold, return to step S14 to obtain the next frame image collected in real time by the visual sensor. If the ROI of the object is greater than the threshold, proceed to the next step; S17, determine whether the object is a known object based on the object name; If the object is known, then assign the corresponding semantic category label to the object; If it is an unknown object, the unknown object is stored as an unknown tag ID, and the unknown object is semantically annotated and stored frame by frame; S18, Update the global point cloud map based on the labeled object semantic information, and construct a 3D environment model containing semantic information; S19. Detect whether the intelligent robot has returned to a previously visited location area based on the SLAM closed-loop detection mechanism: If the judgment result is negative and a closed loop has not been formed, return to step S14 and control the intelligent robot to move and collect images of the new area. If the judgment result is yes, the indoor map construction is completed, and the indoor map is deployed in the AI ​​intelligent body.

3. The intelligent robot handling control method according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31, by using SLAM positioning technology combined with an IMU inertial measurement unit to measure the angular velocity and acceleration of the intelligent robot's movement, the current position coordinates of the intelligent robot are obtained; S32, determine the navigation task type based on the current position coordinates and the object transport position. The navigation task type includes moving the intelligent robot to the pick-up position and / or transporting the object. Calculate the target visible area scattering in a straight line from the transported object to surrounding obstacles. The vector calculation formula for the target visible area is: The center position of the object in the map coordinate system is ( x 0, y 0), the set of obstacle boundary points is represented as ; S33, Based on the current location coordinates and the navigation task type, determine whether the intelligent robot is within the visible area of ​​the target; If the target is within the visible area, proceed to the next step; If the target is not within the visible area, the first navigation path is determined based on the current location coordinates and the position of the transported object. The intelligent robot is then controlled to move to the visible area of ​​the target based on the first navigation path and the path tracking reward function before proceeding to the next step. S34, Let the ROI of the object to be moved in the image at the current moment be... R t The ROI of the object to be moved in the image at the previous moment was . R t 1. Construct an image approach motion strategy, expressed as: in, α As the weight for the rate of change of ROI, β Weights are the absolute values ​​of ROI. γ Directional deviation penalty weight, target The orientation angle of the object to be moved. current The current orientation angle of the intelligent robot; S35, determine whether the ROI of the transported object is increasing; If so, the intelligent robot will approach the object based on the image-based motion approach strategy and then proceed to the next step. If not, proceed to the next step; S36 controls the PWM values ​​of the four servo motors to make the forward direction of the intelligent robot close to the vector angle of the target visible area of ​​the object to be transported, thus completing the navigation of the intelligent robot.

4. The intelligent robot handling control method according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41: Start the intelligent robot grasping task, enter the grasping operation state, and obtain the width dimension Wobject of the object to be moved; S42, compare the width dimension Wobject of the object to be transported with the maximum gripping width Whand of a single-arm dexterous hand to determine the applicable gripping strategy; S43, compare the width Wobject of the object to be moved with the maximum gripping width Whand of a single-arm dexterous hand to determine the applicable gripping strategy: If Wobject≤Whand, then the object to be transported is suitable for single-arm grasping. The robot arm is controlled to grasp the object according to the single-arm grasping strategy. If Wobject > Whand, then the object to be transported is suitable for dual-arm grasping. The robotic arm is controlled to grasp the object according to the dual-arm grasping strategy.

5. The intelligent robot handling control method according to claim 1, characterized in that, The formula for the transport reward function is as follows: rtransport = ξ 1 ∣ x ¨ object ∣ ξ 2 ∣ y ¨ object ∣ ξ 3 z ¨ object ξ 4 σslip in, x ¨ object, y ¨ object, z ¨ object These represent the accelerations of the object to be transported relative to the robotic arm in three directions. σslip For sliding detection signal, ξ 1= ξ 2= ξ 3=1.0 is the acceleration weighting coefficient for the offset of the object to be moved. ξ 4 is the weighting coefficient for the sliding of the object to be moved.

6. The intelligent robot handling control method according to claim 1, characterized in that, The formula for the placement collaborative reward function is as follows: rplace = ζ 1 1 aligned ζ 2 ∣ vdescent vtarget ∣ ζ 3 Fimpac Among them, 1 aligned To align the indicators, vdescent This represents the actual rate of descent. vtarget For the target descent speed, Fimpact To withstand the impact force, ζ 1 = 10.0 ζ 2 = 10.0, ζ 3 = 10.

0.

7. The intelligent robot handling control method according to claim 5, characterized in that, Grasping the object with one arm specifically includes the following steps: A1, Calculate the three-dimensional spatial distance between the end effector of the robotic arm and the gripping point of the object to be transported. d t The calculation formula is: in, This refers to the current position of the end effector of the robotic arm. The center position of the object to be transported; A2, Calculate the degree of attitude alignment between the end effector posture of the robotic arm and the grasping posture of the object to be transported. r orientation The calculation formula is: in, To enable dexterous hands to grasp the normal vector, The surface normal vector of the object; A3, based on the three-dimensional spatial distance d t Based on the degree of alignment with the stated posture, calculate the 3D distance reward. r grasp The calculation formula is: in, d max For the maximum effective distance, µ 1 represents the reward weight based on proximity. µ 2 represents the reward weight for motion smoothness. µ 3 represents the reward weight for motion smoothness. v max This represents the maximum speed at which the arm moves. A4, based on 3D distance reward r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object to be transported; A5. Monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported. If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; A6 sends a clamping command to the dexterous hand of the robotic arm, controlling the fingers to close in order to grasp the object to be transported; The A7 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

8. The intelligent robot handling control method according to claim 5, characterized in that, Grasping the object with both arms specifically includes the following steps: B1, calculate the three-dimensional spatial distance between the end of each robotic arm and the gripping point of the object to be transported, which are respectively the first three-dimensional spatial distances. d L Second and third-dimensional spatial distance d R ; B2. Calculate the degree of posture alignment between the end effector posture of each robotic arm and the grasping posture of the object to be transported; B3, based on the first three-dimensional spatial distance d L Calculate the first 3D distance reward based on the corresponding pose alignment. ; According to the second three-dimensional spatial distance d L The second 3D distance reward is calculated based on the corresponding pose alignment. ; B4. Calculate the symmetry reward of the two robotic arms' positions relative to the center of the object to be transported. The calculation formula is: in, ɛ To prevent division by zero of small constants; B5, Calculate the safe clearance between the ends of the two robotic arms and the surface of the object to be transported. The calculation formula is: in, R max for d arms = d safe The maximum reward value at that time, d arms The distance between the ends of the arms. d safe For safe distance threshold, k 1 and k 2 represents the penalty coefficients for being too close and too far, respectively; B6, based on the first 3D distance reward Second 3D distance bonus 1. Rewards based on the symmetry of the arm positions and safety gaps Calculate the dual-arm cooperative reward function r grasp The calculation formula is: in, ; B7, Based on a Dual-Arm Collaborative Reward Function r grasp Generate PWM control sequences for the motors of each joint of the robotic arm to drive the robotic arm to move toward the object; B8, monitor the spatial position of the end effector of the robotic arm and the gripping point of the object to be transported, and determine whether the end effector of the robotic arm has reached the position of the object to be transported; If the position of the object to be moved is not reached, adjust the position of the robotic arm and return to step A1; If the location of the object to be moved has been reached, proceed to the next step; B9 sends a clamping command to the dexterous hand of the robotic arm, controlling the fingers to close in order to grasp the object to be transported; B10 uses tactile sensors mounted on the fingertips to obtain the current clamping force. F Based on the preset minimum threshold of effective contact force F min and maximum threshold F max Determine the current clamping force F Is it within the safety threshold range? like F < F min or F > F max Then adjust the clamping force until it meets the requirements. F min ≤ F ≤ F max ; like F min ≤ F ≤ F max The robot arm is controlled to lift according to the set lift reward function.

9. The intelligent robot handling control method according to claim 7 or 8, characterized in that, The expression for the lift-up reward function is: in, h The current height of the object to be moved. h target The relative lifting height between the object being lifted and the robotic arm. σh For a high degree of tolerance, dslip, dtol for dslip exist dtol There is no penalty within the tolerance range; beyond the tolerance range, the penalty increases according to a quadratic function. dmax Maximum permissible sliding distance ρ 1 represents a high reward coefficient. ρ 2 represents the sliding penalty coefficient.

10. The intelligent robot handling control method according to claim 3, characterized in that, Constructing the path tracking reward function includes: Let the first navigation path be { p 0, p 1, ..., pm },in pi =( xi , yi The current coordinates of the intelligent robot are: pcurrent =( xc , yc The nearest path point is pnearest The target point is pgoal Then path tracking reward rpath The formula is as follows: rpath = λ 1 ddeviation λ 2 ∣ θrobot θpath ∣+ λ 3 D dgoal in, ddeviation This is the lateral offset distance. θrobot The current orientation of the intelligent robot, θpath Δ represents the tangent direction of the path. dgoal The amount of distance reduction towards the target point , λ 1 represents the lateral offset penalty coefficient. λ 2 represents the penalty coefficient for directional deviation. λ The target is 3, which is close to the reward coefficient.

Images