Control method for improving motion efficiency of load robot

Abstract

The application provides a control method for improving the motion efficiency of a load manipulator, comprising the following steps: obtaining a motion instruction of the manipulator; evaluating the load to obtain a weight parameter of the load; decomposing the motion instruction; optimizing a load path of a grabbing part; the load path is divided into a climbing section and a pushing section, and the top elevation of the climbing section of the grabbing part is higher than the elevation when the grabbing part unloads to a target position; optimizing torque output; the torque output of the manipulator is the maximum torque in the climbing section, the maximum torque in the front section of the pushing section, no torque in the middle section, and no torque or reverse brake torque in the last section; through the above steps, the motion efficiency of the load manipulator is improved. The smooth path in the prior art is converted into a combination of the climbing section and the pushing section, so that the manipulator realizes the use of acceleration in the maximum torque control state to obtain the highest motion efficiency.

Description

Technical Field

[0001] This invention relates to the field of robotic arm motion control technology, and in particular to a control method for improving the motion efficiency of a load-bearing robotic arm. Background Technology

[0002] The development of robots is progressing rapidly, with large numbers being deployed in production workshops and warehouses to reduce labor intensity and improve efficiency. However, the purchase and maintenance costs of robots are high; therefore, significantly improving robot efficiency is equivalent to reducing the cost of robots. The improvement in robot efficiency is related to the movement speed of the robot's moving parts while ensuring accuracy. This speed improvement is related to the robot's drive mechanism, structural stiffness, accuracy constraints, inertia, and control methods. How to construct a balanced control model among these constraints is a technical challenge that the industry urgently needs to solve.

[0003] For an n-DOF robot system, a dynamic model of inertia matrix and joint motion is proposed:

[0004] (1);

[0005] in ∈R n×n For the robot's inertia matrix, ∈R n×n The matrix represents the centrifugal force and the Coriolis force. ∈R n The gravitational torque vector, ∈R n To control the torque, ∈R n This is an external disturbance.

[0006] To overcome the impact of inertia caused by increased robot efficiency and speed, primarily its effect on accuracy, a feedforward control scheme is proposed. This scheme pre-calculates and compensates for inertial forces, centrifugal forces, and Coriolis forces, reducing reliance on feedback loops and improving control accuracy and response speed. The feedforward control model is as follows:

[0007] (2);

[0008] in For feedforward inertial torque, For feedforward centrifugal force and Coriolis torque and This is a feedforward gravity torque. Feedforward control can significantly improve the robot's trajectory tracking accuracy and response speed. However, this scheme has a problem... In the control system, it is usually used to offset the dynamic interference of inertia force. That is, the current mainstream idea is how to offset the influence of the inertia of the robot. Further, the robot is usually considered as a rigid body, but in actual situation, the robot should be considered as an elastic body. For example, under the load state, the manipulator will vibrate due to elastic deformation. The current idea is to suppress the occurrence of vibration in the form of active or passive damping. However, these methods actually limit the motion efficiency of the manipulator.

[0009] Part of the background art is described to facilitate better understanding of the innovative ideas of the present application, not to acknowledge the prior art. SUMMARY

[0010] The technical problem to be solved by the present application is to provide a control method for improving the motion efficiency of a loaded manipulator, which can significantly improve the motion efficiency of the manipulator, especially the motion efficiency of the manipulator under load, and can partially utilize inertia to improve the motion efficiency. In the preferred scheme, the motion efficiency can also be improved by partially utilizing the flexible deformation of the manipulator.

[0011] To solve the above technical problems, the technical scheme adopted by the present application is: a control method for improving the motion efficiency of a loaded manipulator, comprising the following steps:

[0012] S1, obtaining the action instruction of the manipulator;

[0013] S2, evaluating the load to obtain the weight parameter of the load;

[0014] S3, decomposing the action instruction;

[0015] S4, optimizing the load path of the gripping part of the manipulator under load;

[0016] The load path is divided into a climbing section and a pushing section, and the top elevation of the climbing section of the gripping part is higher than the elevation when the gripping part unloads to the target position;

[0017] S5, optimizing the torque output of the manipulator under load;

[0018] The torque output of the manipulator is maximum torque constant torque output in the climbing section, and the torque output of the manipulator is vector adapted according to the motion direction in the pushing section. The maximum torque constant torque output is adopted in the front section of the pushing section, no torque output is adopted in the middle section, and no torque or reverse brake torque output is adopted in the end section;

[0019] Through the above steps, the motion efficiency of the loaded manipulator is improved.

[0020] In the preferred scheme, in the climbing section, the gripping part of the manipulator runs in the direction of reducing the arm span, so as to obtain greater rotational speed under the condition of maximum torque constant torque output.

[0021] In the preferred embodiment, the robotic arm is an articulated robotic arm;

[0022] The robotic arm employs a lightweight design, including lightweight materials and / or a lightweight structure;

[0023] Lightweight materials include aluminum alloys, aluminum-magnesium alloys, or titanium alloys;

[0024] Lightweight structures include using hollow structures or truss structures in the arm of the robotic arm.

[0025] In the preferred embodiment, a power acquisition circuit is provided in the control circuit of each motor of the robotic arm to provide feedback on the torque of each motor;

[0026] Hall sensors or absolute sensors are installed in each motor or motor transmission mechanism of the robotic arm to provide feedback on the motor's rotation angle and speed.

[0027] The arm span data is calculated based on the joint rotation angle.

[0028] In a preferred embodiment, an attitude sensor is provided in the gripping part of the robotic arm to provide feedback on the acceleration, spatial position, and attitude of the gripping part.

[0029] In a preferred embodiment, a strain gauge sensor is also provided on the forearm of the robotic arm, which is used to provide feedback on the flexible deformation of the forearm.

[0030] In the preferred scheme, the low-order vibration parameters of the manipulator under load are collected. When the manipulator reaches the top of the climbing section and the torque output of the manipulator is vector-adapted according to the direction of motion, the torque output is loaded within the range of low-order vibration amplitude reversal of the manipulator, starting from the moment of load vibration reversal.

[0031] In a preferred embodiment, the load assessment includes an inertia assessment of the load along the motion trajectory of the gripper of the robotic arm over time.

[0032] In the preferred scheme, it is verified whether the time in the push segment meets the time required for push in zero gravity. If not, the top elevation of the climbing segment is raised.

[0033] Or reduce the arm span of the robotic arm.

[0034] In the preferred embodiment, parameters from various sensors on the robotic arm are collected, the motion parameters of the robotic arm are analyzed, and an intelligent agent model is constructed to optimize the load path parameters and torque output parameters.

[0035] This invention provides a control method for improving the motion efficiency of a load-bearing manipulator. By employing a path optimization scheme, the smooth path in existing technologies is transformed into a combination of climbing and pushing sections. This allows the manipulator to utilize acceleration under maximum torque constant torque control, thereby achieving the highest motion efficiency. In a preferred embodiment, the optimized path scheme combining climbing and pushing sections, combined with optimized torque output control, enables partial utilization of inertia under load. In a further preferred embodiment, by collecting vibration parameters and controlling the timing of torque output, the vibration generated by the manipulator's elastic body is utilized to further improve motion efficiency. Attached Figure Description

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0037] Figure 1 This is a three-dimensional structural diagram of the present invention.

[0038] Figure 2 This is a schematic diagram of the optimized path of the present invention projected onto the nearest vertical plane.

[0039] Figure 3 This is a schematic diagram of the optimized path of the present invention projected onto a horizontal plane.

[0040] Figure 4 This is a schematic diagram of the first-order vibration fitting curve collected by the gripping part of the robot arm under load conditions according to the present invention.

[0041] Figure 5 This is a torque control diagram for the robotic arm output of the present invention.

[0042] Figure 6 This is a flowchart of the robotic arm control process of the present invention.

[0043] In the diagram: 1. Manipulator, 101. Grasping part, 102. Upper arm, 103. Lower arm, 104. Base, 2. Target position, 3. Nearest vertical plane, 4. Load, 5. Nearest trajectory line, 6. Fastest trajectory line, 7. Climbing section, 8. Pushing section, 9. Horizontal plane. Detailed Implementation

[0044] Example 1:

[0045] According to the inventor's analysis, in classical control theory, the control of a robotic arm follows the shortest trajectory line 5. That is, after receiving a motion command, robotic arm 1 will preferentially drive its gripping part 101 to move along the shortest trajectory line 5. The robotic arm's primary task is to transport loads from lower to higher locations. This results in control typically being performed at a constant speed, where acceleration and vibration are considered detrimental, leading to relatively low work efficiency. In a competition between a certain company and a foreign company, the work efficiency gap of robotic arms reached 20% to 50% due to various reasons. Based on research into improving the control efficiency of robotic arms, this invention proposes a control method to improve the motion efficiency of a load-bearing robotic arm, comprising the following steps:

[0046] like Figure 6 As shown in the figure, S1, obtain the motion command of the robot arm 1;

[0047] Preferred solutions include Figure 1 In the diagram, robotic arm 1 is an articulated robotic arm;

[0048] The robotic arm 1 adopts a lightweight design, including lightweight materials and / or a lightweight structure;

[0049] Lightweight materials include aluminum alloys, aluminum-magnesium alloys, or titanium alloys;

[0050] Lightweight structures include using hollow or truss structures in the arm of the robotic arm. Lightweight design can significantly reduce the impact of inertia on the robotic arm's movement, and may increase inertial deformation. However, the advantage of lightweight design is that it can improve the overall motion efficiency of the robotic arm, thereby increasing overall work efficiency by more than 10%.

[0051] In a preferred embodiment, each motor of the robotic arm 1 is equipped with a power acquisition circuit for feedback of the torque of each motor. With this structure, the output torque of the motor is obtained by acquiring parameters such as current and voltage. In order to improve the motion efficiency of the entire robotic arm, the present invention adopts a control scheme of maximum rated torque constant torque output. That is, during the control process, the torque is output at the maximum rated torque while ensuring accuracy, thereby greatly improving the motion efficiency of the robotic arm.

[0052] Hall sensors or absolute sensors are installed in each motor or motor transmission mechanism of the robotic arm 1 to provide feedback on the motor's rotation angle and speed. The Hall sensors are mounted coaxially with the motor or the output shaft of the transmission mechanism, while the absolute sensors are absolute encoders. Hall sensors are used in low-end robotic arms, while absolute sensors are used in high-end robotic arms. The arm span is calculated based on the joint rotation angle. Arm span refers to the horizontal projected distance between the gripper 101 and the base 104 of the robotic arm. Generally, under constant torque control, a larger arm span results in a lower speed.

[0053] In a preferred embodiment, the gripping part 101 of the robotic arm 1 is equipped with an attitude sensor to provide feedback on the acceleration, spatial position, and attitude of the gripping part 101. The attitude sensor is a 9-axis sensor, such as the Bosch Sensortec BMX160, with a package size of only 2.5 × 3.0 × 0.95 mm³, integrating a 3-axis accelerometer, a 3-axis gyroscope, and a 3-axis magnetometer. Alternatives include the TDK InvenSense ICM-20948+AK09916.

[0054] In a preferred embodiment, a strain gauge sensor is also provided on the forearm 103 of the robotic arm 1. The strain gauge sensor is used to provide feedback on the flexible deformation of the forearm 103. Optional strain gauge sensors include HBM's LY41-3 / 120LY and Vishay Precision Group (VPG) / Micro-Measurements EA-06-125UN-350.

[0055] Instructions given to the robotic arm, such as grabbing a load from one location and transporting it to a target location, typically include: moving above the gripping point, moving down to the gripping height, actuating the gripping unit 101 to grab the load 4, lifting the object, moving it above the target location 2, moving down to the placement height, releasing the gripping unit 101, releasing the load 4, and returning. Instruction formats include programming language instructions, coordinate point instructions, or graphical / teach pendant instructions. This also includes scheduling instructions issued at the task level in intelligent manufacturing or logistics systems. These instructions do not directly control the robotic arm; instead, the robot autonomously plans the path, avoids obstacles, and executes the gripping action. They feature edge computing characteristics. This invention primarily addresses scheduling instructions, but can also be applied to other instruction formats.

[0056] S2. Evaluate load 4 and obtain the weight parameters of load 4;

[0057] Different models of robotic arms employ different methods to obtain the weight parameters of load 4, including: 1. Weight estimation based on joint torque sensors, which involves measuring the actual output torque of the motors at each joint of the robotic arm and combining this with the robot's dynamics model to deduce the weight of the end-effector load; 2. Integrating a miniature weighing sensor into the gripper 101; 3. Installing a six-dimensional force / torque sensor, such as those from ATI, Kistler, or Honeywell, at the end-effector flange or wrist of the robotic arm to directly measure the force and torque acting on the end effector. In this example, method 2, integrating a miniature weighing sensor into the gripper 101, is used.

[0058] In a preferred embodiment, the evaluation of the load 4 includes an evaluation of the inertia of the load along the motion trajectory of the gripper 101 of the robotic arm 1 over time.

[0059] Including the mass, center of mass position, and inertia tensor of load 4, the following steps are included:

[0060] S201: Establish the inertial parameter model of the load;

[0061] S202: Determine the relative position and orientation of the gripper and the load;

[0062] S203: Calculate the total inertial parameters of the assembly in the end coordinate system;

[0063] The parallel axis theorem is used to transfer the load's inertial tensor from its center of mass to the end flange coordinate system;

[0064] (3);

[0065] m : For load quality, : is the position vector of the load centroid relative to the end flange. : is the inertial tensor of the load about its center of mass. : is the inertial tensor of the load about the end flange coordinate system. For the outer product, It is the identity matrix;

[0066] S204: Obtain the sequence of joint angles of the robot arm along the trajectory;

[0067] That is, to obtain time series data from the motion trajectory;

[0068] S205: Calculate the equivalent joint inertia or end-effector inertia mapping at each time point;

[0069] m t=const; The end-effector pose is calculated using forward kinematics, thus obtaining the position of the load's center of mass in the base coordinate system. Due to the change in attitude, the direction of the inertia tensor also rotates with time. A rotation matrix is ​​used to perform a coordinate transformation on the inertia tensor:

[0070] (4);

[0071] S206: Extract key inertia parameters.

[0072] For each time point, the following indicators were extracted, including end-point quality. Distance between the centroids at the ends Maximum principal moment of inertia Diagonal elements of the joint inertia matrix H ii (t) , condition number k(H(q)) .

[0073] S3. Decompose the action instructions;

[0074] It includes steps such as task instructions, task planning, motion planning, trajectory generation, low-level control, and joint drive execution.

[0075] S4. Optimize the load path of the gripping part 101 of the robot arm 1 under load;

[0076] like Figures 1-3 In the process, the load path is divided into an incline section 7 and a pushing section 8. The top elevation of the incline section 7 of the grabbing part 101 is higher than the elevation when the grabbing part 101 unloads to the target position 2.

[0077] In the climbing section 7, the load 4 is lifted to the top of the climbing section 7 with maximum rated torque output. During the torque output process in the climbing section 7, the influence of inertia on motion accuracy does not need to be considered. In the pushing section 8, the inertia generated by the load 4 and the gripping part 101 of the robot arm during the climbing section 7 is used to push the load 4 to the target position 2. During the movement in the pushing section 8, the weight of the load 4 does not need to be considered, and the running speed of the pushing section 8 is maximized. The principle of load path optimization is that in the motion from point a to point b, the velocity of the motion trajectory with acceleration is higher than the velocity of the uniform motion trajectory.

[0078] Preferred solutions include Figure 3 In the climbing section 7, the gripping part 101 of the robotic arm 1 moves in the direction of reducing the arm span, so as to obtain a greater rotational speed under the condition of maximum torque constant torque output.

[0079] In the preferred embodiment, it is verified whether the time in the pushing segment 8 meets the time required for pushing in a zero-gravity state. If not, the top elevation of the climbing segment 7 is raised. The higher elevation makes the pushing segment 8 less affected by the gravity of the load 4.

[0080] Alternatively, the arm span of robotic arm 1 can be reduced. The advantage of reducing the arm span of robotic arm 1 is that, under constant torque output conditions, a smaller arm span allows the gripping part 101 to achieve greater acceleration. This results in a reduction in the time required for the entire task.

[0081] In this example, it is quite difficult to calculate and optimize the motion trajectory to meet the expectations. Therefore, a calibration method was used to obtain the optimized motion trajectory.

[0082] During the calibration process, an appropriate load weight (4) is first selected based on the task. Then, an appropriate incline trajectory (7) is chosen, including selecting multiple incline segments (7) and their top positions. The motion trajectory is then verified to meet the requirements. Constraint 1 is that the speed of the optimized trajectory is higher than the speed of the nearest trajectory line (5). Constraint 2 is that the positional accuracy upon reaching the target position (2) meets the requirements. Through experiments with multiple sets of motion trajectories, the optimal solution is gradually approached.

[0083] like Figure 5 In step S5, the torque output of the robot arm 1 under load is optimized;

[0084] Figure 5 The dashed line in the figure represents the torque output curve of the prior art. The solid line represents the torque output curve of the present invention. The torque output commutation in the figure represents the commutation torque of the motor output portion.

[0085] The torque output of robotic arm 1 is at its maximum constant torque in the climbing section 7. In the pushing section 8, the torque output of robotic arm 1 is vector-adapted according to the direction of motion. That is, the motion of robotic arm 1 is decomposed into the individual motors of robotic arm 1, so that these motors drive robotic arm 1 to move in a preset direction. Maximum constant torque output is used in the early stage of the pushing section 8, no torque output is used in the middle stage, and no torque or reverse braking torque output is used in the final stage. For the vector adaptation described in this example, see [link to documentation]. Figure 2 , Figure 3 In this invention, the motion trajectory has a large bend, namely the bend between the climbing section 7 and the pushing section 8. When the gripping part 101 of the robot arm 1 moves to the top position of the climbing section 7, some motors of the robot arm 1 need to reverse, and the output torque pushes the load 4 along the pushing section 8 toward the target position 2. Due to the influence of the load 4's own inertia during the climbing section 7, the output torque in the pushing section 8 hardly needs to consider the influence of the load 4's gravity, thereby giving the load 4 a large acceleration. Compared with uniform motion, this increases the speed of the entire conveying process, that is, reduces the time required for the entire conveying task.

[0086] In this example, the main challenge in controlling the torque output lies in controlling the timing of the torque output. This is achieved through calibration, with constraint 1 ensuring the positional accuracy at target position 2 meets requirements; and constraint 2 minimizing the duration of the reverse braking torque output. By using multiple sets of torque output timing parameters, the system gradually approaches the optimal solution.

[0087] By following the steps above, the motion efficiency of the load-bearing robot can be improved.

[0088] Example 2:

[0089] In a further optimized scheme, parameters from various sensors on the robotic arm 1 are collected, the motion parameters of the robotic arm are analyzed, and an intelligent agent model is constructed to optimize the load path parameters and torque output parameters. Various parameters during actual motion are collected, an intelligent agent model is constructed, and the collected parameters are trained using the constraints in steps S4 and S5 to obtain the optimal motion trajectory and torque output strategy under different work task conditions. This includes the following steps:

[0090] S601, Data Acquisition and Preprocessing: Obtain the motion state of the robot arm, including joint angles, gripper 101 position, gripper 101 posture, load parameters, structural state, and task type.

[0091] The load parameters include load quality. m The position of the center of mass and the inertia tensor;

[0092] S602, Data synchronization and time alignment;

[0093] This includes using a uniform timestamp and selecting an appropriate sampling frequency, preferably a sampling frequency greater than or equal to 100Hz;

[0094] S603, Data preprocessing, including denoising, normalization, and building time windows;

[0095] S604, Constructing the State Vector s t As input values ​​to the model:

[0096] s t =[ q t , , , m , I , [,task_id](5);

[0097] Where q t Indicates the position or joint angle at time t. This represents the velocity at time t, i.e., the rate of change of position. This represents the torque or control input measured at time t, where m represents mass and I represents the inertia tensor. It represents the amount of deformation or deformation at time t, and task_id represents the number of the current task;

[0098] S605, Use a reinforcement learning framework;

[0099] Preferably, the PPO+LSTM+action space constraint PPO-LSTM-Constraint control model is adopted;

[0100] The Proximal Policy Optimization (PPO) algorithm boasts advantages such as high stability, suitability for continuous control, support for parallel sampling, and high training efficiency. The network structure employs an Actor-Critic network + LSTM, where LSTM handles time-series dependencies, the Actor outputs the optimized trajectory / torque strategy, and the Critic evaluates the strategy's value. Motion space constraints utilize pre-torque compensation, and the reward function employs a multi-objective weighted reward. In this example, reducing the total task time and end-effector acceleration are used as the weighted reward objectives. First, a single objective is optimized, such as reducing the total task time, before introducing multiple objectives, such as higher end-effector acceleration. The model is trained through simulation to verify the optimization effect, and then the training and verification of the agent model are completed through real-world testing. These steps significantly reduce the number of calibration experiments, especially when the working trajectory, load weight, target position 2, and initial position of load 4 change during the task. The optimal motion trajectory and torque output strategy can be quickly derived using the agent model, saving design and planning time for the robotic arm and reducing human intervention.

[0101] Example 3:

[0102] Preferred solutions include Figure 4 In this process, low-order vibration parameters of the manipulator 1 under load are collected. Preferably, in this example, first-order vibration parameters are used. When the manipulator 1 reaches the top of the climbing section 7, and the torque output of the manipulator 1 is vector-adapted according to the direction of motion, the torque output is loaded within the range of the first-order vibration amplitude reversal of the manipulator 1, starting from the moment of load vibration reversal. This scheme makes full use of the vibration, or flexible deformation, generated by the manipulator 1 and the load 4 on the manipulator to help generate the initial thrust of the pushing section 8. Together with the output torque of the manipulator's motor, it forms a thrust on the load 4 to increase the initial speed of the load 4.

[0103] Example 4:

[0104] The solution of this invention was tested in a textile machinery manufacturing enterprise. A FANUC M-20iA six-axis robotic arm was used, featuring a custom-designed aluminum alloy hollow arm and a custom-designed gripping section 101. An expanding gripper was employed; the expanding gripper inserts into the through-hole at the center of the load 4 and then expands its diameter to complete the gripping. A Bosch Sensortec BMX160 attitude sensor was installed in the gripping section 101, and HBM LY41-3 / 120LY strain gauge sensors were installed in both the forearm section 103 and the upper arm section 102. The test load 4 weighed 10 kg, was cylindrical, 260 mm in diameter, and 250 mm in height, with a through-hole at its center. The lifting height from the load to the target position 2 was 900 mm, the linear distance from the load to the target position 2 was 1500 mm, and the initial linear distance between the gripping section 101 of the robotic arm 1 and the load 4 was 1000 mm.

[0105] Comparative Example 1: Under conventional industrial camera vision correction, robot arm 1 takes about 11 seconds using the nearest trajectory line 5 and the FINE mode, which is a control method that precisely stops at each point.

[0106] Comparative Example 2: Under the condition of conventional industrial camera vision correction, the robot arm 1 takes about 8.5 seconds using the nearest trajectory line 5 and the most commonly used CNT50 mode, i.e. continuous path mode, with transition control.

[0107] Comparative Example 3: When the robot arm 1 uses the conventional industrial camera vision correction method and the nearest trajectory line 5 in CNT100 mode, it takes about 6.5 seconds. However, in this mode, due to the influence of inertia, there is a 15% situation where the positioning accuracy error needs to be adjusted later. When the later adjustment occurs, the time is about 8 to 10 seconds.

[0108] Embodiment 4 of the present invention: Under conventional industrial camera visual correction, the robotic arm 1, using the fastest trajectory line 6 selected after calibration, took approximately 6.2 seconds without any positioning accuracy error. Compared with the stable CNT50 mode, the speed is improved by approximately 35%. Compared with the unstable CNT100 mode, the speed is slightly improved. It is speculated that the speed improvement mainly comes from the increased acceleration of the pushing segment 8.

[0109] Embodiment 5 of this invention: Using conventional industrial camera vision correction, the robot arm 1 follows the fastest calibrated trajectory line 6. First-order vibration parameters of the robot arm 1 are collected. After reaching the fixed position on the climbing section 7, a reversing torque is simultaneously output at the amplitude reversal node. The thrust generated by the vibration assists the output torque in pushing the load 4 on the pushing section 8, taking approximately 5.5 seconds. A 10% positioning accuracy error occurs during later adjustments. Compared to the stable CNT50 mode, the speed is improved by approximately 54.5%. Compared to the CNT100 mode, the speed is improved by approximately 18%. Considering that no comparable control model is available for this example, and the optimization of the motion trajectory and torque output strategy has not yet reached its optimal level, the control speed of the robot arm 1 in this example has the potential for further improvement. This has potential value for in-depth research.

[0110] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The embodiments and features described in these embodiments can be arbitrarily combined without conflict. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A control method for improving the motion efficiency of a load-bearing manipulator, characterized in that: Includes the following steps: S1. Obtain the motion command of the robotic arm (1); S2. Evaluate the load (4) and obtain the weight parameters of the load (4); S3. Decompose the action instructions; S4. Optimize the load path of the gripping part (101) of the robot (1) under load; The load path is divided into an incline section (7) and a push section (8). The top elevation of the incline section (7) of the grabbing part (101) is higher than the elevation when the grabbing part (101) unloads to the target position (2). S5. Optimize the torque output of the robot (1) under load; The torque output of the robotic arm (1) is the maximum torque constant torque output in the climbing section (7), and the torque output of the robotic arm (1) is vector-adapted according to the direction of motion in the pushing section (8). The maximum torque constant torque output is used in the front section of the pushing section (8), the no torque output is used in the middle section, and the no torque or reverse braking torque output is used in the final section. By following the steps above, the motion efficiency of the load-bearing robot can be improved.

2. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 1, characterized in that: During the climbing section (7), the gripping part (101) of the robot (1) moves in the direction of reducing the arm span so as to obtain a greater rotational speed under the condition of maximum torque constant torque output.

3. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 1, characterized in that: The robotic arm (1) is an articulated robotic arm; The robotic arm (1) adopts a lightweight design, including lightweight materials and / or lightweight structure; Lightweight materials include aluminum alloys, aluminum-magnesium alloys, or titanium alloys; Lightweight structures include using hollow structures or truss structures in the arm of the robotic arm.

4. A control method for improving the motion efficiency of a load-bearing manipulator according to any one of claims 1 to 3, characterized in that: The control circuit of each motor of the robotic arm (1) is equipped with a power acquisition circuit, which is used to feed back the torque of each motor; Hall sensors or absolute sensors are provided in each motor or motor transmission mechanism of the robotic arm (1) to provide feedback on the rotation angle and speed of the motor. The arm span data is calculated based on the joint rotation angle.

5. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 4, characterized in that: An attitude sensor is provided in the gripping part (101) of the robotic arm (1) to provide feedback on the acceleration, spatial position and attitude of the gripping part (101).

6. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 5, characterized in that: A strain gauge sensor is also provided on the forearm (103) of the robotic arm (1), which is used to provide feedback on the flexible deformation of the forearm (103).

7. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 6, characterized in that: Collect the low-order vibration parameters of the manipulator (1) under load. When the manipulator (1) reaches the top of the climbing section (7) and the torque output of the manipulator (1) is vector-adapted according to the direction of motion, the torque output is loaded within the range of low-order vibration amplitude reversal of the manipulator (1) starting from the moment of load vibration reversal.

8. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 7, characterized in that: The evaluation of the load (4) includes determining the mass, center of mass position and inertia tensor of the load (4), as well as the dynamic analysis of the motion trajectory of the load (4) along the gripper (101) of the manipulator (1) and the corresponding time.

9. A control method for improving the motion efficiency of a load-bearing manipulator according to claim 8, characterized in that: Verify whether the time in the push segment (8) meets the time required for push in a zero-gravity state. If not, raise the top elevation of the climbing segment (7). Alternatively, reduce the arm span of the robotic arm (1).

10. The control method for improving the motion efficiency of a load-bearing manipulator according to claim 7, characterized in that: Collect the parameters of each sensor on the robot (1), analyze the motion parameters of the robot, and construct an intelligent agent model to optimize the load path parameters and torque output parameters.

Images