A multi-joint cooperative motion control method and system of a humanoid robot

Abstract

The application provides a multi-joint cooperative motion control method and system of a humanoid robot, and relates to the technical field of robot control. The application obtains gait phase data of a humanoid robot in different motion states, including a walking state, a running state and an emergency stop state. According to the gait phase data, a gait phase transition matrix is generated to determine a motion mode switching point and further determine a hysteresis compensation opportunity of a pneumatic artificial muscle to perform hysteresis compensation processing on the pneumatic artificial muscle of each lower limb joint of the humanoid robot to obtain a joint motion state. The angle change rate and motion acceleration parameters in the joint motion state are analyzed to generate a biological inspiration electric signal. According to the joint cooperative angle threshold and motion triggering time of the biological inspiration electric signal, each lower limb joint of the humanoid robot is driven to perform cooperative motion to realize seamless switching of multi-modal motion and improve motion flexibility and stability.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and in particular to a multi-joint cooperative motion control method and system for a humanoid robot. Background Technology

[0002] In scenarios such as service, industrial collaboration, and emergency rescue, humanoid robots need to frequently switch between walking, running, and sudden stops. This requires that the multiple joints of their lower limbs (such as hips, knees, and ankles) can achieve precise coordinated movement. They must not only respond quickly to state transition commands, but also ensure the smoothness and reasonable energy consumption of the movement process, so as to meet the requirements of complex tasks in dynamic environments for the flexibility and stability of robot movement.

[0003] Currently, the mainstream solution for addressing these needs adopts a pre-programmed control method based on kinematic models, combined with real-time feedback from joint position sensors. A closed-loop control algorithm adjusts joint drive parameters to track a preset motion trajectory. Simultaneously, by monitoring whether physical quantities such as motion speed and acceleration reach set thresholds, it triggers the switching of motion sequences between different motion modes. However, existing solutions have significant limitations. Firstly, they rely on fixed thresholds to determine the timing of motion mode switching, making it difficult to anticipate switching needs based on subtle changes in the humanoid robot's real-time motion state, resulting in significant lag in mode transitions. Secondly, they do not consider the characteristic delays of drive components during force transmission, leading to deviations between actual joint movement and commands, affecting the humanoid robot's posture stability. Summary of the Invention

[0004] The purpose of this application is to provide a multi-joint cooperative motion control method and system for humanoid robots, in order to solve the problems of lag in mode switching, deviation between actual joint movement and commands, and insufficient posture stability of humanoid robots in the prior art.

[0005] To address the aforementioned technical problems, in a first aspect, this application provides a multi-joint cooperative motion control method for a humanoid robot, comprising:

[0006] Gait phase data of a humanoid robot in different motion states, including walking, running and sudden stop;

[0007] Based on the gait phase data, a gait phase transition matrix is ​​generated to determine the movement mode switching point;

[0008] Based on the motion mode switching point, the timing of hysteresis compensation for pneumatic artificial muscles is determined, so as to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot to obtain the joint motion state.

[0009] Analyze the rate of change of angle and acceleration parameters in the joint motion state to generate a bio-inspired electrical signal;

[0010] Based on the joint coordination angle threshold and motion triggering time of the bio-inspired electrical signal, the lower limb joints of the humanoid robot are driven to perform coordinated movements to achieve seamless switching of multimodal motion.

[0011] Optionally, based on the gait phase data, a gait phase transition matrix is ​​generated to determine the motion mode switching point, including:

[0012] Extract the first phase feature value corresponding to the walking state, the second phase feature value corresponding to the running state, and the third phase feature value corresponding to the sudden stop state from the gait phase data.

[0013] The frequency of occurrence of the first phase feature value, the second phase feature value, and the third phase feature value is counted to calculate the transition probability between each two motion states;

[0014] All conversion probabilities are written into an initial matrix structure of a preset dimension to form a gait phase transition matrix. The row dimension of the initial matrix structure corresponds to the initial motion state, and the column dimension corresponds to the target motion state.

[0015] By traversing the gait phase transition matrix, the termination time of the initial motion state corresponding to the transition probability exceeding the preset probability threshold is determined as the motion mode switching point.

[0016] Optionally, extracting the first phase feature value corresponding to the walking state from the gait phase data includes:

[0017] The data segments in the gait phase data whose movement speed is within a preset walking speed range are defined as walking state data;

[0018] Collect the first rotation angle of each lower limb joint at each sampling time in the walking state data, and arrange all the first rotation angles to form a first joint angle sequence;

[0019] Record the time points of two adjacent foot strikes in the walking state data, calculate the first strike interval duration, and determine the first strike interval duration as the first movement cycle duration;

[0020] In the walking state data, the starting time points of the hip joint and knee joint are selected to calculate the first starting time difference, and the first starting time difference is used as the first adjacent joint movement time difference;

[0021] The first joint angle sequence, the first motion cycle duration, and the motion time difference of the first adjacent joints are combined to form the first phase feature value corresponding to the walking state.

[0022] Optionally, based on the motion mode switching point, the timing for hysteresis compensation of the pneumatic artificial muscles is determined to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot, thereby obtaining the joint motion state, including:

[0023] Extract the extension length and internal pressure values ​​of the pneumatic artificial muscles of the lower limb joint within a first preset time period before the movement mode switching point and within a second preset time period after the movement mode switching point;

[0024] Under the same extension length value, the difference between the internal pressure value and the corresponding standard pressure value in the standard pressure curve is calculated to obtain the hysteresis deviation value of the pneumatic artificial muscle.

[0025] The hysteresis compensation amount is calculated based on the time interval between the hysteresis deviation value and the motion mode switching point.

[0026] Using the motion mode switching point as a time reference, the compensation start time is set to a preset advance time before the motion mode switching, and the compensation end time is set to a preset duration after the motion mode switching point, so as to generate a hysteresis compensation timing.

[0027] Within the hysteresis compensation time, the internal pressure of the pneumatic artificial muscle is adjusted according to the hysteresis compensation amount, and the second rotation angle and rotation speed of the lower limb joint are collected during the adjustment process. The second rotation angle and rotation speed are combined into the joint motion state.

[0028] Optionally, the angular change rate and motion acceleration parameters in the joint motion state are analyzed to generate a bio-inspired electrical signal, including:

[0029] Based on the joint motion state, the angle change rate and motion acceleration of each lower limb joint were calculated;

[0030] By comparing the rate of change of angles of different lower limb joints at the same time, the difference in angle change between adjacent lower limb joints can be calculated to determine the degree of motion correlation between each lower limb joint.

[0031] Multiply all motion correlation coefficients by preset angle scaling factors to obtain the corresponding joint coordination angle thresholds;

[0032] The motion acceleration that appears most frequently under the same motion correlation degree is taken as the reference acceleration. All reference accelerations are multiplied by a preset time coefficient to obtain the time interval parameter of the corresponding motion trigger moment. Combined with the corresponding joint coordination angle threshold, a signal feature group is formed.

[0033] All signal feature groups are arranged according to the motion transmission sequence of each lower limb joint to form a bio-inspired electrical signal, wherein the pulse amplitude of the bio-inspired electrical signal is the corresponding joint coordination angle threshold, and the pulse interval is the corresponding motion triggering time.

[0034] Optionally, based on the joint coordination angle threshold and motion triggering time of the bio-inspired electrical signal, the lower limb joints of the humanoid robot are driven to perform coordinated movements to achieve seamless switching of multimodal motion, including:

[0035] Based on the movement transmission sequence of each lower limb joint, and taking the movement triggering time of the bio-inspired electrical signal as a benchmark, a corresponding action execution period is assigned to each lower limb joint. The action execution period is a preset action window before and after the movement triggering time.

[0036] The rotation angle amplitude is calculated based on the joint coordination angle threshold of each lower limb joint and the corresponding transmission ratio parameter.

[0037] Based on the motion correlation between each lower limb joint, the rotation direction of each lower limb joint is determined, and the rotation direction maintains a coordinated relationship with the motion direction of adjacent joints;

[0038] Combine the rotation angle amplitude and rotation direction corresponding to each lower limb joint into a target rotation angle command;

[0039] During the action execution period, the corresponding lower limb joints are driven to move according to the target rotation angle command, so as to collect the third rotation angle of each lower limb joint;

[0040] The angle deviation between the third rotation angle and the target rotation angle command is calculated, and the angle deviation value is compared with a preset deviation threshold. Based on the comparison result, all lower limb joints are driven to perform coordinated movements until the switching of different movement states is completed, so as to achieve seamless switching of multimodal movements.

[0041] Optionally, the rotation direction of each lower limb joint is determined based on the motion correlation between the joints, wherein the rotation direction maintains a coordinated relationship with the motion direction of adjacent joints, including:

[0042] Establish a cooperative relationship between adjacent lower limb joints, which includes a unidirectional cooperative relationship and a reverse cooperative relationship. Specifically, the cooperative relationship between adjacent lower limb joints with a motion correlation degree within a preset positive range is defined as a unidirectional cooperative relationship, and the cooperative relationship between adjacent lower limb joints with a motion correlation degree within a preset negative range is defined as a reverse cooperative relationship.

[0043] Based on the motion correlation and coordination between different adjacent lower limb joints, the rotation direction of each lower limb joint is determined so that the rotation direction of each lower limb joint maintains a corresponding coordination relationship with the motion direction of the adjacent lower limb joints.

[0044] Secondly, this application provides a multi-joint cooperative motion control system for a humanoid robot, comprising:

[0045] The acquisition module is used to acquire gait phase data of the humanoid robot in different motion states, including walking, running and sudden stop.

[0046] The generation module is used to generate a gait phase transition matrix based on the gait phase data in order to determine the motion mode switching point;

[0047] The compensation module is used to determine the timing of hysteresis compensation for pneumatic artificial muscles based on the motion mode switching point, so as to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot to obtain the joint motion state.

[0048] The analysis module is used to analyze the angle change rate and motion acceleration parameters in the joint motion state and generate bio-inspired electrical signals;

[0049] A driving module is used to drive the lower limb joints of the humanoid robot to perform coordinated movements based on the joint coordination angle threshold and movement trigger time of the bio-inspired electrical signal, so as to achieve seamless switching of multimodal movements. Thirdly, this application provides an electronic device, including:

[0050] Memory, used to store computer programs;

[0051] A processor is configured to execute the computer program to implement the steps of a multi-joint cooperative motion control method for a humanoid robot as described in the first aspect above.

[0052] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the multi-joint cooperative motion control method for a humanoid robot as described in the first aspect above.

[0053] This application provides a multi-joint cooperative motion control method for a humanoid robot. The method acquires gait phase data of the humanoid robot in different motion states, including walking, running, and sudden stop. Based on the gait phase data, a gait phase transition matrix is ​​generated to determine the motion mode switching point. Based on the motion mode switching point, the timing for hysteresis compensation of pneumatic artificial muscles is determined to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot, obtaining the joint motion state. The angle change rate and motion acceleration parameters in the joint motion state are analyzed to generate a bio-inspired electrical signal. Based on the joint cooperative angle threshold and motion trigger time of the bio-inspired electrical signal, the lower limb joints of the humanoid robot are driven to perform cooperative motion to achieve seamless switching of multi-modal motion. The technical solution provided by this invention collects gait data during walking, running, and sudden stops, providing comprehensive and real-time basic information for subsequent analysis of movement pattern characteristics, thus solving the problem of incomplete movement state judgment caused by insufficient data coverage. By dynamically calculating the probability of movement state transitions, it overcomes the limitation of relying on fixed thresholds to determine the timing of switching, enabling early prediction of movement mode switching needs and reducing mode transition lag. Addressing the characteristic delay of pneumatic artificial muscles, it precisely compensates before and after switching, reducing the deviation between actual lower limb joint movement and commands, and solving the problem of motion distortion caused by not considering the delay of drive components. By extracting angle change rate and acceleration parameters to generate electrical signals containing coordination information, it provides accurate timing and angle benchmarks for multi-joint coordinated movement, solving the problem of movement disjointness caused by relying solely on preset sequences for lower limb joint coordination. By coordinating the movements of each lower limb joint according to signal commands, it ensures the continuity and consistency of lower limb joint movements during movement mode switching, solving the problems of unsmooth switching processes and poor posture stability in existing solutions. Furthermore, based on the sequence of lower limb joint movement transmission, the action execution period is allocated with the movement triggering time of the bio-inspired electrical signal as the benchmark. The rotation angle amplitude is calculated based on the joint coordination angle threshold and transmission ratio parameters, and the rotation direction is determined by combining the inter-joint motion correlation, forming a target rotation angle command. During the action execution period, the joints are driven to move, and the actual rotation angle is collected and dynamically adjusted by comparing the deviation value until the motion state switch is completed. This addresses the problems of poor joint motion coordination and large deviations between commands and actual movements, ensuring that multiple joints move synchronously according to a preset coordination relationship during motion mode switching, ultimately achieving seamless switching of multimodal motion and improving the smoothness and posture stability of the humanoid robot's movement. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart of a multi-joint cooperative motion control method for a humanoid robot provided in this application;

[0056] Figure 2 A schematic diagram of a multi-joint cooperative motion control system for a humanoid robot provided in this application;

[0057] Figure 3 This is a schematic diagram of the structure of a computing device provided in this application. Detailed Implementation

[0058] To address the issues of response lag and poor stability in the current mainstream control strategies based on fixed thresholds and idealized models for multi-joint cooperative control of humanoid robots, the core shortcomings lie in their reliance on rigid thresholds for motion mode switching based on macroscopic parameters such as velocity and acceleration. This makes it difficult to capture the continuous evolution of gait phase, leading to lag in intent recognition and potential switching delays or misjudgments, resulting in stiff movements or even instability. Furthermore, existing frameworks often neglect the strong nonlinearity, time-varying hysteresis, and force transmission delay of actuators (such as pneumatic artificial muscles), causing joint output torque to deviate from expectations, introducing cooperative control errors, and affecting dynamic stability. In addition, the lack of reference to biological neural regulation mechanisms results in isolated joint control, hindering the achievement of biomimetic compliant coupling and efficient energy transfer. These problems collectively limit the smooth switching of multimodal motion and overall motion performance of humanoid robots in complex scenarios, making it difficult to meet the comprehensive requirements of flexibility, stability, and energy efficiency for high-dynamic tasks.

[0059] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] The core of this application is to provide a multi-joint cooperative motion control method for a humanoid robot, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0061] Step 101: Obtain gait phase data of the humanoid robot in different motion states, including walking, running and sudden stop.

[0062] In this step, gait phase data refers to a dataset reflecting the lower limb joint movement characteristics of a humanoid robot in different motion states. This includes data such as joint rotation angles, movement speed, movement cycle, and foot landing information collected by sensors, used to analyze the robot's motion state and gait characteristics. Walking refers to the humanoid robot's stable movement at a low speed, with a relatively fixed stride length and movement cycle, including characteristic data reflecting smooth joint movement and moderate speed in this state. Running refers to the humanoid robot's movement at a higher speed, with a larger stride length and shorter movement cycle, including characteristic data reflecting large joint movement amplitude and high speed in this state. Sudden stop refers to the humanoid robot's rapid deceleration from a motion state to a complete stop, including characteristic data reflecting a sharp decrease in speed and large changes in joint forces in this state.

[0063] In this embodiment, angle sensors, speed sensors, and pressure sensors installed on the robot's lower limb joints are used to collect data such as joint rotation angles, movement speed, and foot contact pressure in real time when the robot is walking, running, or stopping suddenly. The dataset formed by arranging these data in chronological order is the gait phase data. In the walking state, the robot's movement speed is in a low range and the gait is stable. In the running state, the movement speed is high and the stride is large. In the sudden stop state, the movement speed drops rapidly from a certain value to zero. The speed data collected by the sensors can distinguish these three movement states, and data for each state is collected to form complete gait phase data.

[0064] Step 102: Generate a gait phase transition matrix based on the gait phase data to determine the motion mode switching point;

[0065] In this embodiment, the first phase feature value corresponding to the walking state, the second phase feature value corresponding to the running state, and the third phase feature value corresponding to the sudden stop state are extracted from the gait phase data; the frequency of occurrence of the three phase feature values ​​is counted to calculate the transition probability between each two movement states; all transition probabilities are written into an initial matrix structure of a preset dimension to form a gait phase transition matrix; the gait phase transition matrix is ​​traversed, and the termination time of the starting movement state corresponding to the transition probability exceeding the preset probability threshold is determined as the movement mode switching point.

[0066] Step 103: Based on the motion mode switching point, determine the timing of hysteresis compensation for the pneumatic artificial muscles, so as to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot to obtain the joint motion state.

[0067] In this embodiment, the extension length and internal pressure values ​​of the pneumatic artificial muscle of the lower limb joint are extracted within a first preset time period before the movement mode switching point and a second preset time period after the movement mode switching point. Under the same extension length value, the difference between the internal pressure value and the corresponding standard pressure value in the standard pressure curve is calculated to obtain the hysteresis deviation value. Based on the time interval between the deviation and the movement mode switching point, the hysteresis compensation amount is calculated. Using the movement mode switching point as the time reference, the compensation start time is set to a preset advance time before the movement mode switching, and the compensation end time is set to a preset duration after the movement mode switching point to generate a hysteresis compensation timing. Within the hysteresis compensation timing, the internal pressure of the pneumatic artificial muscle is adjusted according to the hysteresis compensation amount, and the second rotation angle and rotation speed of the lower limb joint are collected during the adjustment process, and the two are combined into a joint movement state.

[0068] Step 104: Analyze the angle change rate and motion acceleration parameters in the joint motion state to generate a bio-inspired electrostatic signal;

[0069] In this embodiment, the angle change rate and motion acceleration of each lower limb joint are calculated based on the joint motion state; the angle change rate of different lower limb joints at the same time is compared to calculate the angle change difference between adjacent lower limb joints to determine the motion correlation between each lower limb joint; all motion correlations are multiplied by a preset angle proportionality coefficient to obtain the corresponding joint coordination angle threshold; the motion acceleration that occurs most frequently under the same motion correlation is taken as the reference acceleration, and all reference accelerations are multiplied by a preset time coefficient to obtain the time interval parameter of the corresponding motion triggering time, which is combined with the corresponding joint coordination angle threshold to form a signal feature group; all signal feature groups are arranged into a bio-inspired electrical signal according to the motion transmission sequence of each lower limb joint.

[0070] Step 105: Based on the joint coordination angle threshold and motion triggering time of the bio-inspired electrical signal, drive the lower limb joints of the humanoid robot to perform coordinated movements to achieve seamless switching of multimodal motion;

[0071] In this embodiment, based on the motion transmission sequence of each lower limb joint and the motion triggering time of the bio-inspired electrical signal as a benchmark, a corresponding action execution period is assigned to each lower limb joint. The action execution period is a preset action window before and after the motion triggering time. The rotation angle amplitude is calculated based on the joint coordination angle threshold and the corresponding transmission ratio parameter of each lower limb joint. The rotation direction of each lower limb joint is determined according to the motion correlation between each lower limb joint. The rotation angle amplitude and rotation direction corresponding to each lower limb joint are combined into a target rotation angle command. During the action execution period, the corresponding lower limb joint is driven to move according to the target rotation angle command to collect the third rotation angle of each lower limb joint. The angle deviation value between the third rotation angle and the target rotation angle command is calculated, and the angle deviation value is compared with a preset deviation threshold. Based on the comparison result, all lower limb joints are driven to perform coordinated movement until the switching of different motion states is completed, so as to achieve seamless switching of multimodal motion.

[0072] This application's embodiments achieve accurate prediction of motion mode switching; by performing hysteresis compensation on pneumatic artificial muscles before and after the switching point, the characteristic delay of the drive components is considered, reducing the deviation between joint movement and commands; by analyzing the joint movement state to generate bio-inspired electrical signals containing joint coordination information, the joints are driven to move according to the coordination relationship, solving the problem of poor joint coordination, and finally realizing seamless switching between multimodal movements such as walking, running, and sudden stopping of the humanoid robot, improving the stability, flexibility, and posture stability of the movement, and meeting the needs of different scenarios.

[0073] This application provides a specific embodiment. Step 102 involves generating a gait phase transition matrix based on the gait phase data to determine the motion mode switching point. This specifically includes the following steps:

[0074] Step 201: Extract the first phase feature value corresponding to the walking state, the second phase feature value corresponding to the running state, and the third phase feature value corresponding to the sudden stop state from the gait phase data.

[0075] In this step, the first phase feature value refers to the set of motion characteristics of the walking state extracted from walking state data, including the joint angle sequence, movement cycle duration, and time difference of adjacent joint movement during the walking state, used to characterize the gait pattern of the walking state. The second phase feature value refers to the set of motion characteristics of the running state extracted from running state data, including the second joint angle sequence, second movement cycle duration, and time difference of the second adjacent joint movement during the running state, used to characterize the gait pattern of the running state. The third phase feature value refers to the set of motion characteristics of the sudden stop state extracted from sudden stop state data, including the third joint angle sequence, third movement cycle duration, and time difference of the third adjacent joint movement during the sudden stop state, used to characterize the gait pattern of the sudden stop state.

[0076] In this embodiment, the data segment in the gait phase data where the movement speed is within a preset walking speed range is defined as walking state data; the first rotation angle of each lower limb joint is collected at each sampling moment in the walking state data, and all the first rotation angles are arranged to form a first joint angle sequence; the time points of two adjacent foot landings in the walking state data are recorded, and the first landing interval is calculated, which is determined as the first movement cycle duration; in the walking state data, the movement start time points of the hip joint and knee joint are selected to calculate the first start time difference, which is used as the first adjacent joint movement time difference; the first joint angle sequence, the first movement cycle duration, and the first adjacent joint movement time difference are combined to form the first phase feature value corresponding to the walking state. Similarly, the corresponding second joint angle sequence, the second movement cycle duration, and the second adjacent joint movement time difference are extracted from the running state data and combined to form the second phase feature value; the corresponding third joint angle sequence, the third movement cycle duration, and the third adjacent joint movement time difference are extracted from the sudden stop state data and combined to form the third phase feature value.

[0077] Step 202: Count the frequency of occurrence of the first phase feature value, the second phase feature value and the third phase feature value to calculate the transition probability between each two motion states;

[0078] In this step, the transition probability refers to the likelihood of a transition between two different motion states. It is calculated based on the frequency of occurrence of the first, second, and third phase eigenvalues ​​and is used to reflect the trend of motion state transition.

[0079] In this embodiment, the occurrences of the first, second, and third phase feature values ​​in the gait phase data are counted. The total number of occurrences of the corresponding feature values ​​is recorded when the gait transitions from walking to running, walking to a sudden stop, running to walking, running to a sudden stop, a sudden stop to walking, and a sudden stop to running. The number of occurrences of the phase feature value in each transition direction is then divided by the total number of occurrences of the phase feature value in all transition directions to obtain the transition probability between each pair of motion states. The boundary for determining a state transition is defined as follows: when the motion state feature value (such as joint angle sequence or motion cycle) in the gait phase data changes from the initial motion state (such as walking) to the target motion state (such as running) within three consecutive sampling times, it is determined as a complete state transition, and the occurrence of the phase feature value corresponding to that transition direction is recorded as once. For example, if the first phase feature value occurs 20 times in the transition from walking to running, and the total number of occurrences in all transition directions is 100, then the transition probability = 20 ÷ 100 = 0.2.

[0080] Step 203: Write all the conversion probabilities into an initial matrix structure of a preset dimension to form a gait phase transition matrix. The row dimension of the initial matrix structure corresponds to the initial motion state, and the column dimension corresponds to the target motion state.

[0081] In this step, the gait phase transition matrix refers to the set of transition probabilities between various motion states, presented in matrix form. Its row dimensions correspond to the initial motion state before the transition, and its column dimensions correspond to the target motion state after the transition, used to visually display the probability distribution of motion state transitions. The initial motion state refers to the initial motion state before the transition, including walking, running, and sudden stop states, and is the baseline state for calculating the transition probability. The target motion state refers to the final motion state after the transition, including walking, running, and sudden stop states, and is the result state after the transition probability calculation.

[0082] In this embodiment, a 3x3 initial matrix is ​​preset, where the row dimensions correspond to walking, running, and sudden stop states (i.e., the initial movement state) in sequence, and the column dimensions correspond to walking, running, and sudden stop states (i.e., the target movement state) in sequence. The calculated transition probabilities from walking to running are written into the first row and second column, the transition probabilities from walking to sudden stop are written into the first row and third column, the transition probabilities from running to walking are written into the second row and first column, the transition probabilities from running to sudden stop are written into the second row and third column, the transition probabilities from sudden stop to walking are written into the third row and first column, and the transition probabilities from sudden stop to running are written into the third row and second column. The probabilities of the remaining diagonal positions (where the initial and target states are the same) are set to 0, thus forming a gait phase transition matrix.

[0083] Step 204: Traverse the gait phase transition matrix and determine the termination time of the initial motion state corresponding to the transition probability exceeding the preset probability threshold as the motion mode switching point;

[0084] In this step, the preset probability threshold refers to a pre-set probability threshold used to determine whether a motion state needs to be switched. When the switching probability exceeds this threshold, the corresponding motion state is determined to need to be switched. Based on 100 successful switching experiments of a humanoid robot in typical scenarios (such as switching from walking to running on flat ground), the minimum effective switching probability (i.e., the lowest probability to ensure a successful switch) is taken as the benchmark and determined in combination with a safety redundancy coefficient (1.2). The typical value is set to 0.6 (i.e., when the switching probability ≥ 0.6, it is determined that a state switch needs to be triggered). The motion mode switching point refers to the moment when the initial motion state ends and is about to switch to the target motion state, determined based on the switching probability exceeding the preset probability threshold in the gait phase transition matrix, and is used to trigger subsequent motion mode switching operations.

[0085] In this embodiment, each transition probability in the gait phase transition matrix is ​​traversed row by row, and each transition probability is compared with a preset probability threshold. When a certain transition probability (such as the transition probability from walking state to running state) is greater than the preset probability threshold, the starting motion state (i.e. walking state) corresponding to the transition probability is found, and the moment when the first phase feature value of the starting motion state last appears in the gait phase data is determined as the motion mode switching point.

[0086] This application's embodiments overcome the limitations of relying on fixed thresholds to determine the switching timing, and can dynamically predict the motion mode switching needs based on real-time gait characteristics, thus solving the problem of mode transition lag. Based on high-probability transition events in the gait phase transfer matrix, the motion mode switching point is determined, making the switching timing more consistent with the robot's actual motion state. This lays a precise time reference for subsequent hysteresis compensation and joint coordinated motion, improving the accuracy and timeliness of multimodal motion switching.

[0087] This application provides a specific embodiment. Step 201, extracting the first phase feature value corresponding to the walking state from the gait phase data, specifically includes the following steps:

[0088] Step 211: Determine the data segments in the gait phase data whose movement speed is within the preset walking speed range as walking state data;

[0089] In this step, the preset walking speed range refers to a pre-defined speed interval used to distinguish walking from other motion states. It is determined based on the typical speed range of a humanoid robot walking, including a numerical range reflecting a speed that is neither lower than the minimum walking speed nor higher than the maximum walking speed, and is used to filter walking state data from the gait phase data. Walking state data refers to continuous data segments selected from the gait phase data whose movement speed falls within the preset walking speed range. This includes data reflecting joint rotation angles, movement time, and foot contact information during walking, and is used to extract characteristic parameters of the walking state.

[0090] In this embodiment, the robot's lower limbs are equipped with speed sensors to collect movement speed in real time. The collected movement speed is compared with a preset walking speed range, and all continuous data segments with movement speeds within that range are selected. These data segments contain information such as joint angles, movement time, and foot pressure during that time period, which together constitute walking status data.

[0091] Step 212: Collect the first rotation angle of each lower limb joint at each sampling time in the walking state data, and arrange all the first rotation angles to form a first joint angle sequence;

[0092] In this step, the first rotation angle refers to the real-time rotation angle of each joint of the lower limb (including the hip, knee, and ankle joints) collected by the joint angle sensor at each sampling moment in the walking state data. Based on the real-time detection of joint rotation by the sensor, it is used to construct a sequence reflecting the dynamic changes in joint angles. The first joint angle sequence refers to the sequence formed by arranging the first rotation angles at each sampling moment in the walking state data in chronological order, including hip joint angle subsequence, knee joint angle subsequence, and ankle joint angle subsequence, used to reflect the changing pattern of each joint angle over time during walking.

[0093] In this embodiment, walking status data is sampled at a preset sampling interval. At each sampling time, the real-time rotation angle (i.e., the first rotation angle) of the hip joint, knee joint, and ankle joint is collected by a joint angle sensor. The first rotation angles of the same lower limb joint are arranged sequentially according to the sampling time to form the angle subsequence of the lower limb joint. Then, the angle subsequences of each lower limb joint are combined into the first joint angle sequence.

[0094] Step 213: Record the time points of two adjacent foot strikes in the walking state data, calculate the first strike interval duration, and determine the first strike interval duration as the first movement cycle duration;

[0095] In this step, the first landing interval duration refers to the time difference between two adjacent foot landing times in the walking state data. It is calculated based on the landing time recorded by the foot pressure sensor and is used to characterize the duration of a complete gait in the walking state. The first movement cycle duration refers to the time length that reflects a complete gait cycle in the walking state, as determined by the first landing interval duration. It includes the time interval from one foot landing to the next foot landing and is used to characterize the gait periodicity of the walking state.

[0096] In this embodiment of the application, a pressure sensor installed on the foot detects the trigger signal of the foot contacting the ground. When the pressure value exceeds the preset contact threshold, the moment is recorded as the foot landing time point. Two adjacent landing time points are selected, and the difference between the latter time point and the former time point (i.e. the first landing interval duration) is calculated. This difference reflects the time to complete a complete gait in the walking state, so it is determined as the first movement cycle duration.

[0097] Step 214: In the walking state data, select the starting time points of the hip joint and knee joint movement to calculate the first starting time difference, and use the first starting time difference as the first adjacent joint movement time difference;

[0098] In this step, the movement initiation time point refers to the moment when a joint begins to move from a resting or initial position in the walking data. It is determined based on the moment when the rate of change of joint angle exceeds a preset initiation threshold, and includes the hip joint movement initiation time point and the knee joint movement initiation time point. This is used to calculate the time difference between adjacent joint movements. The first initiation time difference refers to the difference between the knee joint movement initiation time point and the hip joint movement initiation time point in the walking data. It is calculated based on the numerical values ​​of the two joint movement initiation time points and is used to reflect the time difference between the hip and knee joints initiating movement. The first adjacent joint movement time difference refers to the time parameter determined by the first initiation time difference, reflecting the synchronicity of hip and knee joint movements during walking, including the difference in the order of the two joint initiation times, and is used to characterize the degree of coordination between adjacent joint movements.

[0099] In this embodiment of the application, by analyzing the angle changes of the hip and knee joints in the first joint angle sequence, when the joint angle deviates from the initial position and the angle change rate exceeds the preset start threshold, the moment is recorded as the starting time point of the corresponding joint movement. The starting time points of the hip and knee joints are selected, and the difference between the starting time point of the knee joint movement and the starting time point of the hip joint movement (i.e., the first starting time difference) is calculated. This difference reflects the time difference between the start of movement of the two joints, and therefore it is used as the first adjacent joint movement time difference.

[0100] Step 215: Combine the first joint angle sequence, the first motion cycle duration, and the motion time difference of the first adjacent joints into the first phase feature value corresponding to the walking state;

[0101] In this embodiment of the application, the first joint angle sequence reflecting the dynamic change of joint angle, the first motion cycle duration reflecting the gait cycle, and the first adjacent joint motion time difference reflecting the synchronicity of joint motion are integrated into a feature set according to a preset format. This set completely represents the gait characteristics of the walking state, which is the first phase feature value corresponding to the walking state.

[0102] The embodiments of this application solve the problems of fuzzy motion state feature extraction and difficulty in accurately distinguishing different motion states; the first phase feature value provides reliable basic data for subsequent calculation of motion state transition probability and generation of gait phase transfer matrix, ensuring the accuracy of motion mode switching point judgment, and thus laying a precise state recognition foundation for seamless switching of multimodal motion.

[0103] For example, in a service scenario, a humanoid robot needs to deliver items while walking, with its preset walking speed range set at 0.5-1.5 m / s. The robot's movement speed is collected via a speed sensor, and continuous data segments within the 0.5-1.5 m / s range are selected as walking state data (including information such as joint angles, time, and foot pressure over a 5-minute period). Sampling intervals of 10 milliseconds are used to collect the first rotation angles of the hip joint (e.g., 85°, 87°...), knee joint (e.g., 60°, 62°...), and ankle joint (e.g., 30°, 32°...) at each moment in this walking state data, and these angles are arranged chronologically. A first joint angle sequence is formed; the foot pressure sensor detects that the two adjacent foot landing times are at 10 seconds and 12 seconds respectively, and the first landing interval is calculated to be 2 seconds, which is determined as the first movement cycle duration; the first joint angle sequence is analyzed and it is found that the hip joint starts to move at 10.2 seconds (the angle change rate exceeds the start threshold), and the knee joint starts to move at 10.3 seconds. The first start time difference is calculated to be 0.1 seconds, which is used as the first adjacent joint movement time difference; finally, the above first joint angle sequence, the 2-second first movement cycle duration, and the 0.1-second first adjacent joint movement time difference are combined to form the first phase feature value corresponding to the walking state.

[0104] This application provides a specific embodiment. Step 103 involves determining the timing of hysteresis compensation for the pneumatic artificial muscles based on the motion mode switching point, so as to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot to obtain the joint motion state. The specific steps include the following:

[0105] Step 301: Extract the extension length and internal pressure values ​​of the pneumatic artificial muscles of the lower limb joint within the first preset time period before the movement mode switching point and within the second preset time period after the movement mode switching point;

[0106] In this step, the first preset duration refers to a specific time period before the movement mode switching point, determined based on historical data of the hysteresis characteristics of the pneumatic artificial muscle. This includes a time interval for collecting pre-compensation state data to analyze the pressure-length relationship before the switch. The second preset duration refers to a specific time period after the movement mode switching point, shorter than the first preset duration. It is determined based on the dynamic response requirements after the movement mode switch, including a time interval for collecting initial state data to supplement the analysis of real-time changes in hysteresis characteristics. The extension / extension length value refers to the length data of the pneumatic artificial muscle during the extension / extension process, directly measured by a length sensor. It includes specific values ​​reflecting the degree of muscle contraction or extension, used to correlate internal pressure with the movement state. The internal pressure value refers to the air pressure data inside the pneumatic artificial muscle, collected in real-time by a pressure sensor. It includes values ​​reflecting the magnitude of the muscle driving force, used to analyze muscle hysteresis deviation.

[0107] In this embodiment, the motion mode switching point is taken as the time origin. The length extension data within a first preset time period (e.g., 1 second) before the switching point is collected by the length sensor installed on the pneumatic artificial muscle, and the internal pressure data at the same time is collected by the pressure sensor. At the same time, the length extension value and internal pressure value within a second preset time period (e.g., 0.5 seconds) after the motion mode switching point are collected. These data together serve as the basis for the analysis of hysteresis compensation.

[0108] Step 302: Under the same extension length value, calculate the difference between the internal pressure value and the corresponding standard pressure value in the standard pressure curve to obtain the hysteresis deviation value of the pneumatic artificial muscle.

[0109] In this step, the standard pressure curve refers to the curve showing the relationship between the extension length and internal pressure of the pneumatic artificial muscle under ideal conditions (i.e., no hysteresis). It is obtained based on factory calibration data or through fitting multiple experiments and includes standard pressure values ​​corresponding to different lengths, serving as a benchmark for calculating the hysteresis deviation value. The standard pressure value refers to the pressure data corresponding to a specific extension length in the standard pressure curve. It is obtained based on the standard pressure curve and includes pressure values ​​reflecting the ideal state at that length, used to compare with the actual internal pressure value to calculate the hysteresis deviation value. The hysteresis deviation value refers to the difference between the actual internal pressure value and the standard pressure value of the pneumatic artificial muscle at the same extension length, calculated by subtracting the two. It includes positive deviation (i.e., actual pressure higher than the standard) and negative deviation (i.e., actual pressure lower than the standard), used to characterize the degree of influence of hysteresis characteristics.

[0110] In this embodiment, the standard pressure curve is first obtained as follows: Under ideal conditions (temperature 25℃, humidity 50%), 50 repeated expansion and contraction experiments are performed on the pneumatic artificial muscle. In each experiment, 20 sampling points are evenly selected from the minimum expansion length (e.g., 2cm) to the maximum expansion length (e.g., 15cm), and the stable internal pressure (the value when the pressure fluctuation is ≤0.01MPa) at each length is recorded. The average pressure value of the 50 experiments is taken, and then the curve corresponding to the expansion length and the standard pressure (e.g., a quadratic function curve) is obtained by fitting the curve using the least squares method. The same expansion length value (e.g., 5cm, 10cm, etc.) is selected from the collected expansion length values, and the standard pressure value corresponding to the expansion length value is found in the standard pressure curve (the standard pressure curve is the curve corresponding to the expansion length value and the pressure value of the pneumatic artificial muscle under ideal conditions). Then, the internal pressure value at the same expansion length value is subtracted from the standard pressure value, and the difference is the hysteresis deviation value at that length. This process is repeated to obtain multiple hysteresis deviation values.

[0111] Step 303: Calculate the hysteresis compensation amount based on the time interval between the hysteresis deviation value and the motion mode switching point;

[0112] In this step, the hysteresis compensation amount refers to the pressure value that needs to be adjusted to offset the hysteresis deviation value. It is calculated based on the weighted average of the hysteresis deviation value and the time interval, including the pressure value that needs to be increased or decreased, and is used to guide the pressure adjustment of the pneumatic artificial muscle.

[0113] In this embodiment, the time interval between the acquisition time and the motion mode switching point corresponding to each hysteresis deviation value is recorded and set as t. The maximum range of the time interval is set to T (T is the sum of the first preset duration and the second preset duration). Based on the time decay characteristics of the hysteresis effect of the pneumatic artificial muscle, the weight coefficient corresponding to the time interval is determined. The influence of the hysteresis deviation on the joint movement under different time intervals (t) is measured experimentally (the greater the influence, the higher the weight). A linear weight model is fitted to determine the weight coefficient k = 1 - (t / T) to map the hysteresis deviation values ​​at different time points to the hysteresis compensation amount, thereby realizing the reverse correction of the hysteresis characteristics of the pneumatic artificial muscle. Wherein, when t=0, k=1; when t=T, k=0, that is, the closer the time is to the motion mode switching point, the larger the weight coefficient; for example, when t=0.3 seconds (close to the motion mode switching point), k=1-(0.3 / 1.5)=0.8; when t=1.5 seconds (far from the motion mode switching point), k=0, ensuring that the closer the hysteresis deviation value is to the motion mode switching point, the greater its impact on the hysteresis compensation amount; multiply each hysteresis deviation value by its corresponding weight coefficient k to obtain multiple weighted hysteresis deviation values, and then sum all the weighted hysteresis deviation values ​​to obtain the hysteresis compensation amount to be applied. The calculation formula can be expressed as: Hysteresis compensation amount = Σ(hysteresis deviation value × k).

[0114] Step 304: Using the motion mode switching point as the time reference, set the compensation start time to a preset advance time before the motion mode switching, and set the compensation end time to a preset duration after the motion mode switching point, so as to generate a hysteresis compensation timing.

[0115] In this step, the time reference refers to the origin of the time coordinate system with the motion mode switching point as the reference. It is used to uniformly calculate the start and end times of compensation, ensuring precise alignment between the compensation timing and the switching point. The compensation start time refers to the specific time at which hysteresis compensation begins, set as a preset advance time before the motion mode switching point. This is determined based on the response speed of the hysteresis characteristics and is used to initiate compensation in advance to offset the lag during switching. The preset advance time refers to the time difference between the compensation start time and the motion mode switching point, determined based on experimentally measured hysteresis compensation response time. It includes an advance amount to ensure the compensation effect takes effect during switching, used to achieve advanced compensation. The compensation end time refers to the specific time at which hysteresis compensation ends, set as a preset duration after the motion mode switching point. This is determined based on the need for dynamic stability after switching and is used to continue compensation until the motion state stabilizes. The preset duration refers to the time difference between the compensation end time and the motion mode switching point, shorter than the preset advance time. It is determined based on the stabilization time of joint movement after switching and is used to avoid overcompensation. The hysteresis compensation timing refers to the time period for hysteresis compensation, defined by the compensation start time and compensation end time, including the time interval covering the key dynamic stages before and after switching, used to implement pressure regulation during the optimal period.

[0116] In this embodiment of the application, the motion mode switching point is used as the time reference. The compensation start time is determined as a preset advance time (e.g., 0.2 seconds) before the switching point, and the compensation end time is determined as a preset duration (e.g., 0.3 seconds) after the switching point. The time period between the two (from 0.2 seconds before the switching point to 0.3 seconds after the switching point) is the hysteresis compensation timing.

[0117] Step 305: During the hysteresis compensation time, the internal pressure of the pneumatic artificial muscle is adjusted according to the hysteresis compensation amount, and the second rotation angle and rotation speed of the lower limb joint are collected during the adjustment process. The second rotation angle and rotation speed are combined into the joint motion state.

[0118] In this embodiment, during the hysteresis compensation time, the internal pressure of the pneumatic artificial muscle is increased or decreased by the pressure regulating device according to the hysteresis compensation amount. At the same time, the rotation angle (i.e., the second rotation angle) and rotation speed of the hip joint, knee joint and ankle joint are collected in real time by the joint angle sensor and speed sensor. The second rotation angle and rotation speed at the same moment are integrated according to the joint classification to form a joint motion state that reflects the real-time motion of the joint.

[0119] This application's embodiments solve the problem of joint movement and command deviation caused by the characteristic delay of the drive component; by presetting the advance time and duration to determine the timing of hysteresis compensation, it realizes advance compensation and dynamic correction during the switching process, ensuring the accuracy of joint movement state during mode switching, providing reliable motion data for subsequent generation of bio-inspired electrical signals and driving joint coordinated movement, and improving the smoothness and posture stability of humanoid robot movement mode switching.

[0120] This application provides a specific embodiment. Step 104 involves analyzing the angle change rate and motion acceleration parameters in the joint motion state to generate a bio-inspired electrical signal, specifically including the following steps:

[0121] Step 401: Based on the joint motion state, calculate the angle change rate and motion acceleration of each lower limb joint;

[0122] In this embodiment, the second rotation angle and rotation speed of each lower limb joint (including hip joint, knee joint, and ankle joint) are extracted from the joint motion state. The angle change rate of each joint is obtained by calculating the difference of the second rotation angle at adjacent moments and dividing it by the time interval. The motion acceleration of each joint is obtained by calculating the difference of the rotation speed at adjacent moments and dividing it by the time interval, thereby quantifying the dynamic characteristics of joint motion.

[0123] Step 402: Compare the angle change rates of different lower limb joints at the same time to calculate the angle change difference between adjacent lower limb joints, so as to determine the motion correlation between each lower limb joint;

[0124] In this step, the angle change difference refers to the difference in the rate of angle change of adjacent lower limb joints at the same moment. It is obtained by subtracting the rate of angle change of the hip joint from that of the knee joint, and the knee joint from that of the ankle joint. It includes a value reflecting the difference in movement speed between adjacent joints and is used to determine the degree of motion correlation between joints. The degree of motion correlation refers to a quantitative index reflecting the degree of joint motion coordination, determined based on the angle change difference of adjacent lower limb joints. It includes positive values ​​(indicating similar movement trends) and negative values ​​(indicating opposite movement trends) and is used to characterize the motion coordination relationship between joints.

[0125] In this embodiment, at the same time point, the difference in the angle change rate between the hip and knee joints (i.e., the hip joint angle change rate minus the knee joint angle change rate) and the difference in the angle change rate between the knee and ankle joints (i.e., the knee joint angle change rate minus the ankle joint angle change rate) are calculated to obtain the angle change difference between adjacent joints. The sign and magnitude of this difference reflect the degree of coordination of the movements of adjacent joints, that is, the degree of motion correlation between the lower limb joints. For example, if the angle change rate of the hip joint is 10° / s and that of the knee joint is 8° / s, then the motion correlation between the two is 2° / s (a positive value indicates similar movement trends); if the angle change rate of the knee joint is 12° / s and that of the ankle joint is 15° / s, then the motion correlation is -3° / s (a negative value indicates opposite movement trends).

[0126] Step 403: Multiply all motion correlation degrees with preset angle ratio coefficients to obtain the corresponding joint coordination angle thresholds;

[0127] In this step, the preset angle ratio coefficient refers to a fixed coefficient pre-set to convert motion correlation into an angle threshold. It is determined based on the biomechanical constraints of the human lower limb joints, specifically referencing the natural synergistic angle ratio of the hip-knee-ankle joints during walking and running (e.g., the angle linkage ratio between the hip and knee joints in the gait cycle). This is combined with the ratio of the upper limit of joint range of motion (e.g., the maximum hip flexion angle and the maximum knee extension angle) to the muscle safety force threshold (the angle corresponding to the critical force value that avoids excessive muscle stretching or contraction). This value is obtained by fitting multiple sets of human motion experimental data and is used to calculate the joint synergistic angle threshold. The joint synergistic angle threshold is the product of motion correlation and the preset angle ratio coefficient, including the maximum allowable angle deviation value during synergistic movement of each joint, used to limit the angle range of joint movement to ensure synergy.

[0128] In this embodiment, based on the biomechanical constraints of the human lower limb joints, a fixed angle conversion coefficient (i.e., a preset angle ratio coefficient) is preset. For example, referring to the hip-knee coordination angle ratio of approximately 1:0.8 in human kinematics research during walking, it is set to 0.6 after being corrected by the robot joint structure parameters. This coefficient is used to convert the motion correlation degree into the joint's executable angle threshold. Each motion correlation degree (such as the motion correlation degree between the hip joint and the knee joint, and the motion correlation degree between the knee joint and the ankle joint) is multiplied by the preset angle ratio coefficient, and the resulting product is the joint coordination angle threshold that must be followed when the corresponding joints coordinate to move, and this threshold does not exceed the safe range of joint biomechanical activity.

[0129] Step 404: Take the motion acceleration that appears most frequently under the same motion correlation degree as the reference acceleration, multiply all the reference accelerations by the preset time coefficient to obtain the time interval parameter of the corresponding motion trigger time, and combine it with the corresponding joint coordination angle threshold to form a signal feature group;

[0130] In this step, the reference acceleration refers to the most frequent motion acceleration under the same motion correlation, determined by statistically analyzing the frequency of acceleration occurrence. It includes acceleration values ​​reflecting typical motion intensity under that motion correlation and is used to calculate the time parameter for motion triggering. The preset time coefficient refers to a pre-set fixed coefficient used to convert the reference acceleration into a time interval. It is determined based on the joint motion response speed and includes conversion parameters reflecting the correspondence between acceleration and time, used to calculate the interval duration of motion triggering. The motion triggering moment refers to the time point at which each lower limb joint begins to move. It is determined based on the time interval parameter and the motion transmission sequence, including the specific time at which each lower limb joint initiates its movement, used to control the order of joint movements. The signal feature set refers to a feature set composed of joint coordination angle thresholds and the corresponding time interval parameters of the motion triggering moment. It is calculated based on the motion correlation and reference acceleration and includes angle and time information reflecting the coordinated movement of a single lower limb joint, used to construct a bio-inspired electrical signal.

[0131] In this embodiment, the statistical analysis of the baseline acceleration must meet the requirement of a data sample size of ≥50 sets: For each motion correlation degree, at least 50 motion acceleration data points under the same correlation degree are extracted from the joint motion state. If the frequency of a certain acceleration value is ≥30% (i.e., ≥15 times), it is determined as the baseline acceleration. For example, among the 50 acceleration data points corresponding to the motion correlation degree 2° / s, 1.5° / s appears 18 times (accounting for 36%), then 1.5° / s is the baseline acceleration, ensuring the reliability of the statistical results under small sample size; a preset time conversion coefficient (i.e., preset time coefficient) is used to multiply each baseline acceleration by this coefficient to obtain the time interval parameter of the motion trigger moment; then the joint coordination angle threshold corresponding to each motion correlation degree is combined with the time interval parameter to form a signal feature group containing angle and time information.

[0132] Step 405: Arrange all signal feature groups according to the motion transmission sequence of each lower limb joint to form a bio-inspired electrical signal, wherein the pulse amplitude of the bio-inspired electrical signal is the corresponding joint coordination angle threshold, and the pulse interval is the corresponding motion triggering time.

[0133] In this step, the motion transmission sequence refers to the order in which lower limb joints move, starting from the hip joint and proceeding sequentially to the knee and ankle joints. This sequence is determined based on the physiological laws of human movement and is used to regulate the initiation order of lower limb joint movements. The bio-inspired electrical signal refers to a pulse sequence formed by arranging signal feature groups according to the motion transmission sequence. Its pulse amplitude corresponds to the joint coordination angle threshold, and the pulse interval corresponds to the motion triggering time. It is used to drive each lower limb joint to move in a coordinated manner, simulating the control characteristics of bioelectrical signals.

[0134] In this embodiment, the signal feature groups corresponding to each lower limb joint are arranged sequentially according to the motion transmission order from the hip joint to the knee joint and then to the ankle joint. The joint coordination angle threshold in each signal feature group is used as the pulse amplitude, and the time interval parameter is used as the interval duration between adjacent pulses, thereby forming a continuous pulse sequence. This pulse sequence is the bio-inspired electrical signal used to drive joint movement.

[0135] The embodiments of this application solve the problems of poor joint coordination and the disconnect between motion commands and actual actions; the bio-inspired electrical signal integrates the joint coordination angle and trigger time into a pulse sequence, so that the movement of each joint can not only follow the preset angle range, but also start in a reasonable timing sequence, ensuring the accuracy and continuity of multi-joint coordinated movement, providing a reliable driving signal for achieving seamless switching of multimodal movement, and improving the movement flexibility and stability of humanoid robots in dynamic environments.

[0136] This application provides a specific embodiment. Step 105 involves driving the lower limb joints of the humanoid robot to perform coordinated movements based on the joint coordination angle threshold and motion triggering time of the bio-inspired electrical signal, in order to achieve seamless switching of multimodal motion. This specifically includes the following steps:

[0137] Step 501: Based on the movement transmission sequence of each lower limb joint, and taking the movement triggering time of the bio-inspired electrical signal as a reference, assign a corresponding action execution period to each lower limb joint. The action execution period is a preset action window before and after the movement triggering time.

[0138] In this step, the motion transmission sequence refers to the order in which lower limb joint movements are initiated, starting from the hip joint and proceeding sequentially to the knee and ankle joints. This sequence is based on the physiological laws of human movement and is used to regulate the timing of each lower limb joint's movements, ensuring coordinated motion. The motion execution period refers to the allocated time interval for each lower limb joint, based on the trigger time of the bio-inspired electrical signal. It includes preset motion windows before and after the trigger time, used to limit the time range of lower limb joint movements and avoid overlapping movements. The preset motion window refers to the specific time range of the motion execution period, consisting of a preset duration before the trigger time and a preset duration after the trigger time. It is set based on the joint's response speed and duration, ensuring that the lower limb joints have sufficient time to complete the movement.

[0139] In this embodiment, according to the motion transmission sequence of the hip joint, knee joint, and ankle joint, and based on the motion triggering time corresponding to each lower limb joint in the bio-inspired electrical signal, a motion execution period is defined for each lower limb joint. This period is a preset motion window consisting of a preset duration before the triggering time (e.g., 0.3 seconds) and a preset duration after the triggering time (e.g., 0.2 seconds), ensuring that each lower limb joint completes its motion within its respective motion execution period and avoiding motion conflicts.

[0140] Step 502: Calculate the rotation angle amplitude based on the joint coordination angle threshold of each lower limb joint and the corresponding transmission ratio parameter;

[0141] In this step, the transmission ratio parameter refers to the conversion coefficient between the joint coordination angle threshold and the actual rotation angle amplitude. It is determined based on the transmission characteristics of the drive mechanism (such as a pneumatic artificial muscle) and includes a value reflecting the correspondence between the input and output angles, used to convert the coordination angle threshold into an actual executable angle. The rotation angle amplitude refers to the actual angle the joint needs to rotate, calculated based on the product of the joint coordination angle threshold and the transmission ratio parameter. It includes a specific value reflecting the range of motion of the lower limb joint, used to define the range of joint movement.

[0142] In this embodiment of the application, for the hip joint, knee joint, and ankle joint, the corresponding joint coordination angle threshold (extracted from bio-inspired electrical signals) and transmission ratio parameter (preset conversion coefficient between joint angle and output angle of drive mechanism) are obtained respectively. The joint coordination angle threshold of each lower limb joint is multiplied by its corresponding transmission ratio parameter to obtain the actual angle that the joint needs to rotate, i.e., the rotation angle amplitude.

[0143] Step 503: Determine the rotation direction of each lower limb joint based on the motion correlation between each lower limb joint, wherein the rotation direction maintains a coordinated relationship with the motion direction of adjacent joints;

[0144] In this step, the coordination relationship refers to the coordination rules of the rotation direction of adjacent lower limb joints, including same-direction coordination (same direction) and opposite-direction coordination (opposite direction), which is determined based on the motion correlation between lower limb joints and is used to ensure the coordination of lower limb joint movements.

[0145] In this embodiment, the coordination relationship is determined by the motion correlation degree (positive or negative value) between each lower limb joint: if the motion correlation degree is positive, adjacent lower limb joints maintain a unidirectional coordination relationship (e.g., when the hip joint rotates forward, the knee joint also rotates forward); if the motion correlation degree is negative, adjacent lower limb joints maintain a reverse coordination relationship (e.g., when the hip joint rotates forward, the ankle joint rotates backward), thereby determining the rotation direction (forward or backward) of each lower limb joint.

[0146] Step 504: Combine the rotation angle amplitude and rotation direction corresponding to each lower limb joint into a target rotation angle command;

[0147] In this step, the target rotation angle command refers to a complete motion command that includes the rotation angle amplitude and rotation direction. It is formed based on the combination of angle and direction parameters of each lower limb joint and is used to directly drive joint movement.

[0148] In this embodiment of the application, for each lower limb joint, its rotation angle amplitude (e.g., 30° for the hip joint and 20° for the knee joint) and rotation direction (forward or backward) are integrated into a complete target rotation angle command. This command clarifies the angle and direction that the lower limb joint needs to rotate, serving as the direct basis for driving the movement of the lower limb joint.

[0149] Step 505: During the action execution period, drive the corresponding lower limb joints to move according to the target rotation angle command, so as to collect the third rotation angle of each lower limb joint;

[0150] In this step, the third rotation angle refers to the actual rotation angle of the lower limb joint during movement, which is obtained based on real-time detection.

[0151] In this embodiment, during the action execution period allocated to each lower limb joint, the movement of the lower limb joint is controlled by a drive mechanism (such as a pneumatic artificial muscle) according to the target rotation angle command. At the same time, the actual rotation angle (i.e., the third rotation angle) during the movement of the lower limb joint is collected in real time by an angle sensor for subsequent deviation calculation.

[0152] Step 506: Calculate the angle deviation value between the third rotation angle and the target rotation angle command, compare the angle deviation value with the preset deviation threshold, and drive all lower limb joints to perform coordinated movements according to the comparison result until the switching of different movement states is completed, so as to achieve seamless switching of multimodal movements;

[0153] In this step, the angle deviation value refers to the difference between the third rotation angle and the target rotation angle command, calculated by subtracting the two. It includes a value reflecting the degree of deviation between the actual movement and the target rotation angle command, used to judge the movement accuracy. The preset deviation threshold refers to the maximum critical value of the allowable angle deviation, set based on the movement accuracy requirements. It includes a reference value used to determine whether the movement command needs to be adjusted, ensuring that the joint movement error is within an acceptable range.

[0154] In this embodiment, the difference between the third rotation angle of each lower limb joint and the target rotation angle command (i.e., the angle deviation value) is calculated. If the angle deviation value is less than a preset deviation threshold (e.g., 2°), the current command is maintained and the movement continues. If the angle deviation value is greater than or equal to the preset deviation threshold, the rotation angle amplitude and rotation direction of adjacent lower limb joints are adjusted. The angle deviation value is reduced through collaborative compensation. This process is repeated until all lower limb joints complete the movement, thereby achieving seamless switching between different movement states.

[0155] This application's embodiments solve the problems of poor joint coordination and large deviations between motion commands and actual actions; by combining target rotation angle commands with dynamic deviation adjustments, it ensures that each lower limb joint moves according to a preset coordination relationship, reducing stuttering and instability during motion mode switching, realizing seamless switching of the humanoid robot between different states, and improving the stability, accuracy, and environmental adaptability of the motion.

[0156] This application provides a specific embodiment. Step 503 involves determining the rotation direction of each lower limb joint based on the motion correlation between the joints, wherein the rotation direction maintains a coordinated relationship with the motion direction of adjacent joints. This specifically includes the following steps:

[0157] Step 511: Set the coordination relationship between adjacent lower limb joints. The coordination relationship includes unidirectional coordination relationship and reverse coordination relationship. Specifically, the coordination relationship between adjacent lower limb joints with a motion correlation degree within a preset positive range is determined as a unidirectional coordination relationship, and the coordination relationship between adjacent lower limb joints with a motion correlation degree within a preset negative range is determined as a reverse coordination relationship.

[0158] In this step, the coordination relationship refers to the rules governing the rotational directions of adjacent lower limb joints during movement. Determined based on the numerical range of motion correlation, it includes unidirectional and antidirectional coordination relationships, used to regulate the directional coordination between lower limb joints to ensure motion coordination. Unidirectional coordination refers to the coordination relationship where adjacent lower limb joints rotate in the same direction. This relationship is established when the motion correlation of adjacent lower limb joints is within a preset positive range, including cases where the hip and knee joints rotate forward together, or the knee and ankle joints rotate forward together, ensuring consistent lower limb joint movement trends. Antidirectional coordination refers to the coordination relationship where adjacent lower limb joints rotate in opposite directions. This relationship is established when the motion correlation of adjacent lower limb joints is within a preset negative range, including cases where the knee joint rotates backward when the hip joint rotates forward, or the ankle joint rotates backward when the knee joint rotates forward, achieving complementary coordination of lower limb joint movements.

[0159] In this embodiment, the numerical range of motion correlation is predefined: the preset positive value range is the range greater than 0 and less than or equal to a certain positive number (such as 5), and the preset negative value range is the range less than 0 and greater than or equal to a certain negative number (such as -5); when the motion correlation of adjacent lower limb joints (such as hip joint and knee joint) falls within the preset positive value range, it is determined that the two need to maintain a cooperative relationship in the same direction (i.e., the rotation direction is the same); when the motion correlation falls within the preset negative value range, it is determined that the two need to maintain a cooperative relationship in opposite directions (i.e., the rotation direction is opposite), thereby clarifying the coordination rules of adjacent joints.

[0160] Step 512: Determine the rotation direction of each lower limb joint based on the motion correlation and coordination relationship between different adjacent lower limb joints, so that the rotation direction of each lower limb joint maintains a corresponding coordination relationship with the motion direction of the adjacent lower limb joints;

[0161] In this step, the preset positive value range refers to a pre-defined range of motion correlation values ​​used to determine whether adjacent lower limb joints have a unidirectional synergistic relationship. This range is determined based on historical data and motion characteristics of joint synergistic movements and includes values ​​greater than 0 but not exceeding a certain upper limit. This range is used to quantify the criteria for unidirectional synergy. The preset negative value range refers to a pre-defined range of motion correlation values ​​used to determine whether adjacent lower limb joints have a reversible synergistic relationship. This range is determined based on the requirements of complementary joint movements and includes values ​​less than 0 but not lower than a certain lower limit. This range is used to quantify the criteria for reversible synergy.

[0162] In this embodiment, the initial rotation direction (forward or backward) of the starting joint (such as the hip joint) is first determined; then the motion correlation between the hip joint and the knee joint is obtained. If it is within a preset positive range (corresponding to a unidirectional coordination relationship), the rotation direction of the knee joint is the same as that of the hip joint. If it is within a preset negative range (corresponding to a reverse coordination relationship), the rotation direction of the knee joint is opposite to that of the hip joint. Similarly, based on the motion correlation between the knee joint and the ankle joint and the corresponding coordination relationship, the rotation direction of the ankle joint is determined to ensure that the rotation directions of the ankle joint and the knee joint conform to the coordination relationship, so that the rotation directions of all lower limb joints are matched with those of adjacent joints.

[0163] This application's embodiments solve the problems of ambiguous joint rotation direction coordination and poor synergy; it provides a clear basis for the directional coordination of adjacent lower limb joints, avoids conflicting or asynchronous movement directions, ensures the coordination of multi-joint collaborative movements, provides precise directional control for seamless switching of multimodal movements, and improves the smoothness and posture stability of humanoid robot movements.

[0164] Figure 2 This application provides a schematic diagram of the structure of a multi-joint cooperative motion control system for a humanoid robot, as shown in the embodiment of the present application. Figure 2 As shown, the system includes:

[0165] The acquisition module 21 is used to acquire gait phase data of the humanoid robot in different motion states, including walking, running and sudden stop.

[0166] The generation module 22 is used to generate a gait phase transition matrix based on the gait phase data in order to determine the motion mode switching point;

[0167] The compensation module 23 is used to determine the timing of hysteresis compensation for pneumatic artificial muscles based on the motion mode switching point, so as to perform hysteresis compensation processing on the pneumatic artificial muscles of each lower limb joint of the humanoid robot to obtain the joint motion state.

[0168] The analysis module 24 is used to analyze the angle change rate and motion acceleration parameters in the joint motion state and generate a bio-inspired electrical signal;

[0169] The drive module 25 is used to drive the lower limb joints of the humanoid robot to perform coordinated movements based on the joint coordination angle threshold and movement trigger time of the bio-inspired electrical signal, so as to achieve seamless switching of multimodal movements.

[0170] This application provides a humanoid robot multi-joint cooperative motion control system to implement the aforementioned humanoid robot multi-joint cooperative motion control method. Therefore, the specific implementation of the humanoid robot multi-joint cooperative motion control system can be found in the embodiment section of the humanoid robot multi-joint cooperative motion control method above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0171] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the multi-joint cooperative motion control method for a humanoid robot as described above.

[0172] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-joint cooperative motion control method for a humanoid robot described above.

[0173] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0174] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the multi-joint cooperative motion control method for humanoid robots described above.

[0175] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0176] The foregoing has provided a detailed description of a multi-joint cooperative motion control method and system for a humanoid robot provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A multi-joint cooperative motion control method for a humanoid robot, characterized by, The method comprises the following steps: acquiring gait phase data of a humanoid robot in different motion states, the motion states including a walking state, a running state and an emergency stop state; generating a gait phase transition matrix according to the gait phase data to determine a motion mode switching point; determining a hysteresis compensation opportunity of a pneumatic artificial muscle according to the motion mode switching point to perform hysteresis compensation processing on the pneumatic artificial muscle of each lower limb joint of the humanoid robot to obtain a joint motion state; analyzing an angular rate and a motion acceleration parameter in the joint motion state to generate a bio-inspired electrical signal; driving each lower limb joint of the humanoid robot to perform coordinated motion according to a joint coordination angle threshold and a motion trigger time of the bio-inspired electrical signal to realize seamless switching of multi-modal motion; generating a gait phase transition matrix according to the gait phase data to determine a motion mode switching point, comprising: extracting a first phase characteristic value corresponding to the walking state, a second phase characteristic value corresponding to the running state and a third phase characteristic value corresponding to the emergency stop state from the gait phase data; counting the occurrence frequencies of the first phase characteristic value, the second phase characteristic value and the third phase characteristic value to calculate the conversion probability between each two motion states; writing all the conversion probabilities into an initial matrix structure of a preset dimension to form a gait phase transition matrix, wherein the row dimension of the initial matrix structure corresponds to a starting motion state and the column dimension corresponds to a target motion state; iterating through the gait phase transition matrix to determine the termination time of the starting motion state corresponding to the conversion probability exceeding a preset probability threshold as the motion mode switching point; analyzing the angular rate and the motion acceleration parameter in the joint motion state to generate a bio-inspired electrical signal, comprising: calculating the angular rate and the motion acceleration of each lower limb joint according to the joint motion state; comparing the angular rates of different lower limb joints at the same time to calculate the angular rate difference between adjacent lower limb joints to determine the motion correlation degree between each lower limb joint; performing product operation of all the motion correlation degrees and a preset angle proportion coefficient to obtain corresponding joint coordination angle thresholds; taking the motion acceleration with the highest occurrence frequency under the same motion correlation degree as a reference acceleration and performing product operation of all the reference accelerations and a preset time coefficient to obtain time interval parameters of the motion trigger time, and combining the corresponding joint coordination angle thresholds to form a signal feature group; arranging all the signal feature groups according to the motion transmission sequence of each lower limb joint to form a bio-inspired electrical signal, wherein the pulse amplitude of the bio-inspired electrical signal is the corresponding joint coordination angle threshold and the pulse interval is the corresponding motion trigger time; driving each lower limb joint of the humanoid robot to perform coordinated motion according to the joint coordination angle threshold and the motion trigger time of the bio-inspired electrical signal to realize seamless switching of multi-modal motion, comprising: According to the motion transmission sequence of each lower limb joint, a motion execution time period corresponding to each lower limb joint is allocated based on the motion trigger time of the bio-inspired electrical signal, and the motion execution time period is a preset motion window before and after the motion trigger time; Based on the joint coordination angle threshold of each lower limb joint and the corresponding transmission ratio parameter, the rotation angle amplitude is calculated; According to the motion correlation between each lower limb joint, the rotation direction of each lower limb joint is determined, and the rotation direction maintains a coordinated relationship with the motion direction of the adjacent joint; The rotation angle amplitude and the rotation direction of each lower limb joint are combined into a target rotation angle instruction; During the motion execution time period, the corresponding lower limb joint is driven to move according to the target rotation angle instruction to collect the third rotation angle of each lower limb joint; The angle deviation value of the third rotation angle and the target rotation angle instruction is calculated, and the angle deviation value is compared with a preset deviation threshold to drive all lower limb joints to move coordinately according to the comparison result until the switching of different motion states is completed to realize seamless switching of multi-modal motion.

2. The method of claim 1, wherein, From the gait phase data, a first phase feature value corresponding to a walking state is extracted, including: Data segments in the gait phase data with a motion speed within a preset walking speed range are determined as walking state data; The first rotation angle of each lower limb joint at each sampling time in the walking state data is collected, and all first rotation angles are arranged to form a first joint angle sequence; The time points of adjacent two foot landings in the walking state data are recorded, and a first landing interval duration is calculated, which is determined as a first motion cycle duration; In the walking state data, the motion starting time points of the hip joint and the knee joint are selected to calculate a first starting time difference, which is taken as a first adjacent joint motion time difference; The first joint angle sequence, the first motion cycle duration, and the first adjacent joint motion time difference are combined into a first phase feature value corresponding to the walking state.

3. The method of claim 1, wherein, According to the motion mode switching point, a hysteresis compensation opportunity is determined to perform hysteresis compensation processing on the pneumatic artificial muscle of each lower limb joint of the humanoid robot to obtain a joint motion state, including: The extension length value and the internal pressure value of the pneumatic artificial muscle of the lower limb joint within a first preset duration before the motion mode switching point and within a second preset duration after the motion mode switching point are extracted; Under the same extension length value, the internal pressure value and the corresponding standard pressure value in the standard pressure curve are calculated to obtain the hysteresis deviation value of the pneumatic artificial muscle; According to the time interval between the hysteresis deviation value and the motion mode switching point, a hysteresis compensation amount is calculated; Taking the motion mode switching point as a time reference, the compensation start time is set as a preset advance time before the motion mode switching, and the compensation end time is set as a preset duration after the motion mode switching point to generate a hysteresis compensation opportunity. In the hysteresis compensation opportunity, the internal pressure of the pneumatic artificial muscle is adjusted according to the hysteresis compensation amount, and the second rotation angle and rotation speed of the lower limb joint in the adjustment process are collected, and the second rotation angle and rotation speed are combined into a joint motion state.

4. The method of claim 1, wherein, According to the motion correlation degree between each lower limb joint, the rotation direction of each lower limb joint is determined, and the rotation direction and the motion direction of the adjacent joint maintain a cooperative relationship, including: The cooperative relationship between adjacent lower limb joints is set, and the cooperative relationship includes a same direction cooperative relationship and an opposite direction cooperative relationship, wherein the cooperative relationship between adjacent lower limb joints with a motion correlation degree in a preset positive value range is determined as the same direction cooperative relationship, and the cooperative relationship between adjacent lower limb joints with a motion correlation degree in a preset negative value range is determined as the opposite direction cooperative relationship; According to the motion correlation degree and the cooperative relationship between different adjacent lower limb joints, the rotation direction of each lower limb joint is determined, so that the rotation direction of each lower limb joint maintains a corresponding cooperative relationship with the motion direction of the adjacent lower limb joint.

5. A multi-joint cooperative motion control system of a humanoid robot for executing a multi-joint cooperative motion control method of a humanoid robot as claimed in claim 1, characterized by It includes: An acquisition module is configured to acquire gait phase data of a humanoid robot in different motion states, including a walking state, a running state, and an emergency stop state; A generation module is configured to generate a gait phase transition matrix according to the gait phase data to determine a motion mode switching point; A compensation module is configured to determine a hysteresis compensation opportunity of a pneumatic artificial muscle according to the motion mode switching point, to perform hysteresis compensation processing on the pneumatic artificial muscle of each lower limb joint of the humanoid robot, and obtain a joint motion state; An analysis module is configured to analyze the angle change rate and motion acceleration parameters in the joint motion state, and generate a biological heuristic electrical signal; A driving module is configured to drive each lower limb joint of the humanoid robot to perform cooperative motion according to a joint cooperative angle threshold and a motion triggering time of the biological heuristic electrical signal, to realize seamless switching of multi-modal motion.

6. An electronic device, comprising: It includes: A memory is configured to store a computer program; A processor is configured to execute the computer program to implement the steps of the multi-joint cooperative motion control method of the humanoid robot according to any one of claims 1 to 4.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the multi-joint cooperative motion control method of the humanoid robot according to any one of claims 1 to 4.

Images