Track generation method and system based on action intention recognition and medium

Abstract

The invention relates to a trajectory generation method and system based on action intention recognition and a medium. The trajectory generation method comprises the steps that a force value component of force applied by a rehabilitation patient for a recovery position in the rehabilitation exercise process is acquired; identifying a motion intention of the rehabilitation patient based on the strength value component; according to the motion intention, a tail end track of the rehabilitation type exoskeleton robot is generated, and each position in the tail end track is restrained and limited within a preset space range. According to the track generation method, the force exerting condition of the wrist of the patient is collected in real time through the force sensor, the action intention is accurately distinguished and recognized according to the force exerting condition, the problem of distinguishing the motion state of the mechanical arm through the force sensor at the tail end is solved, and the starting state and the stopping state are conveniently distinguished. And through position accumulation and adoption of spatial constraint, a suitable motion track is quickly generated to support and realize the motion, grabbing and obstacle avoidance tasks of the mechanical arm, and support is provided for the motion tasks of the rehabilitation type exoskeleton robot.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory generation, and in particular to motion trajectory generation of a mechanical arm suitable for a rehabilitation exoskeleton robot. A trajectory generation method, device, system and computer-readable medium based on action intention recognition are proposed. Background Art

[0002] Motion trajectory generation is a prerequisite for completing tasks such as robot arm movement, grasping, and obstacle avoidance. A good trajectory generation front end can provide good input for the accurate and timely execution of specific tasks at the back end and reduce the workload of back end optimization.

[0003] At present, the development of the classic collaborative robot arm SDK has been relatively complete. It can realize the mutual conversion of the collaborative arm end posture to each joint angle, that is, forward and inverse solution. For example, the posture angle of the collaborative arm end is expressed by Euler angle representation. When performing forward and inverse solution, the Euler angle is converted into a rotation matrix. When performing trajectory planning, the trajectory of the path point is generated through polynomial fitting function (such as fifth-order polynomial) and optimization solver.

[0004] However, the trajectory planning scheme of the existing collaborative robot is mainly designed to solve and plan the joint movement of the serial manipulator arm robot, and use the force sensor built into the joint motor to record the angular momentum of each joint during the dragging process in a dragging teaching manner. The optimization of the terminal motion trajectory of the rehabilitation exoskeleton robot is insufficient. The dragging method of the joint force sensor relied on by this type of trajectory planning cannot accurately evaluate the terminal input force of the exoskeleton robot after being worn (generated by the wearer under specific movement intentions), while patients are more concerned about the recovery level of the terminal force in the recovery position such as the hands. The existing trajectory planning method is not accurate enough in mapping the terminal force when the rehabilitation exoskeleton robot is used.

[0005] In addition, in the research on trajectory planning of collaborative robots, a robot-based modeling method was proposed, which combined with dynamic optimization algorithms, acceleration continuity and other methods to generate trajectory for the robot's end manipulator motion trajectory. However, this type of method also cannot focus on the mapping and evaluation of the end force in the recovery position such as the hands of rehabilitation patients. Summary of the invention

[0006] In view of the defects of the prior art, according to the first aspect of the present invention, a trajectory generation method based on motion intention recognition is proposed, which aims to achieve accurate recognition of motion intention and generate the terminal trajectory of the rehabilitation exoskeleton robot accordingly through real-time collection and analysis of the force conditions of the rehabilitation position of the rehabilitation patient.

[0007] As an optional example, a trajectory generation method based on action intention recognition includes the following steps: Obtain the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise; Based on the force value component, identifying the exercise intention of the rehabilitation patient; According to the motion intention, a terminal trajectory of the rehabilitation exoskeleton robot is generated, and each position in the terminal trajectory is constrained to be within a preset spatial range.

[0008] As an optional example, obtaining the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise includes: A three-dimensional force sensor is obtained to sense the force exerted by the rehabilitation patient on the recovery position in real time to obtain the force value components of the three axes of X, Y, and Z; wherein the three-dimensional force sensor is arranged at the recovery position of the rehabilitation patient.

[0009] As an optional example, the identifying the exercise intention of the rehabilitation patient based on the strength value component includes: According to the force value components, obtaining the statistical variation trend of the force value components in each axis direction; In response to a statistical change trend on any axis satisfying a preset condition, determining that there is force input on the axis; and Activate the motion intention flag.

[0010] Therefore, based on the collection of the force actually applied by the rehabilitation patient to the rehabilitation position, its movement intention is identified, and the patient's actual strength and intention are mapped accordingly. The rehabilitation exoskeleton robot moves according to the motion trajectory generated by the intention, assisting the rehabilitation patient in motor rehabilitation.

[0011] As an optional example, obtaining the statistical variation trend of the force value component in each axis direction according to the force value component includes: According to the obtained force magnitude components in the three axial directions of X, Y, and Z, the mean and variance of the force magnitude components in each axial direction are respectively obtained as the statistical variation trend.

[0012] As an optional example, in response to a statistical change trend on any axis satisfying a preset condition, determining that there is force input on the axis includes: In response to the mean and variance of the force value component on any axis reaching a preset stable state, it is determined that there is force input on the axis. The preset condition is met including: within a preset continuous sampling period, the change in the mean and the change in the variance are both within a preset threshold range, and the preset condition is determined to be met.

[0013] Therefore, by calculating the variance and mean in real time, the statistical trend of the three force value components can be measured. The mean of the data indicates the concentration level of the data, and the variance reflects its degree of dispersion. Therefore, when the mean is stable, the movement intention is identified, and it is determined that the rehabilitation patient wants to use this force input to control the movement of the rehabilitation exoskeleton robot. The movement intention flag is activated, and position accumulation and trajectory generation are performed accordingly. The remaining stages (i.e., unstable stages) are not activated and do not participate in trajectory generation. In addition, for each uniaxial force input, the stable and stable stage is taken as the movement intention flag as the activation stage to avoid possible jitter and ensure the smooth progress of rehabilitation movement.

[0014] As an optional example, generating the terminal trajectory of the rehabilitation exoskeleton robot according to the motion intention includes: Based on the activation of the motion intention flag, the preset search constraint rules are used to accumulate the motion positions, and the terminal trajectory of the rehabilitation exoskeleton robot is generated with the accumulated position sequence.

[0015] As an optional example, the motion position accumulation is configured to be performed in the following manner: For the force value component obtained on any axis, the force value component of the current position is multiplied by the preset step adjustment factor, and the result is added to the accumulated force value component on the axis as the current accumulated position.

[0016] Therefore, by limiting the spatial search under the preset constraint rules, it is ensured that the spatial search process will not exceed the limit range, and the search progress can be accelerated and the efficiency can be improved. The motion trajectory of the rehabilitation exoskeleton robot can be generated through the trajectory sequence formed by position accumulation.

[0017] As an optional example, by adjusting the size of the preset step length adjustment factor, motion trajectories corresponding to different ranges are generated to adapt to different motor rehabilitation action task requirements.

[0018] According to a second aspect of the present invention, a computer system is also provided, comprising: one or more processors, and A memory for storing instructions that can be operated; Wherein, when the instructions are executed by the one or more processors, the process of the aforementioned trajectory generation method based on action intention recognition is implemented.

[0019] According to a third aspect of the present invention, a computer-readable medium storing a computer program is also proposed, wherein the computer program includes instructions that can be executed by one or more processors, and when the instructions are executed by the one or more processors, the process of the aforementioned trajectory generation method based on action intention recognition is implemented.

[0020] In combination with the implementation of the above aspects, the trajectory generation method based on action intention recognition proposed in the present invention realizes efficient and accurate trajectory generation by combining motion intention recognition and differentiation, search space constraints and trajectory integral accumulation motion position. Compared with the existing trajectory generation system, the significant beneficial effects of the method proposed in the present invention are: (1) For the terminal trajectory of the rehabilitation exoskeleton robot, by accurately distinguishing the movement intention, it can accurately identify the user's movement intention, choose whether to generate the trajectory, and generate a suitable movement trajectory to support the development of robot arm movement, grasping and obstacle avoidance tasks. It solves the problem of the terminal force sensor distinguishing the movement state of the robot arm, facilitates the robot to distinguish the start and stop states, and can avoid possible jitters to ensure the smooth progress of rehabilitation movements; (2) It can generate safe and necessary trajectories by cutting unnecessary sampling space as needed through the constraints within the trajectory sampling space, which can reduce the amount of calculation and improve the efficiency of trajectory generation; (3) By flexibly adjusting the integral parameters, trajectories of different ranges and step lengths can be generated to meet different motion requirements.

[0021] It should be understood that all combinations of the aforementioned concepts and the additional concepts described in more detail below can be considered as part of the inventive subject matter of the present disclosure as long as such concepts are not mutually inconsistent. In addition, all combinations of the claimed subject matter are considered as part of the inventive subject matter of the present disclosure.

[0022] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are not intended to be drawn to scale. In the accompanying drawings, each identical or nearly identical component shown in each figure may be represented by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings.

[0024] Figure 1 4 is a flow chart of a method for generating a trajectory based on action intention recognition according to an embodiment of the present invention.

[0025] Figure 2 4 is a flow chart of identifying the exercise intention of a rehabilitation patient based on the strength value component according to an embodiment of the present invention.

[0026] Figure 3 is a schematic diagram of a three-dimensional force sensor according to an embodiment of the present invention.

[0027] Figure 4 4 is a schematic diagram of an execution process of a trajectory generation method based on action intention recognition according to an embodiment of the present invention.

[0028] Figure 5 It is a mean value curve diagram of force value component data based on a three-dimensional force sensor according to an embodiment of the present invention.

[0029] Figure 6 4 is a graph showing a variance of force component data based on a three-dimensional force sensor according to an embodiment of the present invention.

[0030] Figure 7 This is a diagram of the result of motion intention recognition based on the force value component data of a three-dimensional force sensor according to an embodiment of the present invention.

[0031] Figure 8 Schematic diagram of a spherical constrained sampling space according to an embodiment of the present invention.

[0032] Fig. 9 Schematic diagram of a cube-constrained sampling space according to an embodiment of the present invention.

[0033] Fig.10 Schematic diagram of a trajectory generated under spherical constraints according to a method in an embodiment of the present invention.

[0034] Fig.11 It is a schematic diagram of a trajectory generated under cube constraints according to the method of an embodiment of the present invention.

[0035] Fig.12 is a schematic diagram of a trajectory generated without constraints according to the method of an embodiment of the present invention.

[0036] Fig.13 It is a test comparison chart of the method according to the embodiment of the present invention and the existing method. DETAILED DESCRIPTION

[0037] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

[0038] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone or in any appropriate combination with other aspects disclosed by the present invention.

[0039] {Trajectory generation method based on action intention recognition} According to the embodiments disclosed in the present invention, a trajectory generation method based on motion intention recognition is proposed, which is particularly suitable for motion trajectory generation of a mechanical arm of a rehabilitation exoskeleton robot. The method aims to collect the force applied to a rehabilitation patient by a three-dimensional force sensor and recognize the motion intention, and on this basis, to achieve an efficient and fast trajectory generation method, thereby providing good support for the motion tasks of the rehabilitation exoskeleton robot, meeting the motion trajectory planning requirements of different types of rehabilitation exoskeleton robots, and realizing the function of "move as soon as you put your hands on it, stop as soon as you let go".

[0040] As an example, Figure 1 The trajectory generation method based on action intention recognition of the exemplary embodiment shown includes the following steps: Step S101: obtaining the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise; Step S102: identifying the exercise intention of the rehabilitation patient based on the force value component; Step S103: generating a terminal trajectory of the rehabilitation exoskeleton robot according to the movement intention, wherein each position in the terminal trajectory is constrained to be within a preset spatial range.

[0041] As an optional embodiment, in the aforementioned step S101, obtaining the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise includes: A three-dimensional force sensor is obtained to sense the force exerted by the rehabilitation patient on the recovery position in real time to obtain the force value components of the three axes of X, Y, and Z; wherein the three-dimensional force sensor is arranged at the recovery position of the rehabilitation patient.

[0042] It should be understood that Figure 3In the example shown, the aforementioned three-dimensional force sensor is configured as a sensor device capable of real-time decomposition force detection in the three directions of X, Y, and Z. The three-dimensional force sensor is installed in a suitable form at a predetermined position of the rehabilitation patient, such as the wrist, ankle, and other joint positions. After calibration and initialization, it can continuously collect the force applied by the rehabilitation patient to the rehabilitation position in real time (i.e., indicating the rehabilitation patient's intention of motor rehabilitation training), and obtain the force value decomposition data in the three-axis directions of X, Y, and Z, i.e., the force value component on each axis.

[0043] For ease of explanation, wrist rehabilitation training is used as an example in the following embodiments. Through the three-dimensional force sensor, the patient's wrist force is collected to obtain the components on the X, Y, and Z axes, and the user's movement intention is accurately identified. The problem of the end force sensor distinguishing the movement state of the robotic arm is solved, which facilitates the rehabilitation exoskeleton robot to distinguish between the start and stop states, and can realize the movement intention distinction at any hovering angle, which can easily realize the "move when hands are put on, stop when hands are taken off" function.

[0044] In an embodiment of the present invention, the resultant force vector of the three-dimensional force sensor is represented by components on the X, Y, and Z axes, and the components of the force magnitude on each axis are represented in the form of a matrix (in N), that is:

[0045] in, , , Respectively represent the force value components on the X, Y, and Z axes.

[0046] Combination Figure 2 In the example shown, in the aforementioned step S102, based on the strength value component, identifying the exercise intention of the rehabilitation patient includes: Step S201: according to the force value component, obtaining the statistical variation trend of the force value component in each axis direction; Step S202: in response to the statistical change trend on any axis satisfying a preset condition, determining that there is force input on the axis; Step S203: Activate the motion intention flag.

[0047] In an embodiment of the present invention, the mean and variance of the force magnitude component are used to judge and distinguish the exercise intention of the rehabilitation patient.

[0048] As an optional embodiment, the above-mentioned obtaining the statistical variation trend of the force value component in each axis direction according to the force value component includes: According to the obtained force magnitude components in the three axial directions of X, Y and Z, the mean and variance of the force magnitude components in each axial direction are respectively obtained as the criterion of the statistical change trend.

[0049] As an optional embodiment, in response to a statistical change trend on any axis satisfying a preset condition, determining that there is force input on the axis includes: In response to the mean and variance of the force value component on any axis reaching a preset stable state, it is determined that there is force input on the axis.

[0050] As an optional example, the aforementioned preset conditions being met include: In a preset continuous sampling period, if the change in the mean and the change in the variance are both within a preset threshold range, it is determined that the preset condition is met.

[0051] Therefore, through the variance and mean, the statistical change trend of the three force value components can be measured, among which the mean of the data indicates the concentration level of the data, and the variance reflects its degree of dispersion.

[0052] On this basis, through the continuous acquisition and preservation of the force value component data of the three axes, and the real-time calculation of the mean and variance of the force value component data on the three axes, when the mean and variance are in a relatively stable state, it is determined that the rehabilitation patient wishes to use this force input to control the movement of the robot and conduct motor rehabilitation training. Based on this, the movement intention flag is activated, and the movement intention flag is not activated in other stages.

[0053] Combination Figure 1 As shown, as an optional embodiment, the aforementioned generation of the terminal trajectory of the rehabilitation exoskeleton robot according to the movement intention includes: Based on the activation of the motion intention flag, the preset search constraint rules are used to accumulate the motion positions, and the terminal trajectory of the rehabilitation exoskeleton robot is generated with the accumulated position sequence.

[0054] In the embodiment of the present invention, only when the movement intention flag is activated can the movement position be accumulated to form a position sequence for guiding the movement of the robot and helping rehabilitation patients to conduct rehabilitation training on rehabilitation positions, such as the wrist.

[0055] As an optional example, the aforementioned motion position accumulation is configured to be performed in the following manner: For the force value component obtained on any axis, the force value component of the current position is multiplied by the preset step adjustment factor, and the result is added to the accumulated force value component on the axis as the current accumulated position.

[0056] In an embodiment of the present invention, a search space that is too large may affect the calculation speed of the integrator. In other scenarios, there are requirements for the safety range of the trajectory, so it is necessary to limit the range of the search space. As an example, the spatial search of the present invention uses the following three forms of constraint rules to limit the trajectory space, namely: unconstrained form, spherical constraint form, and cubic constraint form.

[0057] like Figure 8 , 9 The figures show examples of sampling spaces for spherical constraint form and cube constraint form respectively.

[0058] Therefore, in the trajectory generation process of the embodiment of the present invention, based on the constraint method within the trajectory sampling space, unnecessary sampling space can be trimmed as needed to generate safe and necessary trajectories, and the efficiency of spatial search and trajectory generation can be improved, so that suitable motion trajectories can be quickly generated to support the execution of various rehabilitation training action tasks of the robot's mechanical arm, including but not limited to: swinging, grasping and obstacle avoidance.

[0059] In an optional embodiment, the aforementioned step size adjustment factor can be pre-set and adjusted as an integral parameter to adjust the size of a single step size, thereby achieving generation of trajectories in different large, medium and small ranges.

[0060] In some embodiments, by adjusting the aforementioned integral parameter, i.e., the step length adjustment factor, trajectories of different ranges can be generated to meet different rehabilitation exercise action requirements. Therefore, in the trajectory generation method proposed in the present invention, by flexibly adjusting the integral parameter, more flexible and diversified trajectory generation can be achieved to meet different rehabilitation training exercise task requirements.

[0061] {Example 1} In order to more specifically explain the implementation of the trajectory generation method of the present invention, we combine the attached Figure 3-12 The exemplary implementation of the aforementioned trajectory generation method is further described. In the following embodiments, the wrist movement recovery of the upper limb of a rehabilitation patient is taken as an example, and combined with Figure 4 A specific process of trajectory generation is described below.

[0062] The implementation process of the trajectory generation method of this embodiment is configured to be implemented through four parts, namely, obtaining the strength value component of the rehabilitation patient, judging the movement intention, searching for space constraints, and generating trajectories based on space constraints. Figure 4 as well as Figure 5-12 The figure illustrates the implementation of the above four parts in more detail.

[0063] 1.1 Wrist force analysis and triaxial force component acquisition In this example, a three-dimensional force sensor is deployed on the wrist of a rehabilitation patient to collect the patient's wrist force according to a preset sampling period and output the force value decomposition, i.e., component force data.

[0064] The following matrix represents the matrix expression of a resultant force vector acquired by a three-dimensional force sensor:

[0065] This means that the force components on the X, Y, and Z axes are 3.12N, -0.4N, and 5.12N respectively.

[0066] On this basis, the modulus of the resultant force vector is obtained, which is expressed as: .

[0067] Thus, the angle between the resultant force vector and the X, Y, and Z axes is obtained. , , (in radians), respectively:

[0068] Therefore, according to the aforementioned angle , , , the direction and quadrant of the resultant force vector can be determined, and the magnitude of the resultant force can be determined based on the modulus of the resultant force vector.

[0069] In an optional embodiment, the force component acquisition module may be configured as a data interface to continuously obtain component data collected by the sensor.

[0070] 1.2 Movement Intention Judgment In this example, the motion intention is judged and distinguished based on the mean and variance of the force value components of the X, Y, and Z axes obtained by continuous acquisition.

[0071] As an optional example, the mean of the sampled data on three axes is calculated as follows:

[0072] in, , , Represents the mean of the force components of the X, Y, and Z axes, respectively. n Indicates the sampling frequency of the 3D force sensor, that is, the number of samples of the component data on each axis.

[0073] , , Represents the force value components of the X, Y, and Z axes respectively. iThe sub-sampling value, that is, the observed value.

[0074] As an optional example, the variance of the sampled data on three axes is calculated as follows:

[0075] in, , , Represents the variance of the force value components of the X, Y, and Z axes respectively.

[0076] Therefore, by calculating the variance and mean, the statistical change trend of the three-axis component data can be obtained. The mean of the data represents the concentration level of the data, and the variance reflects its degree of dispersion.

[0077] In this example, by continuously obtaining the collected three-axis component data and calculating the mean and variance on the three axes in real time, when the mean and variance are stable, for example, the change amount or rate of change is within the allowable threshold range, it is determined that the rehabilitation patient wishes to use this force input to control the movement of the robot's mechanical arm, and the movement intention flag is activated, while the other stages are not activated.

[0078] As an optional method, the real-time calculation of the mean and variance and the judgment of the movement intention can be realized by configuring a movement intention judgment module. For example, the movement intention judgment module realizes the judgment and distinction of the movement intention by maintaining the force value component data container of the three axes and the movement intention flag, and outputs the judgment result.

[0079] Among them, the three-axis force value component data container is set to a data structure in matrix form to save the adopted data of the three-dimensional force sensor in real time, and calculate the mean and variance of the three-axis component data in the container in real time. When the mean and variance are stable, the motion intention flag is activated and output to participate in trajectory generation, and it is not activated in the other stages (that is, it does not participate in trajectory generation).

[0080] Combination Figure 5 , 6 As shown in ,7, the process of judging the motion intention is explained by taking the X-axis as an example.

[0081] When the mean value of the X-axis sampling data of the three-dimensional force sensor exceeds a preset threshold, it can be considered that a certain force input occurs in the axial direction of the sensor. Figure 5 It exemplifies the process of the mean value rising, stabilizing and then releasing under the input of force. The mean value will experience four stages: sudden rise (10 to 12 stages, 50 to 54 stages), fluctuation (12 to 20 stages, 52 to 62 stages), stabilization (20 to 30 stages, 60 to 70 stages) and release to zero (30 to 50 stages, after 70 stages).

[0082] Combination Figure 6 As shown in , compared with the change process of the mean, the variance also experiences four processes: first a sudden rise, then fluctuation and then stability within a certain range, then a decline and then stability within a range, and finally zero. Figure 6 The data changes during the same period are shown as an example.

[0083] Combination Figure 7 As shown, for each uniaxial force input process, in the embodiment of the present invention, only a relatively stable and steady stage is taken as the motion intention flag as the activation stage to avoid possible partial jitter, such as Figure 7 The data changes during the same period are shown as an example.

[0084] Combination Figure 5 , 6 As shown in Figure 7, in an embodiment of the present invention, motion intention is distinguished by using variance and mean for synchronous comparison. By analyzing the variance value characteristics and mean characteristics of the user, the motion intention can be distinguished according to the stage calculation results to accurately identify the user's action intention and provide accurate input and stop states for subsequent trajectory generation.

[0085] 1.3 Search Space Constraints A search space that is too large may affect the subsequent calculation speed of the integrator used for position accumulation. At the same time, in order to limit the safety range of the trajectory in the training and rehabilitation scenario, in an embodiment of the present invention, the range of the search space can be constrained according to preset rules.

[0086] As an optional implementation mode, the following three forms of constraint methods are used in the examples of the present invention to restrict the trajectory space, specifically including: a spherical constraint form, a cubic constraint form and an unconstrained form.

[0087] 1.3.1 Spherical Constraint First, the cumulative values ​​on the three axes ( , , ) By accumulating the input values ​​of the force components of each axis at the current moment, the radius of the sphere radius The assignment is preconfigured.

[0088] The min and max functions are used in the position accumulation update process to ensure that the accumulated value is within - radius arrive radius within the scope, that is: .

[0089] Then, by calculating the three-axis Euclidean distance distance to measure the length of the cumulative vector.

[0090] As an example, the aforementioned Euclidean distance is calculated using the square root of the sum of the squares of the values ​​on the three coordinate axes, namely:

[0091] If the distance exceeds radius , a scaling operation is required to ensure that the excess part is within the limit boundary and the non-exceeding part is within the limit boundary.

[0092] The scaling factor Scale_factor The calculation method is: radius Divide by distance The accumulated value is then scaled by multiplying it by a scaling factor to ensure that the length of the accumulated vector does not exceed the radius of the sphere. radius .

[0093] Therefore, through the scaling operation, all points beyond the constraint range are restricted to the boundary of the constraint range, and the previous search direction is maintained.

[0094] like Figure 8 It shows an example of using a normal distribution random dotting method to dot the space, and constraining the final result to scale the distribution of each random point in the space. It can be seen that all randomly distributed points are constrained to the inside of the sphere, and the space constraint is successful.

[0095] As an optional example, the process of spherical constraint includes: 1) Get the cumulative value on the three axes , , : 2) Get the three-axis Euclidean distance distance ; 3) Determine the Euclidean distance of the three axes distance Is it greater than the radius of the sphere? radius , if it is larger than the sphere radius, calculate the scaling factor Scale_factor And based on the scaling factor Scale_factor Scale the accumulated values ​​of the three axes to ensure that the length of the accumulated vector does not exceed the radius of the sphere radius , the accumulated positions are constrained to be within the spherical space.

[0096] 1.3.2 Cube Space Constraints Based on the aforementioned spherical constraint, for the cube constraint, there is no need for the scaling operation of the sphere constraint. You only need to keep the first constraint operation of the sphere constraint to limit it to the corresponding range.

[0097] As an optional embodiment, the min and max functions are used in the position accumulation update process to ensure that the accumulated value is within - length arrive length within the scope, that is:

[0098] in, length Represents the length of the semi-axis of the cube.

[0099] Fig. 9 It shows an example of using a normal distribution random dotting method to dot the space, and constraining the distribution of random points in the space after scaling the final result. It can be seen that all randomly distributed points are constrained to the internal range of the cube, and the space constraint is successful.

[0100] 1.3.3 Unconstrained form The unconstrained search space is a simple update operation on the cumulative value. The cumulative value on each axis is updated by summing the force input values ​​of each axis at the current moment.

[0101] The unconstrained form of position update is to add the coordinate value of the current point directly to the accumulated value without any range restriction, which means that the accumulated value can exceed the range of the cube or the sphere.

[0102] In the embodiments of the present invention, the operation can be accelerated by considering the influence of different search spaces, such as selecting a spherical constraint method or a cubic constraint method. By constraining the search space, the amount of calculation can be reduced and the efficiency of trajectory generation can be improved. By reasonably designing the constraints of the search space, the operation speed can be improved while ensuring accuracy.

[0103] As an optional approach, the search space constraint module is configured to implement spatial constraints based on predetermined rules, including but not limited to spherical constraints, cubic constraints or unconstrained rules.

[0104] 1.4 Trajectory Generation In this example, the trajectory generator is configured to generate the motion trajectory of the rehabilitation robot arm. Whether the trajectory generator is started is executed according to the result of the motion intention judgment module, that is, if there is a motion intention, the generation is started, otherwise the generation is not started.

[0105] In this example, when the trajectory generator responds to the generation of motion intention (i.e., the motion intention flag is activated), the search space constraint module is called to restrict the search space according to preset rules to ensure that the spatial search process does not exceed the restricted range and to accelerate the search.

[0106] After the search space constraint module completes the spatial constraints, the specific trajectory generation work is completed by a position accumulator, which is used to calculate the accumulated position and limit and scale it.

[0107] The position accumulator is internally configured with a position sequence container, which is used to maintain and update the accumulated position sequence of the three axes in real time.

[0108] Taking the X-axis as an example, in one calculation, the X-axis force value component of the current position is multiplied by the step adjustment factor and added to the accumulated X-axis component. Then check whether the three-axis Euclidean distance of the accumulated position exceeds the limit range. If it exceeds the limit range, the corresponding scaling and constraints are performed.

[0109] In this example, a sphere constraint is taken as an example, the sphere radius is preconfigured, and the cumulative position is scaled to the limit range of the sphere space by calculating the scaling factor.

[0110] At the same time, each component of the accumulated position is multiplied by the scaling factor to achieve overall scaling.

[0111] According to the above method, the position accumulator processes the force value components on the three axes obtained continuously, and the maximum integration result thereof will not exceed the limit position of the scaling.

[0112] like Fig.10 As shown, an example of a trajectory generated using a spherical constraint rule is exemplarily shown.

[0113] In other examples, we use cube constraints and unconstrained rules to combine motion intentions to generate trajectories. The resulting trajectories are as follows: Fig.11 and 12 shown.

[0114] In an optional embodiment, the aforementioned step size adjustment factor can be pre-set and adjusted as an integral parameter to adjust the size of a single step size, thereby achieving generation of trajectories in different large, medium and small ranges.

[0115] In some embodiments, by adjusting the aforementioned integral parameter, i.e., the step length adjustment factor, trajectories of different ranges can be generated to meet different rehabilitation exercise action requirements. Therefore, in the trajectory generation method proposed in the present invention, by flexibly adjusting the integral parameter, more flexible and diversified trajectory generation can be achieved to meet different rehabilitation training exercise task requirements.

[0116] 1.5 Test Data Comparison Combination Fig.13As shown, we compare the results obtained by using the trajectory generation method based on action intention of the above embodiment of the present invention with the trajectory generation method based on minimal jerk dynamics optimization in the prior art. The two methods are used to generate 70 data points respectively. The trajectory generation method based on minimal jerk uses a 5th-order polynomial to calculate the parameter matrix of 70 path points. The method of the above embodiment of the present invention is used to calculate the generated position points of 70 path points. By recording the time consumption of 5 groups of data, combined with Fig.13 It can be seen that the time spent on path calculation by the method of the present invention is greatly improved, and the time spent is shortened by more than 95%.

[0117] {Example 2} In combination with the implementation of the trajectory generation method based on action intention recognition in the above embodiment, according to the embodiment disclosed in the present invention, a computer system is also proposed, including: one or more processors, and a memory for storing operable instructions.

[0118] When the instructions are executed by the one or more processors, the process of the trajectory generation method based on action intention recognition of any of the aforementioned embodiments is implemented.

[0119] {Example 3} In combination with the implementation of the trajectory generation method based on action intention recognition in the above embodiments, according to the embodiments disclosed in the present invention, a computer-readable medium storing a computer program is also proposed, wherein the computer program includes instructions that can be executed by one or more processors, and when the instructions are executed by the one or more processors, the process of the trajectory generation method based on action intention recognition in any of the aforementioned embodiments is implemented.

[0120] As an optional example, the aforementioned computer-readable medium may be implemented using, but is not limited to, random access memory, read-only memory, electrically erasable programmable memory, optical disk memory, magnetic disk memory, and a combination of the aforementioned types of computer-readable storage media.

[0121] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A trajectory generation method based on action intention recognition, characterized in that: The following steps are involved: Obtain the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise; Based on the force value component, the movement intention of the rehabilitation patient is identified, specifically including: obtaining the statistical change trend of the force value component in the three-axis directions of X, Y, and Z according to the force value component; in response to the statistical change trend on any axis meeting the preset condition, determining that there is force input on the axis, that is, determining that the rehabilitation patient wants to use this force input to control the movement of the robot at this time, and activating the movement intention flag accordingly; According to the motion intention, the terminal trajectory of the rehabilitation exoskeleton robot is generated, specifically including: based on the activation of the motion intention flag, using the preset search constraint rules to accumulate the motion positions, and generating the terminal trajectory of the rehabilitation exoskeleton robot with the accumulated position sequence; each position in the terminal trajectory is constrained to be within a preset spatial range.

2. The trajectory generation method based on action intention recognition according to claim 1 is characterized in that: The step of obtaining the force value component of the rehabilitation patient's force applied to the recovery position during the rehabilitation exercise includes: The three-dimensional force sensor is used to sense the force exerted by the rehabilitation patient on the recovery position in real time, and the force value components of the X, Y, and Z axes are obtained; Wherein, the three-dimensional force sensor is arranged at the recovery position of the rehabilitation patient.

3. The trajectory generation method based on action intention recognition according to claim 1 is characterized in that: The step of obtaining the statistical variation trend of the force value component in each axis direction according to the force value component includes: According to the obtained force magnitude components in the three axial directions of X, Y, and Z, the mean and variance of the force magnitude components in each axial direction are respectively obtained as the statistical variation trend.

4. The trajectory generation method based on action intention recognition according to claim 3 is characterized in that: In response to the statistical change trend on any axis satisfying a preset condition, determining that there is force input on the axis includes: In response to the mean and variance of the force value component on any axis reaching a preset stable state, it is determined that there is force input on the axis.

5. The trajectory generation method based on action intention recognition according to claim 3 is characterized in that: The preset conditions are met as follows: In a preset continuous sampling period, if the change in the mean and the change in the variance are both within a preset threshold range, it is determined that the preset condition is met.

6. The trajectory generation method based on action intention recognition according to claim 1 is characterized in that: The motion position accumulation is configured to be performed in the following manner: For the force value component obtained on any axis, the force value component of the current position is multiplied by the preset step adjustment factor, and the result is added to the accumulated force value component on the axis as the current accumulated position.

7. The trajectory generation method based on action intention recognition according to claim 6 is characterized in that: The method further comprises the following steps: The size of the preset step size adjustment factor is adjusted to generate motion trajectories corresponding to different ranges.

8. A computer system, characterized in that: include: one or more processors and memory for storing instructions that can be operated; Wherein, when the instruction is executed by the one or more processors, the process of the trajectory generation method based on action intention recognition described in any one of claims 1-7 is implemented.

9. A computer-readable medium storing a computer program, characterized in that: The computer program includes instructions that can be executed by one or more processors, and when the instructions are executed by the one or more processors, the process of the trajectory generation method based on action intention recognition described in any one of claims 1-7 is implemented.

Images