Human gait simulation method, device, equipment, medium and product
By using muscle space-based predictive gait simulation technology, realistic simulated gait is generated using musculoskeletal models and reflection loops. This solves the problem of gait dynamics simulation under complex terrain, realizes stable simulated gait under multiple terrains, and improves simulation effect and equipment performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-14
AI Technical Summary
Existing human gait simulation technologies cannot achieve effective gait dynamics simulation in complex terrain, especially in situations involving going up and down stairs and complex terrain, where realistic gait control is not possible.
Predictive gait simulation technology based on muscle space is adopted. The current gait state is determined by the motion sensing information of the musculoskeletal model, the reflection loop is dynamically matched, and muscle excitation signals that conform to physiological laws are generated to drive the musculoskeletal model to form a stable and realistic simulated gait on complex terrain.
It improves the adaptability and realism of human gait simulation, can generate realistic simulated gait in complex terrain, reduces the implementation complexity of multi-terrain gait simulation, and enhances the performance of auxiliary equipment and task generalization ability.
Smart Images
Figure CN122376397A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of biomechanics technology, and in particular to a method, device, equipment, medium and product for simulating human gait. Background Technology
[0002] Human gait simulation can be applied to wearable robots (such as assistive exoskeletons, intelligent powered prostheses, human-computer interaction, etc.) and biomechanical analysis, etc., to verify the system in a simulation environment, which is a low-cost and safe way to verify the system.
[0003] Existing human gait simulation technologies based on muscle space models have limitations in generating anthropomorphic gait across various terrains, which is crucial for the evaluation and optimization of assistive robots. Most human gait simulation technologies are limited to simulations on flat ground and fail to achieve gait dynamics simulation across different motion modes, making it impossible to simulate anthropomorphic gait in complex terrains. For example, gait control for ascending and descending stairs, as well as for complex ascending stairs (height > 10 cm, which is common in practice), cannot be achieved. Summary of the Invention
[0004] This application provides a human gait simulation method, device, equipment, medium, and product to solve the problem that existing technologies cannot achieve gait dynamics simulation in complex terrain.
[0005] Firstly, this application provides a human gait simulation method, including: The current gait state is determined based on the motion sensing information from the musculoskeletal model; Based on the current gait state, a reflection loop corresponding to the current gait state is determined, and the motion sensing information is input into the reflection loop for calculation and summation to obtain the muscle excitation signal that drives each muscle in the musculoskeletal model. The musculoskeletal model is driven to move using the muscle excitation signals described above, thereby generating a simulated gait.
[0006] In one embodiment, determining the current gait state based on motion sensing information from a musculoskeletal model includes: Based on the motion sensing information of the musculoskeletal model, the relative position of the foot and the normalized foot load are determined. The current gait state is determined from multiple predefined states based on the relative position of the foot and the normalized foot load; the multiple predefined states refer to multiple discrete states obtained based on gait cycle division.
[0007] In one embodiment, the plurality of discrete states include several discrete states during the support phase and several discrete states during the swing phase; the several discrete states during the support phase include the collision state, the mid-term support state, the late-term support state, and the liftoff state; the several discrete states during the swing phase include the swing state and the landing state.
[0008] In one embodiment, determining the reflection loop corresponding to the current gait state based on the current gait state, and inputting the motion sensing information into the reflection loop for calculation and summation to obtain the muscle excitation signals driving each muscle in the musculoskeletal model includes: Determine the reflex loops corresponding to each muscle in the current gait state; the reflex loops include muscle reflex loops and degree-of-freedom reflex loops. For each muscle, the motion sensing information is input into the corresponding reflection loop of the muscle for calculation and summation to obtain the muscle excitation signal of the muscle in the current gait state.
[0009] In one embodiment, the reflection circuit is optimized in the following way: Initialize the parameter set of the reflection loop; Human gait simulation is performed based on the current parameter set, and gait performance is evaluated based on the motion data generated by the simulation to obtain an fitness score; Update the parameter set based on the fitness score; The process involves iteratively performing human gait simulation based on the current parameter set, evaluating gait performance based on the motion data generated by the simulation, and obtaining a fitness score, until a preset termination condition is met. The parameter set with the highest fitness score is then determined as the optimized parameter set for the reflection loop.
[0010] In one embodiment, the step of driving the musculoskeletal model to move using the muscle excitation signals to generate a simulated gait includes: Based on the muscle excitation signals described, the corresponding muscle force is calculated using a muscle dynamics model. Based on the muscle forces and the geometric information of the muscle attachment points of the musculoskeletal model, calculate the joint torques acting on the joints of the musculoskeletal model. The joint torques are input into a multibody dynamics model to solve for the motion data of the musculoskeletal model in the simulation environment, thereby generating a simulated gait.
[0011] Secondly, this application also provides a human gait simulation device, comprising: The gait state determination module is used to determine the current gait state based on the motion sensing information of the musculoskeletal model. The muscle excitation signal calculation module is used to determine the reflection loop corresponding to the current gait state based on the current gait state, and input the motion sensing information into the reflection loop for calculation and summation to obtain the muscle excitation signal driving each muscle in the musculoskeletal model. The neuromechanical simulation module is used to drive the musculoskeletal model to move using the muscle excitation signals to generate a simulated gait.
[0012] Thirdly, this application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the above-described human gait simulation methods.
[0013] Fourthly, this application also provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described human gait simulation methods.
[0014] Fifthly, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium, and which, when executed by the processor, implements the steps of any of the above-described human gait simulation methods.
[0015] The human gait simulation method, device, equipment, medium, and product provided in this application determine the current gait state through motion sensing information of a musculoskeletal model, laying the foundation for adaptive control. Based on the gait state, the corresponding reflex loop is dynamically matched to simulate the human neural reflex mechanism, ensuring an immediate response to terrain changes. Then, the motion sensing information is input into the reflex loop for calculation to generate muscle excitation signals that conform to physiological laws. Finally, these muscle excitation signals are used to drive the musculoskeletal model to form a stable and realistic simulated gait on complex terrain, effectively improving the adaptability and realism of human gait simulation and overcoming the shortcomings of traditional methods in complex terrain. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is one of the flowcharts of the human gait simulation method provided in this application.
[0018] Figure 2 This is the second flowchart of the human gait simulation method provided in this application.
[0019] Figure 3 This is a schematic diagram comparing simulation data and experimental data under different terrains provided in this application.
[0020] Figure 4 This is a schematic diagram of the human gait simulation device provided in this application.
[0021] Figure 5 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0023] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein.
[0024] Human gait simulation technology is mainly divided into two categories based on the different driving methods of the human body model: joint space model-based and muscle space model-based. The joint space model constructs the human limbs as a multi-rigid-body dynamic system, while the muscle space model constructs a musculoskeletal simulation model of the human body based on anatomical features. Methodologically, it can be divided into inverse simulation and predictive simulation. Unlike inverse simulation, which relies on recording motion data for backward inference, predictive simulation does not rely on prior motion data. Instead, it is based on forward dynamics and directly optimizes and generates the optimal motion trajectory to complete a specific task based on high-level objectives such as stability and energy efficiency.
[0025] This application employs a muscle-space-based predictive gait simulation technology. This technology relies on a mathematical model of the neuromuscular skeletal system, integrating physical laws and neural control principles to generate and predict the kinematic and dynamic characteristics of human gait. This technology boasts high fidelity, particularly in its ability to generate gait across terrain and provide real-time, human-like responses to external forces. Simulation-based evaluation can accelerate the development and optimization of assistive devices, directly improving their performance and task generalization capabilities.
[0026] This application presents the following embodiments, and in conjunction with Figures 1-5This application describes the human gait simulation method provided. The human gait simulation method provided in this application can be implemented based on a human gait simulation device; therefore, this application uses a human gait simulation device as the execution subject to describe the human gait simulation method.
[0027] Combination Figure 1 and Figure 2 , Figure 1 This is one of the flowcharts illustrating the human gait simulation method provided in this application. Figure 2 This is the second flowchart of the human gait simulation method provided in this application.
[0028] like Figure 1 As shown, the human gait simulation method includes the following steps: Step 101: Determine the current gait state based on the motion sensing information of the musculoskeletal model; Step 102: Based on the current gait state, determine the reflection loop corresponding to the current gait state, and input the motion sensing information into the reflection loop for calculation and summation to obtain the muscle excitation signal driving each muscle in the musculoskeletal model; Step 103: Use the muscle excitation signals to drive the musculoskeletal model to move and generate a simulated gait.
[0029] Specifically, the gait generated by existing technology models does not closely resemble real human data. Human gait mechanisms differ fundamentally across different terrains. For example, in multi-terrain walking scenarios: walking on flat ground mainly relies on ankle propulsion, walking on slopes / climbing stairs requires additional thrust from the hip joint, walking on slopes / climbing stairs relies on the continuous energy expenditure of the knee muscles to control the descent speed, and walking down stairs requires a forefoot-first landing gait pattern. Existing technologies have failed to fully simulate these key human biomechanical characteristics.
[0030] Therefore, this application employs a muscle-space-based predictive gait simulation technology, which relies on a musculoskeletal model. The musculoskeletal model is a pre-constructed human dynamics model integrating muscle activity, skeletal movement, and biomechanical transmission relationships. The musculoskeletal model has multiple degrees of freedom, covering the major joints and muscle groups of the human lower limbs.
[0031] In one embodiment, the complex multi-terrain scene includes flat ground, up / down slopes, up / down stairs, etc. The neuromechanical simulation uses a musculoskeletal model with 9 degrees of freedom. Each lower limb is modeled as 9 muscles, namely the gluteus maximus (GLU), iliopsoas (IL), hamstring (HAM), rectus femoris (RF), vastus lateralis and other quadriceps femoris parts (VAS), biceps femoris short head (BFSH), gastrocnemius (GAS), soleus (SOL) and tibialis anterior (TA). Among them, the gluteus maximus is mainly responsible for hip joint extension and stability. In scenarios where the hip joint needs to provide additional thrust when going uphill or up stairs, the gluteus maximus will maintain high-intensity activation for a longer period of time, providing sufficient power for the body to move upward. The iliopsoas, as the main hip flexor, is responsible for flexing the hip joint and driving the lower limb forward during the walking swing phase. In scenarios where going down stairs, the iliopsoas will also work with the center of gravity to adjust the activation intensity to avoid excessive stride and body imbalance. The hamstrings are involved in both hip joint extension and knee joint flexion. During downhill and stair descent, they will maintain low-intensity activation to control the speed of the lower limb descent and avoid excessive impact on the knee joint due to gravity accelerating the descent. The rectus femoris, as a bi-joint muscle with both knee extension and hip flexion functions, will work with the iliopsoas to complete the hip flexion and forward swing when going up stairs, while maintaining knee joint stability during the support phase and sharing the pressure from the body weight. The lateral quadriceps femoris and other quadriceps muscles are mainly responsible for knee extension during the support phase. Their activation intensity increases significantly during the support phase when going uphill or up stairs, counteracting the extra load from the uphill climb. The short head of the biceps femoris is responsible for knee flexion during the swing phase, adjusting the swing amplitude of the lower leg to adapt the stride length to the spacing between steps or slopes on different terrains. The gastrocnemius and soleus muscles together form the triceps surae, which is the main source of power for ankle propulsion when walking on flat ground. During the push-off phase of walking on flat ground, both muscles are activated simultaneously and explosively, propelling the body forward. When going downhill, the soleus muscle is continuously and slightly activated to maintain ankle stability and cushion the downward force of the body. The tibialis anterior is responsible for ankle dorsiflexion, lifting the toes during the swing phase to avoid dragging the ground. When using a forefoot-first landing gait pattern when going down stairs, the tibialis anterior adjusts its activation level in advance, controlling the angle and speed of the forefoot landing to cushion the impact of landing. These muscles cover the three main lower limb joints: the hip, knee, and ankle. Each joint is equipped with independent mechanical parameters and activity constraints, which can accurately match the stress and range of motion requirements of human joints under different terrains.
[0032] Real-time acquisition of multi-dimensional motion sensing information from musculoskeletal models, such as joint angles, angular velocities, muscle force data, foot contact status, and other multi-dimensional kinematic and dynamic information.
[0033] This application proposes a controller for simulating human gait muscle reflexes in a multi-terrain simulation environment. This controller, simulating a human motion control system, employs a two-layer structure: Upper spinal cord layer (achieving higher-level control): This layer is modeled as a finite state machine, and its function is to divide the gait cycle into different stages and activate the corresponding reflex loops according to these stages; Spinal cord layer (enabling lower-level control): This layer consists of a series of reflex circuits that can calculate muscle excitation signals based on muscle length, velocity, force, and other states. This means that when muscles are in different motion states, the spinal cord layer can quickly respond through these reflex circuits, generating corresponding muscle excitations to drive muscle movement.
[0034] The upper spinal cord is designed as a finite state machine with several states to achieve fine division of leg movement states, and the key biomechanical principles of multi-terrain walking are encoded as muscle reflex circuits to simulate gait adapted to various terrains.
[0035] Based on the real-time motion sensing information of the musculoskeletal model, it is matched with the refined leg movement state to determine which specific stage of the gait cycle the musculoskeletal model is currently in.
[0036] A pre-configured correspondence between gait states and reflex loops is established, based on the physiological activity characteristics of different gait phases during normal human walking. Generally, a complete human walking gait cycle can be divided into two main phases: the support phase and the swing phase. Each main phase can be further subdivided into multiple sub-phases. The muscle activation patterns differ significantly in each gait state; therefore, an independent reflex loop is configured for each gait state to match the muscle exertion requirements of different gait phases. After identifying the current gait state, the reflex loop that needs to be activated for that state is matched. This reflex loop receives real-time sensor information such as length, contraction speed, and force from the muscle itself, calculates the appropriate muscle excitation signal for the current gait state, and transmits it to the musculoskeletal model in the simulation environment to drive the movement.
[0037] Each muscle outputs a corresponding contractile force based on the received muscle excitation signal, driving the corresponding joint to move according to biomechanical laws. The entire musculoskeletal model can then form a simulated gait that matches the current terrain and conforms to the laws of real human movement. Each simulation cycle repeats the above process of state recognition-reflection matching-excitation calculation-model driving, thus generating a continuous and stable multi-terrain natural simulated gait.
[0038] Compared to traditional methods, this simulation method does not require retraining a completely new control model for each terrain. It only needs to adjust the parameters of the corresponding reflection loop to adapt to the gait characteristics of different terrains. This preserves the biological realism of human motion control and reduces the implementation complexity of multi-terrain gait simulation. The generated simulated gait has a higher similarity to real human measured data in multiple dimensions such as joint motion trajectory, muscle force mode, and plantar pressure distribution.
[0039] The human gait simulation method provided in this application determines the current gait state through motion sensing information of a musculoskeletal model, laying the foundation for adaptive control. Based on the gait state, it dynamically matches the corresponding reflex loop to simulate the human neural reflex mechanism, ensuring an immediate response to terrain changes. Then, the motion sensing information is input into the reflex loop for calculation to generate muscle excitation signals that conform to physiological laws. Finally, these muscle excitation signals are used to drive the musculoskeletal model to form a stable and realistic simulated gait on complex terrain, effectively improving the adaptability and realism of human gait simulation and overcoming the shortcomings of traditional methods in complex terrain.
[0040] In one embodiment, determining the current gait state based on motion sensing information from a musculoskeletal model includes: Based on the motion sensing information of the musculoskeletal model, the relative position of the foot and the normalized foot load are determined. The current gait state is determined from multiple predefined states based on the relative position of the foot and the normalized foot load; the multiple predefined states refer to multiple discrete states obtained based on gait cycle division.
[0041] Specifically, the human gait cycle is divided into two major phases, the support phase and the swing phase, according to the contact relationship between the foot and the ground. These can be further subdivided into six predefined discrete states. The support phase can be further subdivided into four discrete states: collision (CO), mid-stance (MS), late-stance (LS), and takeoff (LO). The swing phase can be further subdivided into two discrete states: swing (SW) and landing (LD).
[0042] The impact state is the moment the foot first contacts the ground. At this point, the foot bones are subjected to the impact load from the ground, and the muscles need to contract rapidly to buffer the impact force, causing the foot load to rise rapidly to its peak. The mid-stability state is when the foot is fully in contact with the ground, supporting the entire body's center of gravity as it shifts forward. At this point, the foot load remains in a relatively stable range, and positional information shows that the foot remains relatively stationary relative to the ground, with only a slight shift accompanied by the movement of the center of gravity. The late-stability state is the transitional phase where the body's center of gravity has passed the foot's support surface, preparing to take a step forward. At this point, the foot load begins to gradually decrease, with the forefoot area accounting for a larger proportion of the load. The gait will noticeably improve, with the heel gradually lifting off the ground; the off-ground state is the critical point where the foot completely leaves the ground, at which point the overall load on the foot drops to near zero, and the position begins to shift significantly as the leg swings forward; the swinging state refers to the stage where the foot moves forward in the air after leaving the ground, preparing to complete the next landing step. In this state, the load on the foot remains close to zero, and the position and speed will exhibit regular dynamic changes as the leg swings; the landing state is the preparation stage where the foot is about to touch the ground. At this time, the vertical speed of the foot gradually decreases, the position approaches the ground, and it waits to enter the collision state of the next gait cycle.
[0043] In the state determination process, the three-dimensional coordinates of key points on the foot are first extracted from motion sensing information to calculate the overall spatial position and posture of the foot, determining its relative position, i.e., the position of the foot's center of mass relative to the body's center of mass in the sagittal plane of walking. Simultaneously, the ground reaction force of the foot is extracted from the motion sensing information, and the ground reaction force is divided by body weight to obtain the normalized foot load. By comparing the real-time values of these two types of data with the feature thresholds of six discrete states, the gait state corresponding to the current moment can be quickly matched, providing an accurate basis for subsequent dynamic matching of the reflection loop.
[0044] State transition rules are as follows Figure 2 As shown in the mid-to-high-level control module, the high-level control monitors the normalized foot load ( ) and the relative position of the feet ( Two core parameters, and the first normalized foot load threshold ( The normalized foot load threshold used to determine the transition from the supporting leg to the swing leg, and the second normalized foot load threshold ( Used to determine the normalized foot load threshold when switching from the swing leg to the supporting leg, and the first foot relative position threshold ( The first is used to determine the foot relative position threshold when switching from the middle to the late support phase, and the second foot relative position threshold. A comparison between the relative foot position thresholds used to determine the transition from the swing phase to the landing phase is used to divide the gait into a support phase and a swing phase, and the transition is implemented through a state machine. Normalized foot load on the lateral leg is also considered. Less than the first normalized foot load threshold At this time, it indicates a transition from the collision state to the mid-stage support state; the relative position of the feet Less than the first foot relative position threshold This indicates a transition from a mid-stage support state to a late-stage support state; the normalized foot load on the lateral leg. Greater than the second normalized foot load threshold This indicates a transition from a post-support state to a ground-free state; the normalized foot load of the foreleg. Less than the first normalized foot load threshold This indicates a transition from an off-ground state to a swinging state; the relative position of the feet... Greater than the relative position threshold of the second foot This indicates a transition from a swinging state to a landing state; the normalized foot load of the current leg. Greater than the second normalized foot load threshold When the state transitions from landing to collision, it indicates a change in state. One gait cycle is completed according to the aforementioned state transition rules.
[0045] Based on two key biomechanical signals extracted from real-time motion sensing information—the relative position of the foot and the normalized foot load—this application accurately determines which stage of the gait cycle the user is currently in, simulating the high-level control mechanism of the human central nervous system for walking, and constructs a stable and robust gait state recognition mechanism.
[0046] In one embodiment, the step of determining a reflection loop corresponding to the current gait state based on the current gait state, and inputting the motion sensing information into the reflection loop for calculation and summation to obtain muscle excitation signals driving each muscle in the musculoskeletal model includes: Determine the reflex loops corresponding to each muscle in the current gait state; the reflex loops include muscle reflex loops and degree-of-freedom reflex loops. For each muscle, the motion sensing information is input into the corresponding reflection loop of the muscle for calculation and summation to obtain the muscle excitation signal of the muscle in the current gait state.
[0047] Specifically, human walking is essentially a coordinated movement of multiple muscles and joint degrees of freedom. Each muscle is responsible for movement output in different directions, and each joint degree of freedom requires regulation according to corresponding movement rules. In different phases, the contraction tasks that muscles need to perform are completely different. For example, during the support phase, leg muscles need to contract to maintain body weight and resist ground impact, while during the swing phase, muscles only need to drive the legs forward without bearing the weight of the entire body. If a uniform reflex rule is used to regulate muscle movement in all phases, it is easy to cause movement imbalance, stiff gait, or even the simulation model falling over. However, by breaking down the reflex rule into reflex loops for different phases, the movement regulation of each muscle can be more targeted. Furthermore, through simple summation and superposition, it can quickly adapt to the movement requirements of different phases without designing complex global weight adjustment algorithms. This ensures both the efficiency of simulation calculations and improves the naturalness of the final gait output.
[0048] Reflex circuits include muscle reflex circuits and degree-of-freedom reflex circuits. Muscle reflex circuits are basic control units designed for individual muscles. They can sense changes in muscle length, contraction speed, and other physical parameters, outputting corresponding muscle excitation signals to regulate muscle contraction strength and ensure the muscle maintains its intended working state. For example, when a muscle is rapidly stretched, the muscle reflex circuit can quickly increase muscle tension, preventing overstretching and tissue damage. Degree-of-freedom reflex circuits, on the other hand, are control units designed for the degrees of freedom of human joints. They can simultaneously act on multiple muscle groups controlling the same degree of freedom. By sensing changes in joint angle and angular velocity, they collaboratively adjust the activation level of antagonistic muscle groups to achieve stable control of joint movement.
[0049] The muscle reflex circuit is represented by the following formula: A degree-of-freedom reflection loop is represented by the following formula: in, and These are the outputs of the muscle reflex loop and the degree-of-freedom reflex loop, respectively, both representing the driving force of the target muscle. Activated muscle excitation; It is the input to the muscle reflex circuit, originating from the source muscle. Physiological information, including muscle strength Muscle length or speed ; and These are the inputs to the degree-of-freedom (DoF) reflex loop, representing the position (e.g., joint angle) and velocity (e.g., joint angular velocity) of the source DoF, respectively; ± indicates excitatory / inhibitory reflexes, with the corresponding loop output being greater than / less than 0. It is a non-negative truncation function, ensuring that the output of the reflection loop is non-negative. , , , , These are control parameters that can be optimized. It is the gain parameter of the muscle reflex circuit, which controls the source muscle. Information on target muscles The intensity of the effect of excitement; It is the proportional gain of the degree-of-freedom reflection loop (similar to the proportional term in PID control), and the position deviation of the control source degree of freedom. Target muscle The effect of excitement; It is the differential gain of the degree-of-freedom reflection loop (similar to the differential term in PID control), controlling the velocity of the source degree of freedom. Target muscle The effect of excitement; This represents the baseline value for information about the muscle reflex circuit; The reference value for the position of the reflection loop representing the degree of freedom.
[0050] For walking on stairs and slopes, biomechanical mechanisms such as landing cushioning and mid-stance knee joint damping control are introduced and encoded into corresponding reflex loops, thereby improving the realism and coordination of the movement posture during the stance phase.
[0051] The correspondence between gait states and reflection loops is as follows: Figure 2 The table, as shown in the low-to-mid-level control module, displays nine different muscles in its rows and six gait states in its columns. The cell content indicates the reflex loop activated by that muscle in the corresponding state, formatted as follows: +L / -L: excitatory / inhibitory muscle length reflex loop; +F / -F: excitatory / inhibitory muscle force reflex loop; PD: degree-of-freedom reflex loop; +V / -V: excitatory / inhibitory muscle velocity reflex loop. Parentheses indicate the source muscle / degree of freedom (tilt represents the sagittal trunk tilt angle). If no parentheses are present, it indicates that the source muscle is identical to the target muscle.
[0052] Taking the IL muscle as an example, it is +L in CO, MS, and LS states, indicating that when the leg is in CO, MS, or LS states, the muscle has an excitatory reflex from its own muscle length. In LD state, it is -V (HAM), indicating that when the leg is in LD state, the muscle has an inhibitory reflex from the HAM muscle velocity. Taking the GLU muscle as an example, it is +L and PD (tilt) in CO, MS, and LS states, indicating that when the leg is in CO, MS, or LS states, the muscle has an excitatory reflex from its own muscle length, which, in conjunction with trunk tilt feedback, makes fine adjustments to maintain pelvic stability. In LD state, it is +L (HAM), indicating that when the leg is in LD state, the muscle has an excitatory reflex from the HAM muscle length.
[0053] Utilizing the biomechanical principles of multi-terrain coding, a low-level reflection controller is constructed. The overall structure of the controller is as follows: Figure 2 As shown in the table.
[0054] In the actual process of calling the reflection loop, the mapping relationship in Table 2 is used to retrieve all the reflection loops of each muscle in the current gait state. The calculation and summation of each reflection loop is used as the muscle excitation signal output by the controller to that muscle.
[0055] Based on the current gait state and combined with the biomechanical principles of human gait when walking on various terrains, this application's embodiments design several reflex loops for each muscle, substitute motion sensing information into these reflex loops for calculation and summation, and obtain the muscle excitation signal of the target muscle in the current state, so that the changes in the excitation signal when walking on different terrains are more in line with the physiological regulation law of real human walking.
[0056] In one embodiment, the reflection loop is optimized in the following way: Initialize the parameter set of the reflection loop; Human gait simulation is performed based on the current parameter set, and gait performance is evaluated based on the motion data generated by the simulation to obtain an fitness score; Update the parameter set based on the fitness score; The process involves iteratively performing human gait simulation based on the current parameter set, evaluating gait performance based on the motion data generated by the simulation, and obtaining a fitness score, until a preset termination condition is met. The parameter set with the highest fitness score is then determined as the optimized parameter set for the reflection loop.
[0057] Specifically, in the simulation process, the controller calculates muscle excitation signals based on the motion sensing information from the simulation feedback and drives the musculoskeletal model to move. Based on this, the optimizer uses the motion data generated by the simulation and the covariance matrix adaptive evolution strategy (CMA-ES) to iteratively update the control parameters, thereby achieving automatic gait optimization.
[0058] The optimizer first initializes its core multivariate Gaussian search distribution, which includes: a mean vector, which is randomly generated and represents the initial search center of the parameter space; a covariance matrix, which is usually initialized as an identity matrix, indicating that each parameter is initially independent and searches uniformly around the mean; and a step size, which controls the overall radius of the search distribution.
[0059] After initialization, sampling and evaluation are required. Based on the Gaussian distribution defined by the current mean, covariance matrix, and step size, the optimizer randomly samples and generates... There are several candidate parameter vectors. Each vector represents a complete set of reflection controller parameter configurations. For each candidate parameter vector in the population, the system initiates an independent human gait simulation process. The controller uses these candidate parameters to drive the musculoskeletal model to walk on the target terrain for a predetermined time or distance. After the simulation, motion data, such as angles and torques, are collected. Based on a preset objective function, the gait performance of this simulation is scored. This score is the fitness score. The objective function can comprehensively consider various gait performance requirements, including walking stability requirements, walking speed requirements, energy efficiency requirements, and other objectives.
[0060] After obtaining the fitness scores of all candidate parameter vectors, the update phase begins. From the current... Select from candidate parameter vectors The candidate parameter vector with the highest fitness is selected as the elite individual. The information from the elite individuals is used to adaptively update the Gaussian search distribution (including updating the mean, covariance matrix, and step size) to make it more likely to sample better solutions.
[0061] The updated Gaussian search distribution will be used for sampling in the next iteration. This process is repeated continuously, forming a closed-loop optimization cycle. The loop stops when a termination condition is met, such as reaching the maximum number of iterations or finding a solution that meets the performance requirements.
[0062] After the optimization process is completed, the set of reflection loop parameters with the highest fitness score in all previous evaluations will be used as the final output.
[0063] This application embodiment constructs a closed loop of parameter sampling, simulation evaluation, fitness feedback, and strategy update, and uses an adaptive evolution strategy to achieve iterative optimization of the parameter space. The optimized parameter space is then used to reorganize the reflection loop, thereby achieving automatic gait optimization.
[0064] In one embodiment, the step of using the muscle excitation signals to drive the musculoskeletal model to move and generate a simulated gait includes: Based on the muscle excitation signals described, the corresponding muscle force is calculated using a muscle dynamics model. Based on the muscle forces and the geometric information of the muscle attachment points of the musculoskeletal model, calculate the joint torques acting on the joints of the musculoskeletal model. The joint torques are input into a multibody dynamics model to solve for the motion data of the musculoskeletal model in the simulation environment, thereby generating a simulated gait.
[0065] Specifically, the muscle dynamics model is based on the muscle excitation signal and takes into account physiological constraints such as muscle length, velocity characteristics, and activation dynamics, to solve the muscle force generated by each muscle in real time. .
[0066] Then, based on the anatomical structure of the musculoskeletal system, the force generated by each muscle is converted into a torque contribution to the relevant joint along its lever arm. The torque contributions of all muscles crossing the same joint are then vector-summed to obtain the joint torque acting on each degree of freedom of the joint. .
[0067] Finally, the converted joint torques, along with the external forces such as gravity and ground contact friction acting on the musculoskeletal model, are substituted into the multibody dynamics equations. Through positive integration of dynamics, the position, velocity, and acceleration data of each part of the musculoskeletal model in the next simulation time step are calculated, thereby obtaining the continuous kinematic state, i.e., generating a complete simulated gait.
[0068] The embodiments of this application obtain a simulated gait based on muscle excitation signals that conform to the physiological laws of human movement, through forward solving of muscle dynamics, musculoskeletal geometry and multibody dynamics. The whole process is closer to the neural control and mechanical transmission logic of actual human movement, and the obtained simulated gait is more in line with the movement characteristics of real human gait.
[0069] Figure 3 This is a comparative diagram of simulation data and experimental data under different terrains provided in this application. By introducing biomechanical principles under different terrains and encoding them into human motion control using muscle reflex loops, it is possible to generate human-like gait for walking on flat ground (speed from 0.8-1.3m / s), going up and down slopes (slope from 5° to 18°), and going up and down stairs (height from 10cm to 18cm, depth of 40cm). Figure 3 The study demonstrates a comparison between simulation data and experimental data under different terrain conditions, primarily comparing the angles and torques of the three degrees of freedom of the lower limbs (hip extension / flexion, knee extension / flexion, and ankle plantar flexion / dorsiflexion).
[0070] Figure 4 This is a schematic diagram of the human gait simulation device provided in this application.
[0071] like Figure 4 As shown, the human gait simulation device includes: The gait state determination module 410 is used to determine the current gait state based on the motion sensing information of the musculoskeletal model. The muscle excitation signal calculation module 420 is used to determine the reflection loop corresponding to the current gait state based on the current gait state, and input the motion sensing information into the reflection loop for calculation and summation to obtain the muscle excitation signal driving each muscle in the musculoskeletal model. The neuromechanical simulation module 430 is used to drive the musculoskeletal model to move using the muscle excitation signals to generate a simulated gait.
[0072] The human gait simulation device provided in this application determines the current gait state through motion sensing information of a musculoskeletal model, laying the foundation for adaptive control. Based on the gait state, it dynamically matches the corresponding reflex loop to simulate the human neural reflex mechanism, ensuring an immediate response to terrain changes. Then, the motion sensing information is input into the reflex loop for calculation to generate muscle excitation signals that conform to physiological laws. Finally, these muscle excitation signals are used to drive the musculoskeletal model to form a stable and realistic simulated gait on complex terrain, effectively improving the adaptability and realism of human gait simulation and overcoming the shortcomings of traditional methods in complex terrain.
[0073] In one embodiment, the gait state determination module 410 is further configured to: Based on the motion sensing information of the musculoskeletal model, the relative position of the foot and the normalized foot load are determined. The current gait state is determined from multiple predefined states based on the relative position of the foot and the normalized foot load; the multiple predefined states refer to multiple discrete states obtained based on gait cycle division.
[0074] In one embodiment, the plurality of discrete states include several discrete states during the support phase and several discrete states during the swing phase; the several discrete states during the support phase include the collision state, the mid-term support state, the late-term support state, and the liftoff state; the several discrete states during the swing phase include the swing state and the landing state.
[0075] In one embodiment, the reflection loop mapping module 420 is further configured to: Determine the reflex loops corresponding to each muscle in the current gait state; the reflex loops include muscle reflex loops and degree-of-freedom reflex loops. For each muscle, the motion sensing information is input into the corresponding reflection loop of the muscle for calculation and summation to obtain the muscle excitation signal of the muscle in the current gait state.
[0076] In one embodiment, the reflection circuit is optimized in the following way: Initialize the parameter set of the reflection loop; Human gait simulation is performed based on the current parameter set, and gait performance is evaluated based on the motion data generated by the simulation to obtain an fitness score; Update the parameter set based on the fitness score; The process involves iteratively performing human gait simulation based on the current parameter set, evaluating gait performance based on the motion data generated by the simulation, and obtaining a fitness score, until a preset termination condition is met. The parameter set with the highest fitness score is then determined as the optimized parameter set for the reflection loop.
[0077] In one embodiment, the neuromechanical simulation module 440 is further configured to: Based on the muscle excitation signals described, the corresponding muscle force is calculated using a muscle dynamics model. Based on the muscle forces and the geometric information of the muscle attachment points of the musculoskeletal model, calculate the joint torques acting on the joints of the musculoskeletal model. The joint torques are input into a multibody dynamics model to solve for the motion data of the musculoskeletal model in the simulation environment, thereby generating a simulated gait.
[0078] It should be noted that the human gait simulation device provided in this application can execute the human gait simulation method described in any of the above embodiments during actual operation, which will not be elaborated in this embodiment.
[0079] Figure 5 This is a schematic diagram of the structure of the electronic device provided in this application, such as... Figure 5As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a human gait simulation method, which includes: determining the current gait state based on motion sensing information of a musculoskeletal model; determining a reflection loop corresponding to the current gait state based on the current gait state, and inputting the motion sensing information into the reflection loop for calculation and summation to obtain muscle excitation signals that drive each muscle in the musculoskeletal model; and using each muscle excitation signal to drive the musculoskeletal model to move, thereby generating a simulated gait.
[0080] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0081] On the other hand, this application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the human gait simulation method provided in the above embodiments. The method includes: determining the current gait state based on the motion sensing information of a musculoskeletal model; determining a reflection loop corresponding to the current gait state based on the current gait state, and inputting the motion sensing information into the reflection loop for calculation and summation to obtain muscle excitation signals that drive each muscle in the musculoskeletal model; and using each muscle excitation signal to drive the musculoskeletal model to move, thereby generating a simulated gait.
[0082] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the human gait simulation method provided in the above embodiments. The method includes: determining a current gait state based on motion sensing information of a musculoskeletal model; determining a reflection loop corresponding to the current gait state based on the current gait state, and inputting the motion sensing information into the reflection loop for calculation and summation to obtain muscle excitation signals that drive each muscle in the musculoskeletal model; and using each muscle excitation signal to drive the musculoskeletal model to move, thereby generating a simulated gait.
[0083] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for simulating human gait, characterized in that, The human gait simulation method includes: The current gait state is determined based on the motion sensing information from the musculoskeletal model; Based on the current gait state, a reflection loop corresponding to the current gait state is determined, and the motion sensing information is input into the reflection loop for calculation and summation to obtain the muscle excitation signal that drives each muscle in the musculoskeletal model. The musculoskeletal model is driven to move using the muscle excitation signals described above, thereby generating a simulated gait.
2. The human gait simulation method according to claim 1, characterized in that, Determining the current gait state based on motion sensing information from the musculoskeletal model includes: Based on the motion sensing information of the musculoskeletal model, the relative position of the foot and the normalized foot load are determined. The current gait state is determined from multiple predefined states based on the relative position of the foot and the normalized foot load; the multiple predefined states refer to multiple discrete states obtained based on gait cycle division.
3. The human gait simulation method according to claim 2, characterized in that, The multiple discrete states include several discrete states during the support phase and several discrete states during the swing phase; the several discrete states during the support phase include the collision state, the mid-term support state, the late-term support state, and the liftoff state; the several discrete states during the swing phase include the swing state and the landing state.
4. The human gait simulation method according to claim 2, characterized in that, The step involves determining a reflection loop corresponding to the current gait state, inputting the motion sensing information into the reflection loop for calculation and summation to obtain muscle excitation signals driving each muscle in the musculoskeletal model, including: Determine the reflex loops corresponding to each muscle in the current gait state; the reflex loops include muscle reflex loops and degree-of-freedom reflex loops. For each muscle, the motion sensing information is input into the corresponding reflection loop of the muscle for calculation and summation to obtain the muscle excitation signal of the muscle in the current gait state.
5. The human gait simulation method according to claim 4, characterized in that, The reflection circuit is optimized in the following ways: Initialize the parameter set of the reflection loop; Human gait simulation is performed based on the current parameter set, and gait performance is evaluated based on the motion data generated by the simulation to obtain an fitness score; Update the parameter set based on the fitness score; The process involves iteratively performing human gait simulation based on the current parameter set, evaluating gait performance based on the motion data generated by the simulation, and obtaining a fitness score, until a preset termination condition is met. The parameter set with the highest fitness score is then determined as the optimized parameter set for the reflection loop.
6. The human gait simulation method according to any one of claims 1 to 5, characterized in that, The process of using the muscle excitation signals to drive the musculoskeletal model to move and generate a simulated gait includes: Based on the muscle excitation signals described, the corresponding muscle force is calculated using a muscle dynamics model. Based on the muscle forces and the geometric information of the muscle attachment points of the musculoskeletal model, calculate the joint torques acting on the joints of the musculoskeletal model. The joint torques are input into a multibody dynamics model to solve for the motion data of the musculoskeletal model in the simulation environment, thereby generating a simulated gait.
7. A human gait simulation device, characterized in that, The human gait simulation device includes: The gait state determination module is used to determine the current gait state based on the motion sensing information of the musculoskeletal model. The muscle excitation signal calculation module is used to determine the reflection loop corresponding to the current gait state based on the current gait state, and input the motion sensing information into the reflection loop for calculation and summation to obtain the muscle excitation signal driving each muscle in the musculoskeletal model. The neuromechanical simulation module is used to drive the musculoskeletal model to move using the muscle excitation signals to generate a simulated gait.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the human gait simulation method as described in any one of claims 1 to 6.
9. A non-transitory computer-readable storage medium, wherein a computer program is stored on the non-transitory computer-readable storage medium, characterized in that, When the computer program is executed by the processor, it implements the steps of the human gait simulation method as described in any one of claims 1 to 6.
10. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the human gait simulation method as described in any one of claims 1 to 6.