A control method and system for preventing a humanoid robot arm from falling

Abstract

This application discloses a control method and system for preventing fall in a humanoid robot arm, belonging to the field of robot control technology. The method first collects temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm. Based on the above data, it determines the remaining time before the temperature of each joint rises to the safe upper limit and the shortest movement time for the arm to move from the current posture to a safe position. The remaining time is compared with the shortest movement time, and the output torque limit of each joint is dynamically adjusted according to the comparison result. Based on the output torque limit, a smooth descent control command for the arm is generated. This invention can achieve active graded protection against joint overheating, preventing the arm from falling uncontrollably. It can also generate a descent trajectory without mechanical impact and automatically resume operation after the temperature recovers, significantly improving the safety of robot operation, the continuity of operation, and the service life of the equipment.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a control method and system for preventing humanoid robot arms from falling. Background Technology

[0002] When humanoid robot arms perform tasks such as handling and manipulation for extended periods, the joint motors generate heat due to continuous operation. To prevent overheating and damage to the motors, a shutdown method based on temperature thresholds is typically used. Temperature sensors are usually installed at the joint motors. When the monitored real-time temperature exceeds the preset safety limit, the control system directly cuts off the current output of that joint, causing the output torque to return to zero instantly.

[0003] However, this method of shutting down the robot by directly cutting off the power poses serious safety hazards in practical applications. When the arm is in a high position such as carrying heavy objects or extending horizontally, the sudden loss of joint power will cause the arm to fall rapidly under its own weight and load, forming an uncontrollable free fall motion, which is very likely to collide with surrounding people, damage the robot body, or damage the equipment below. Summary of the Invention

[0004] This invention provides a control method and system for preventing humanoid robot arms from falling, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a control method for preventing a humanoid robot arm from falling, comprising: S1. Collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm; S2. Based on the temperature data and output torque data of each joint, determine the remaining time for the temperature of each joint to rise to the safe upper limit; based on the current posture and load information of the arm, determine the shortest movement time for the arm to move from the current posture to the safe position. S3. Compare the remaining time with the shortest movement time, dynamically adjust the output torque limit of each joint based on the comparison result, generate a smooth descent control command for the arm based on the output torque limit, and drive the arm to a safe position according to the smooth descent control command.

[0006] Preferably, based on the temperature data and output torque data of each joint, the remaining time for the temperature of each joint to rise to the safe upper limit is determined, including: Pair historical torque data within the same time window with historical temperature data at the corresponding time point by point to generate a set of torque-temperature change correlation pairs. Trend fitting was performed on the set of torque-temperature change correlations to extract the influence coefficient of torque on temperature change; The current real-time output torque data is weighted and calculated based on the influence coefficient to obtain the torque contribution to temperature rise component. The torque contribution temperature rise component is superimposed with the base temperature rise component to obtain the corrected real-time temperature rise rate. Based on the corrected real-time temperature rise rate, the remaining time for the temperature of each joint to rise to the safe upper limit is derived.

[0007] Preferably, based on the current posture and load information of the arm, the shortest movement time for the arm to move from its current posture to a safe position is determined, including: Extract the physical motion constraint boundaries of each joint, and substitute the current posture and load information into the physical motion constraint boundaries to generate a set of joint torque constraints; Within the set of joint moment constraints, find the optimal time sequence for joint motion of the arm from its current position to a safe position; Extract the total motion duration corresponding to the optimal joint motion sequence, and use it as the shortest motion time.

[0008] Preferably, within the range of the joint moment constraint set, the optimal joint motion sequence for the arm to move from the current posture to a safe position is determined, including: Set the current posture as the starting point of the movement and the safe position as the ending point of the movement; Based on the joint moment constraint set, generate a set of all feasible joint motion paths from the start point to the end point of motion; For each path in the set of feasible joint motion paths, calculate its corresponding total motion duration; Compare the total motion duration of all paths and select the joint motion path with the shortest total motion duration; The joint motion path with the shortest total motion duration is converted into a time-optimal joint motion sequence arranged in chronological order.

[0009] Preferably, the output torque limit of each joint is dynamically adjusted based on the comparison results, including: The risk level is determined based on the difference between the remaining time and the shortest exercise time. Based on the risk level, determine the maximum permissible output torque ratio for each joint.

[0010] Preferably, the smooth descent control command for the arm is generated based on the output torque limit, including: Based on the current attitude, safe position, and the maximum allowable output torque ratio of each joint, a continuous slow descent trajectory that satisfies the torque constraint is generated; The continuous descent trajectory is decomposed into a real-time target angle sequence for each joint; Control each joint to track the real-time target angle sequence, while simultaneously limiting the output torque of each joint to not exceed the maximum allowable output torque ratio.

[0011] Preferably, generating a continuous descent trajectory that satisfies torque constraints includes: Extract the maximum permissible acceleration boundary and the maximum permissible jerk boundary for each joint; The torque constraint, the maximum allowable acceleration boundary, and the maximum allowable jerk boundary are integrated into a multi-dimensional motion constraint condition that is satisfied simultaneously. Define the starting and ending motion states of the trajectory. The ending motion state is a stationary state where the velocity, acceleration, and jerk are all zero. Under the constraints of multi-dimensional motion conditions, a set of candidate trajectory segments that satisfy full continuity of position, velocity, acceleration, and jerk is generated; By splicing the trajectory segments in the candidate trajectory segment set without abrupt changes, a continuous slow descent trajectory without mechanical impact is obtained.

[0012] Preferably, based on the risk level, the maximum permissible output torque ratio of each joint is determined, including: Obtain the remaining time for the temperature of each joint to rise to the safe upper limit; The baseline torque ratio is determined based on the risk level. Based on the ratio of the remaining time of each joint to the minimum remaining time of all joints, the reference torque ratio is adjusted differentially to obtain the independent maximum allowable output torque ratio for each joint.

[0013] Preferably, after driving the arm to a safe position according to the smooth descent control command, the method further includes: Continuously collect real-time temperature data for each joint; When the real-time temperature data of all joints drops to the safe temperature range, record the last working posture and task progress before the arm reaches the safe position. The control arm automatically moves from a safe position to the final working posture and automatically resumes unfinished tasks based on the task progress.

[0014] To address the aforementioned problems, the present invention also provides a fall prevention control system for a humanoid robot arm, the system comprising: The data acquisition module is used to collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm; The time prediction module is used to determine the remaining time for the temperature of each joint to rise to the safe upper limit based on the temperature data and output torque data of each joint; and to determine the shortest movement time for the arm to move from the current posture to the safe position based on the current posture and load information of the arm. The slow descent control module compares the remaining time with the shortest movement time, dynamically adjusts the output torque limits of each joint based on the comparison results, generates a smooth slow descent control command for the arm based on the output torque limits, and drives the arm to a safe position according to the smooth slow descent control command.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By comparing the remaining time for the temperature of each joint to rise to the safe upper limit with the shortest movement time for the arm to move from the current posture to the safe position, and dynamically adjusting the output torque limit of each joint based on the comparison results, a smooth descent control command is generated based on the output torque limit to drive the arm movement. This can proactively initiate a graded protection process before the joint temperature reaches the safe upper limit, avoiding the arm's uncontrolled fall caused by a sudden interruption of joint power. This achieves a deep integration of overheat protection and movement safety, effectively preventing permanent damage to the joint motors due to continuous overheating, and ensuring the stability and controllability of the arm's movement during the protection process. This significantly improves the safe operation level of the humanoid robot arm in high-load, continuous operation scenarios.

[0016] 2. Based on the precise temperature rise remaining time derivation through torque-temperature change correlation fitting and the joint-independent differentiated torque ratio adjustment, the upper limit of output torque can be precisely set according to the actual heating state of each joint. While ensuring the overall protection effect, it maximizes the preservation of the effective motion capability of the joint. Combined with the fully continuous impact-free descent trajectory generation technology that meets multi-dimensional motion constraints, it can completely eliminate mechanical shock and vibration during the movement process, protect the core transmission components such as the joint motor and reducer, and automatically resume unfinished tasks after all joint temperatures drop to a safe range. This effectively reduces the impact of the protection process on the continuity of production operations and further improves the overall operating efficiency of the robot and the service life of the equipment. Attached Figure Description

[0017] Figure 1 A flowchart of a control method for preventing a humanoid robot arm from falling, provided by the present invention; Figure 2 This invention provides a modular structure diagram of a control system for preventing falls in a humanoid robot arm. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Reference Figure 1 The diagram shown is a flowchart illustrating a fall prevention control method for a humanoid robot arm according to an embodiment of the present invention. In this embodiment, the fall prevention control method for a humanoid robot arm includes: S1. Collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm.

[0020] In practical implementation, the heat generated by the joints of the humanoid robot arm is mainly concentrated in two core areas: the stator winding of the motor and the reducer housing. A single temperature acquisition point is prone to missed detection due to local overheating, resulting in temperature rise prediction deviations. Estimating the output torque solely based on the motor current will be affected by reducer friction and transmission clearance, leading to errors. Furthermore, the lack of timing alignment for the four types of data—temperature, torque, attitude, and load—will cause timing misalignment in subsequent temperature rise prediction and motion time calculations, ultimately resulting in delayed or false triggering of the anti-fall control mechanism. Based on this, the present invention provides a preferred implementation method.

[0021] Specifically, in this embodiment of the invention, temperature sensing components are respectively installed at the motor stator winding and reducer housing of each joint of the humanoid robot arm. Temperature values ​​at the two locations are collected synchronously at fixed time intervals, and the average value of the two is taken as the real-time temperature data of the joint. For example, when the robot performs continuous assembly operations, the temperature sensing component of the shoulder joint will continuously capture the heat changes generated by the motor operation, and generate a set of temperature values ​​and upload them after a fixed time interval.

[0022] The output torque is indirectly calculated through the current feedback of the joint motor. At the same time, a torque detection component is installed at the joint output shaft for calibration. The calibrated value is used as the output torque data of the joint. For example, when the arm grasps workpieces of different weights, the torque detection component will sense the torsional force on the joint output bearing in real time and convert the force into the corresponding output torque data. If the arm is in an unloaded hovering state, the collected output torque data only corresponds to the static torque generated by the weight of the arm itself.

[0023] Attitude sensing components are deployed at the upper arm, forearm, and wrist of the arm. By combining the attitude data of each component with the linkage structure parameters of the arm, the spatial position and angle of each link of the arm are calculated in sequence, thereby determining the overall current attitude of the arm. For example, when the arm extends above the worktable, each attitude sensing component will report its own pitch, yaw, and roll angles. The spatial coordinates and orientation of the end effector of the arm can be obtained by transforming the linkage relationship.

[0024] A load detection component is installed at the root of the end effector. The load information is determined by detecting the vertical downward force exerted on the end effector. For example, when the arm grasps a workpiece, the load detection component will sense the weight of the workpiece. If the arm does not grasp any object, the load information collected is zero.

[0025] All collected temperature data, output torque data, posture data, and load information are aligned according to a unified timestamp to ensure that various types of data correspond to each other at the same moment, thus obtaining complete temperature data, output torque data, and current posture and load information of each joint of the humanoid robot.

[0026] In summary, this solution employs a dual-point temperature acquisition and averaging method for the joint, combined with dual torque detection through current estimation and output shaft torque calibration. Simultaneously, all acquired data are aligned with a unified timestamp, comprehensively covering the core heating area of ​​the joint, eliminating temperature measurement errors from a single acquisition point, correcting torque estimation deviations caused by the transmission link, and ensuring the timing consistency and accuracy of all input data.

[0027] S2. Based on the temperature data and output torque data of each joint, determine the remaining time for the temperature of each joint to rise to the safe upper limit; based on the current posture and load information of the arm, determine the shortest movement time for the arm to move from the current posture to the safe position.

[0028] Specifically, based on the temperature and output torque data of each joint, the remaining time for the temperature of each joint to rise to the safe upper limit is determined, including: Pair historical torque data within the same time window with historical temperature data at the corresponding time point by point to generate a set of torque-temperature change correlation pairs. Trend fitting was performed on the set of torque-temperature change correlations to extract the influence coefficient of torque on temperature change; The current real-time output torque data is weighted and calculated based on the influence coefficient to obtain the torque contribution to temperature rise component. The torque contribution temperature rise component is superimposed with the base temperature rise component to obtain the corrected real-time temperature rise rate. Based on the corrected real-time temperature rise rate, the remaining time for the temperature of each joint to rise to the safe upper limit is derived.

[0029] Among these, based on the current posture and load information of the arm, the shortest movement time for the arm to move from its current posture to a safe position is determined, including: Extract the physical motion constraint boundaries of each joint, and substitute the current posture and load information into the physical motion constraint boundaries to generate a set of joint torque constraints; Within the set of joint moment constraints, find the optimal time sequence for joint motion of the arm from its current position to a safe position; Extract the total motion duration corresponding to the optimal joint motion sequence, and use it as the shortest motion time.

[0030] Within the set of joint moment constraints, the optimal time sequence for the arm to move from its current position to a safe position is determined, including: Set the current posture as the starting point of the movement and the safe position as the ending point of the movement; Based on the joint moment constraint set, generate a set of all feasible joint motion paths from the start point to the end point of motion; For each path in the set of feasible joint motion paths, calculate its corresponding total motion duration; Compare the total motion duration of all paths and select the joint motion path with the shortest total motion duration; The joint motion path with the shortest total motion duration is converted into a time-optimal joint motion sequence arranged in chronological order.

[0031] In practice, the most recent continuous working duration is taken as the time window. The historical torque data and corresponding historical temperature data of each sampling moment within the window are extracted from the stored historical data. The temperature difference between two adjacent sampling moments is calculated as the temperature change at that moment. Then, the historical torque data of each sampling moment is paired with the corresponding temperature change. All the pairing results are combined to form the torque-temperature change correlation pair set.

[0032] For example, when a robot performs assembly work continuously, the most recent working time is taken as a time window. The shoulder joint torque data and temperature data at each sampling moment within this time are extracted, the temperature change at each moment is calculated relative to the previous moment, and then the torque data and temperature change are paired to generate a set of torque-temperature change correlation pairs for that joint.

[0033] Furthermore, the torque-temperature change correlation is established by sorting all data points in the set according to their torque values ​​from smallest to largest, then connecting adjacent data points to form a broken line, and finally drawing a smooth trend line along the overall direction of the broken line. The slope of the trend line is the influence coefficient of the torque on the temperature change. For example, when the torque value increases, the temperature change also increases, so the trend line slopes upward and the influence coefficient is positive. If the torque value increases and the temperature change decreases, the trend line slopes downward and the influence coefficient is negative.

[0034] The current real-time output torque data is then multiplied by the extracted influence coefficient. The result is the current torque contribution to the temperature rise of the joint. For example, if the current real-time output torque data of a certain joint is large and the corresponding influence coefficient is positive, then the calculated torque contribution to the temperature rise is also large, indicating that the current temperature rise rate of the joint due to the output torque is relatively fast.

[0035] It should be noted that the base temperature rise component is the inherent temperature rise rate of the joint when it is in an unloaded and stationary state due to the conduction of ambient temperature and the heating of its own electronic components. This value is pre-determined and stored through the unloaded temperature rise test before the robot leaves the factory. The calculated torque contribution temperature rise component is added to the pre-stored base temperature rise component, and the result is the corrected real-time temperature rise rate. For example, when the joint is in a high-load working state, the torque contribution temperature rise component is much larger than the base temperature rise component, and the corrected real-time temperature rise rate is mainly determined by the torque contribution temperature rise component.

[0036] It should be emphasized that the calculation process of the above-mentioned corrected real-time temperature rise rate can be quantitatively expressed by the following formula:

[0037] In the formula, Indicates the first Each joint The real-time temperature rise rate after correction is the rate at which the joint's temperature rises at the current moment, taking into account both the joint's inherent heat generation and the heat generation from the current output torque.

[0038] Indicates the first Each joint The baseline temperature rise component at any given moment is the inherent rate of temperature rise of the joint when it is in an unloaded and stationary state, caused solely by conduction of ambient temperature and heating of its own electronic components. This parameter is pre-determined and stored through an unloaded temperature rise test before the robot leaves the factory.

[0039] Indicates the first The influence coefficient of torque on temperature change of a joint is the contribution of unit output torque to the joint temperature rise rate. This parameter is obtained by trend fitting of the set of torque-temperature changes within the same time window.

[0040] Indicates the first Each joint The real-time output torque data is indirectly calculated through the current feedback of the joint motor and calibrated by the torque detection component installed at the joint output shaft.

[0041] This formula combines the inherent heat generation of the joint with the additional heat generation generated by the current output torque through linear superposition, resulting in a more accurate real-time temperature rise rate. When the real-time output torque of the joint increases, the corrected real-time temperature rise rate will increase linearly, and when the output torque decreases, the corrected real-time temperature rise rate will decrease linearly.

[0042] For example, in continuous assembly operations, the basic temperature rise component of the shoulder joint is a fixed value at the corresponding ambient temperature. When the output torque of the joint increases to move heavier workpieces, the temperature rise component contributed by the torque will increase significantly, and the corrected real-time temperature rise rate will rise rapidly.

[0043] Furthermore, first obtain the current real-time temperature data of the joint and the preset safe upper limit temperature, calculate the difference between the current temperature and the safe upper limit temperature, and then divide the difference by the corrected real-time temperature rise rate. The result is the remaining time for the joint temperature to rise to the safe upper limit.

[0044] For example, if the current temperature of a joint is still some distance from the upper limit of the safe temperature, a faster real-time temperature rise rate after correction will result in a shorter remaining time, while a slower real-time temperature rise rate after correction will result in a longer remaining time.

[0045] The derivation of the remaining time can be quantified using the following formula:

[0046] In the formula, Indicates the first The remaining time required for a joint temperature to rise from its current temperature to its safe upper limit temperature is the remaining time for the joint to reach the overheat protection threshold, assuming the current rate of temperature rise remains constant.

[0047] This indicates the upper limit of the joint's safe operating temperature. It is the highest temperature at which the joint can work stably for a long time without mechanical damage or electronic component failure. This parameter is preset and stored based on the temperature resistance rating of the joint's mechanical materials, the insulation class of the motor windings, and the operating temperature range of the reducer grease.

[0048] Indicates the first Each joint Real-time temperature data is collected by two temperature sensing components located on the stator winding of the joint motor and the reducer housing. The average of the two values ​​is taken as the real-time temperature of the joint.

[0049] The parameter has the same meaning as in the previous formula, and is the first... Each joint The real-time temperature rise rate after time correction.

[0050] This formula is based on the difference between the current temperature and the upper limit of safety temperature, divided by the corrected real-time temperature rise rate, to obtain the remaining time for the joint to reach the overheating state. This provides a quantitative basis for subsequent risk level classification and torque limit adjustment. When the corrected real-time temperature rise rate increases, the remaining time will decrease accordingly. The closer the current temperature is to the upper limit of safety temperature, the smaller the remaining time will be.

[0051] For example, if the safe upper limit temperature of a certain shoulder joint is a preset fixed value, and the current real-time temperature is still a certain difference from the safe upper limit temperature, when the joint's real-time temperature rise rate increases rapidly due to continuous high load operation, the calculated remaining time will be significantly shortened.

[0052] In addition, in this embodiment of the invention, the physical motion constraint boundaries of each joint include the maximum allowable output torque, the maximum allowable rotational speed, and the maximum allowable rotational angle of the joint. These boundaries can be pre-determined and stored through mechanical performance testing before the robot leaves the factory. The current posture and load information of the arm are substituted into the physical motion constraint boundaries of each joint to calculate the maximum torque range that each joint can output under the current posture and load. The maximum torque range of all joints combined is the joint torque constraint set. For example, when the arm is in an extended posture and the load is large, the maximum torque range that the shoulder joint can output will be smaller than its maximum torque range when unloaded.

[0053] The safe position can be set to a posture where the arm hangs naturally and is close to the body. In this posture, the stress on each joint is minimal and the heat dissipation effect is optimal. The current posture of the arm is collected and set as the starting point of the movement, and then the pre-set safe position is set as the ending point of the movement.

[0054] Furthermore, based on different combinations of the rotation sequence and rotation angle of each joint, all joint motion paths that can reach the end point from the starting point of the motion and whose output torque of each joint does not exceed the range corresponding to the joint torque constraint set are generated. All paths that meet the conditions are combined to form the set of feasible joint motion paths. If the output torque of a certain joint in a certain path exceeds the range of the joint torque constraint set, the path is determined to be infeasible and will not be included in the set of feasible joint motion paths.

[0055] Then, for each feasible joint motion path, the time required for each joint to complete the rotation is calculated according to the rotation angle and the corresponding maximum allowable rotation speed of each joint in the path. The maximum value among all joint rotation times is taken as the total motion time of the path. For example, if the shoulder joint needs to rotate a large angle in a certain path, and the time required to complete the rotation is the longest, then the total motion time of the path is equal to the rotation time of the shoulder joint.

[0056] Furthermore, the total motion durations corresponding to all feasible joint motion paths are arranged in ascending order, and the path ranked first is the joint motion path with the shortest total motion duration. If there are multiple shortest paths with the same total motion duration, the path with the smallest fluctuation in the output torque of each joint is selected as the final shortest joint motion path.

[0057] Then, the rotational movements of each joint in the shortest joint motion path are broken down according to the chronological order to obtain the angle that each joint needs to rotate at each moment. The combination of the joint rotation angles at all moments is the time-optimal joint motion sequence. For example, at the beginning of the movement, the shoulder joint and elbow joint start to rotate at the same time. At the middle of the movement, the wrist joint starts to rotate. At the end of the movement, all joints stop rotating at the same time.

[0058] Finally, extract the time value corresponding to the last moment from the generated time-optimal joint motion sequence. This value is the shortest movement time for the arm to move from the current posture to the safe position.

[0059] In summary, this scheme obtains the influence coefficient of torque on temperature by pairing and fitting historical torque and temperature changes within the same time window. It then combines the torque contribution to temperature rise component with the base temperature rise component to obtain the corrected real-time temperature rise rate, accurately deriving the remaining time for each joint temperature to rise to the safe upper limit. Simultaneously, it extracts the physical motion constraint boundary of the joint and substitutes it with attitude and load information to generate a joint torque constraint set. Within the constraint range, it solves the time-optimal joint motion sequence to determine the shortest motion time.

[0060] It is certain that this solution, through the above settings, can significantly improve the accuracy of temperature rise prediction and movement time calculation, providing a reliable quantitative basis for subsequent risk level classification and torque limit adjustment, ensuring that the arm moves smoothly to a safe position along the optimal path before the joint overheats, fundamentally avoiding control failure and the risk of arm falling due to data calculation deviations, and balancing control response speed and movement safety.

[0061] S3. Compare the remaining time with the shortest movement time, dynamically adjust the output torque limit of each joint based on the comparison result, generate a smooth descent control command for the arm based on the output torque limit, and drive the arm to a safe position according to the smooth descent control command.

[0062] Among them, the output torque limit of each joint is dynamically adjusted based on the comparison results, including: The risk level is determined based on the difference between the remaining time and the shortest exercise time. Based on the risk level, determine the maximum permissible output torque ratio for each joint.

[0063] Among them, the smooth descent control commands for the arm, generated based on the output torque limit, include: Based on the current attitude, safe position, and the maximum allowable output torque ratio of each joint, a continuous slow descent trajectory that satisfies the torque constraint is generated; The continuous descent trajectory is decomposed into a real-time target angle sequence for each joint; Control each joint to track the real-time target angle sequence, while simultaneously limiting the output torque of each joint to not exceed the maximum allowable output torque ratio.

[0064] The generation of a continuous, slowly descending trajectory that satisfies the torque constraint includes: Extract the maximum permissible acceleration boundary and the maximum permissible jerk boundary for each joint; The torque constraint, the maximum allowable acceleration boundary, and the maximum allowable jerk boundary are integrated into a multi-dimensional motion constraint condition that is satisfied simultaneously. Define the starting and ending motion states of the trajectory. The ending motion state is a stationary state where the velocity, acceleration, and jerk are all zero. Under the constraints of multi-dimensional motion conditions, a set of candidate trajectory segments that satisfy full continuity of position, velocity, acceleration, and jerk is generated; By splicing the trajectory segments in the candidate trajectory segment set without abrupt changes, a continuous slow descent trajectory without mechanical impact is obtained.

[0065] Among these, the maximum permissible output torque ratio for each joint is determined based on the risk level, including: Obtain the remaining time for the temperature of each joint to rise to the safe upper limit; The baseline torque ratio is determined based on the risk level. Based on the ratio of the remaining time of each joint to the minimum remaining time of all joints, the reference torque ratio is adjusted differentially to obtain the independent maximum allowable output torque ratio for each joint.

[0066] This includes, after the arm is driven to a safe position according to the smooth descent control command, the process also includes: Continuously collect real-time temperature data for each joint; When the real-time temperature data of all joints drops to the safe temperature range, record the last working posture and task progress before the arm reaches the safe position. The control arm automatically moves from a safe position to the final working posture and automatically resumes unfinished tasks based on the task progress.

[0067] In practice, the minimum remaining time for all joints to reach their safe upper limit is compared with the shortest movement time for the arm to move from its current position to a safe position. The difference between the two is calculated. For example, if the shoulder joint has the shortest remaining time among all joints, the difference is obtained by subtracting the remaining time of the shoulder joint from the shortest movement time.

[0068] Then, the risk level is determined based on the difference between the remaining time and the shortest exercise time. It should be noted that multiple consecutive difference intervals are set here, and each difference interval corresponds to a risk level. If the difference is large, it corresponds to a low risk level; if the difference is in the middle range, it corresponds to a medium risk level; and if the difference is small, it corresponds to a high risk level.

[0069] Based on the risk levels obtained from the classification, the reference torque ratio of each joint is determined. Different risk levels correspond to different preset reference torque ratios. Low risk level corresponds to a higher reference torque ratio, medium risk level corresponds to a medium reference torque ratio, and high risk level corresponds to a lower reference torque ratio. For example, when it is determined to be a high risk level, the reference torque ratio will be significantly reduced to strictly limit the output torque of the joint.

[0070] Furthermore, the remaining time for the temperature of each joint to rise to the safe upper limit is obtained. From the previously calculated remaining time results for each joint, the independent remaining time values ​​for all arm joints, such as the shoulder joint, elbow joint, and wrist joint, are extracted respectively.

[0071] Then, calculate the minimum of the remaining time for all joints, divide the remaining time for each joint by this minimum, and obtain the ratio for each joint. Then multiply this ratio by the reference torque ratio to obtain the independent maximum allowable output torque ratio for each joint. For example, if the remaining time of a certain joint is twice the minimum of the remaining time of all joints, then the maximum allowable output torque ratio of that joint is twice the reference torque ratio. The longer the remaining time of a joint, the higher the allowable output torque ratio.

[0072] Furthermore, the maximum permissible acceleration boundary and the maximum permissible jerk boundary of each joint are extracted. These two boundaries are pre-determined and stored through mechanical performance testing before the robot leaves the factory. The maximum permissible acceleration boundary is the maximum rate of change of velocity that the joint can withstand, and the maximum permissible jerk boundary is the maximum rate of change of acceleration that the joint can withstand.

[0073] Then, the torque constraints, maximum allowable acceleration boundaries, and maximum allowable jerk boundaries corresponding to the maximum allowable output torque ratio of each joint are integrated to generate multi-dimensional motion constraints that all joints must satisfy simultaneously during the motion process. The motion parameters of any joint at any time cannot exceed the range of these multi-dimensional motion constraints.

[0074] Furthermore, the starting motion state and the ending motion state of the trajectory are first set. The starting motion state is the position, velocity, acceleration and jerk of each joint in the current posture of the arm. The ending motion state is the stationary state in the safe position where the velocity, acceleration and jerk of each joint are all zero. For example, when the arm is currently in a horizontally extended posture, the starting motion state includes the actual position of each joint and the current motion parameters in that posture.

[0075] Then, under the constraints of multi-dimensional motion, the entire motion process is divided into multiple continuous motion stages. For each motion stage, a trajectory segment is generated that satisfies the condition of full continuity of position, velocity, acceleration, and jerk. All trajectory segments that meet the constraints are combined to form the candidate trajectory segment set.

[0076] Furthermore, the trajectory segments in the candidate trajectory segment set are spliced ​​without abrupt changes to ensure that the endpoint position, velocity, acceleration and jerk of the previous trajectory segment are completely consistent with the starting position, velocity, acceleration and jerk of the next trajectory segment. After splicing, a continuous slow descent trajectory without mechanical impact is obtained.

[0077] For example, the final velocity of the trajectory during the acceleration phase is the same as the starting velocity of the trajectory during the constant velocity phase, and the final velocity of the trajectory during the constant velocity phase is the same as the starting velocity of the trajectory during the deceleration phase.

[0078] Then, at fixed time intervals, the target angles corresponding to each joint at each moment are extracted sequentially from the continuous descent trajectory. The joint target angles at all moments are arranged in chronological order to obtain the real-time target angle sequence of each joint.

[0079] Then, in each control cycle, the actual angle of each joint is compared with the target angle at the corresponding moment in the real-time target angle sequence. The output torque of the joint is adjusted to make the actual angle track the target angle. At the same time, the output torque of each joint is monitored in real time. If the output torque of a certain joint is about to exceed its maximum allowable output torque ratio, the output torque of that joint is limited to not exceed that ratio.

[0080] Furthermore, the arm joints are driven to move synchronously according to the generated smooth descent control command. When all joints have completed the real-time tracking of the target angle sequence, the arm moves to a pre-set safe position, such as the arm moving smoothly from a horizontally extended posture to a safe position that hangs naturally close to the body.

[0081] After the arm reaches a safe position, real-time temperature data of each joint is continuously collected to monitor the temperature drop of each joint.

[0082] When the real-time temperature data of all joints is detected to have dropped to the preset safe temperature range, the last working posture and task progress before the arm reaches the safe position are retrieved and recorded. The last working posture is the spatial posture of the arm just before entering the slow descent control, and the task progress is the completion status of the work task being performed at that time.

[0083] Finally, the control arm automatically moves from a safe position to the recorded final working posture. During the movement, the motion constraints of each joint are followed to ensure smooth and shock-free movement. After reaching the final working posture, the unfinished work tasks are automatically resumed based on the recorded task progress. For example, if the workpiece assembly work was being performed, the remaining assembly work can be completed from the previously interrupted assembly steps after reaching the final working posture.

[0084] In summary, this solution classifies risk levels and determines the baseline torque ratio by the difference between the remaining time and the shortest movement time. Then, it performs differentiated torque adjustments based on the ratio of the remaining time to the minimum remaining time for each joint. At the same time, it integrates multi-dimensional motion constraints to generate a continuous descent trajectory without mechanical impact throughout the entire process. Finally, it adds an automatic operation recovery mechanism after the joints cool down.

[0085] This embodiment achieves dynamic grading and differentiated control of torque limitation, which not only avoids arm falls caused by local joint overheating, but also ensures the timeliness of slow descent movement, eliminates mechanical impact during movement, protects the joint mechanical structure, and can automatically resume unfinished work after the temperature recovers. It greatly improves the accuracy, stability and continuity of fall prevention control, and takes into account both equipment safety and production efficiency.

[0086] like Figure 2 The diagram shown is a modular structure diagram of a fall prevention control system for a humanoid robot arm provided by the present invention, which includes: The data acquisition module 100 is used to collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm. The time prediction module 200 is used to determine the remaining time for the temperature of each joint to rise to the safe upper limit based on the temperature data and output torque data of each joint; and to determine the shortest movement time for the arm to move from the current posture to the safe position based on the current posture and load information of the arm. The slow descent control module 300 is used to compare the remaining time with the shortest movement time, dynamically adjust the output torque limit of each joint according to the comparison result, generate a smooth slow descent control command for the arm based on the output torque limit, and drive the arm to move to a safe position according to the smooth slow descent control command.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A control method for preventing a humanoid robot arm from falling, characterized in that, The method includes: S1. Collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm; S2. Based on the temperature data and output torque data of each joint, determine the remaining time for the temperature of each joint to rise to the safe upper limit; based on the current posture and load information of the arm, determine the shortest movement time for the arm to move from the current posture to the safe position. S3. Compare the remaining time with the shortest movement time, dynamically adjust the output torque limit of each joint based on the comparison result, generate a smooth descent control command for the arm based on the output torque limit, and drive the arm to a safe position according to the smooth descent control command.

2. The control method for preventing fall of a humanoid robot arm as described in claim 1, characterized in that, Based on the temperature and output torque data of each joint, determine the remaining time for the temperature of each joint to rise to the safe upper limit, including: Pair historical torque data within the same time window with historical temperature data at the corresponding time point by point to generate a set of torque-temperature change correlation pairs. Trend fitting was performed on the set of torque-temperature change correlations to extract the influence coefficient of torque on temperature change; The current real-time output torque data is weighted and calculated based on the influence coefficient to obtain the torque contribution to temperature rise component. The torque contribution temperature rise component is superimposed with the base temperature rise component to obtain the corrected real-time temperature rise rate. Based on the corrected real-time temperature rise rate, the remaining time for the temperature of each joint to rise to the safe upper limit is derived.

3. The control method for preventing fall of a humanoid robot arm as described in claim 2, characterized in that, Based on the arm's current posture and load information, determine the shortest movement time for the arm to move from its current posture to a safe position, including: Extract the physical motion constraint boundaries of each joint, and substitute the current posture and load information into the physical motion constraint boundaries to generate a set of joint torque constraints; Within the set of joint moment constraints, find the optimal time sequence for joint motion of the arm from its current position to a safe position; Extract the total motion duration corresponding to the optimal joint motion sequence, and use it as the shortest motion time.

4. The control method for preventing fall of a humanoid robot arm as described in claim 3, characterized in that, Within the set of joint moment constraints, solve for the optimal time sequence of joint motions for the arm to move from its current position to a safe position, including: Set the current posture as the starting point of the movement and the safe position as the ending point of the movement; Based on the joint moment constraint set, generate a set of all feasible joint motion paths from the start point to the end point of motion; For each path in the set of feasible joint motion paths, calculate its corresponding total motion duration; Compare the total motion duration of all paths and select the joint motion path with the shortest total motion duration; The joint motion path with the shortest total motion duration is converted into a time-optimal joint motion sequence arranged in chronological order.

5. The control method for preventing fall of a humanoid robot arm as described in claim 1, characterized in that, The output torque limits of each joint are dynamically adjusted based on the comparison results, including: The risk level is determined based on the difference between the remaining time and the shortest exercise time. Based on the risk level, determine the maximum permissible output torque ratio for each joint.

6. The control method for preventing fall of a humanoid robot arm as described in claim 5, characterized in that, Based on the output torque limit, smooth descent control commands for the arm are generated, including: Based on the current attitude, safe position, and the maximum allowable output torque ratio of each joint, a continuous slow descent trajectory that satisfies the torque constraint is generated; The continuous descent trajectory is decomposed into a real-time target angle sequence for each joint; Control each joint to track the real-time target angle sequence, while simultaneously limiting the output torque of each joint to not exceed the maximum allowable output torque ratio.

7. The control method for preventing a humanoid robot arm from falling as described in claim 6, characterized in that, Generate a continuous, slowly descending trajectory that satisfies the torque constraint, including: Extract the maximum permissible acceleration boundary and the maximum permissible jerk boundary for each joint; The torque constraint, the maximum allowable acceleration boundary, and the maximum allowable jerk boundary are integrated into a multi-dimensional motion constraint condition that is satisfied simultaneously. Define the starting and ending motion states of the trajectory. The ending motion state is a stationary state where the velocity, acceleration, and jerk are all zero. Under the constraints of multi-dimensional motion conditions, a set of candidate trajectory segments that satisfy full continuity of position, velocity, acceleration, and jerk is generated; By splicing the trajectory segments in the candidate trajectory segment set without abrupt changes, a continuous slow descent trajectory without mechanical impact is obtained.

8. The control method for preventing fall of a humanoid robot arm as described in claim 5, characterized in that, Based on the risk level, determine the maximum permissible output torque ratio for each joint, including: Obtain the remaining time for the temperature of each joint to rise to the safe upper limit; The baseline torque ratio is determined based on the risk level. Based on the ratio of the remaining time of each joint to the minimum remaining time of all joints, the reference torque ratio is adjusted differentially to obtain the independent maximum allowable output torque ratio for each joint.

9. The control method for preventing a humanoid robot arm from falling as described in claim 1, characterized in that, After the arm is driven to a safe position according to the smooth descent control command, the following steps are also included: Continuously collect real-time temperature data for each joint; When the real-time temperature data of all joints drops to the safe temperature range, record the last working posture and task progress before the arm reaches the safe position. The control arm automatically moves from a safe position to the final working posture and automatically resumes unfinished tasks based on the task progress.

10. A control system for preventing fall of a humanoid robot arm, used to implement the control method for preventing fall of a humanoid robot arm as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition module is used to collect temperature data and output torque data of each joint of the humanoid robot, as well as the current posture and load information of the arm; The time prediction module is used to determine the remaining time for the temperature of each joint to rise to the safe upper limit based on the temperature data and output torque data of each joint; and to determine the shortest movement time for the arm to move from the current posture to the safe position based on the current posture and load information of the arm. The slow descent control module compares the remaining time with the shortest movement time, dynamically adjusts the output torque limits of each joint based on the comparison results, generates a smooth slow descent control command for the arm based on the output torque limits, and drives the arm to a safe position according to the smooth slow descent control command.

Images