Sage Journals: Discover world-class research

Abstract

In order to improve the control ability of the human body in the process of physical training, it is necessary to carry out the inverse simulation analysis of the physical training process and establish the process control model of the physical training. The complex problem of high-dimensional spatial motion planning involved in physical training is decomposed into a series of sub-problems in low-dimensional space, and the inertial attitude parameter fusion is carried out according to the position and pose state of the human body in the end of the workspace during the process of physical training. The design of sensor node and base station in the system can realize real-time collection of motion parameters of motion collectors. The multi-dimensional control of physical training process is carried out by fuzzy constraint and inverse integral control, and the attitude parameters of human body are adjusted by means of mechanical analysis model and inertial parameter analysis method. The simulation results show that the inversion simulation control has better convergence, higher control quality, and better inverse simulation performance in the process of physical training, which can effectively guide physical training and improve the effect of physical training.

Keywords

Physical training inversion integral human attitude parameter adjustment wireless sensor network simulation

Introduction

With the rapid development of science and technology, people pay more and more attention to improving the quality of physical training by means of science and technology. In the past sports training process, the trainer used the less difficult video replay, and the analysis management way to explain the movement key points for the athlete is not intuitive and scientific, and lacks authenticity and interaction. It cannot meet the basic evaluation requirements of athletes’ movement norms and training results. However, the sports video can extract the training parameters of athletes, and can further analyze the training parameters by constructing models and observing the training movements, so as to obtain the results of the analysis of the pros and cons of the athletes’ training actions.¹ Wireless sensor network (WSN) is a distributed sensor network whose terminals are sensors that can sense and examine the outside world. Sensors in WSN communicate wirelessly, so network settings are flexible, device locations can be changed at any time, and wired or wireless connections can be made to the Internet. A multi-hop self-organizing network formed by wireless communication ensures athletes know their shortcomings intuitively and improve their quality of physical training according to the guidance of coaches. The movement of human body is a fuzzy control process in physical training. By simulating the posture of human body in physical training and combining with the method of mechanical analysis, the motion model of human body can be analyzed and the effect of physical training can be improved. It has important application value to study the simulation method of human body posture in physical training.²

Traditionally, the research of human body attitude simulation control algorithm in sports training is mainly divided into path offset correction and path tracking control algorithm based on particle filter motion rigid trajectory decomposition and motion attitude simulation control algorithm based on genetic evolution algorithm.³ In physical training based on genetic algorithm (GA), the human body control algorithm is commonly used. Wang and Shao⁴ proposed a reinforcement learning algorithm based on neural network and applied to human trajectory correction and tracking control in physical training. And in the physical training of the human body to avoid obstacles in good practice, based on neural network reinforcement learning algorithm, it is combined with intelligent control structure applied in mobile sports training of human path and motion posture simulation control. The algorithm takes the human body in physical training as the model, which has certain guiding significance, but the model is not detailed enough and cannot completely reflect the characteristics of human body positioning and sports posture simulation in sports training, and the tracking control performance is not good. Zhang et al.⁵ proposed an algorithm of simultaneous location of human body and map creation in sports training based on particle filter, and the pose model of human body and the creation of map of surrounding environment in sports training were obtained. The attitude tracking control of human body in physical training is realized, but the calculation accuracy of the algorithm is not high, the realization is more complex, and the application value is not good.⁶ On the basis of designing human posture integral synovium control in physical training, multi-tree hybrid genetic evolution method was used to track human pose target in physical training, and human body trajectory control was carried out under multi-information fusion. The control accuracy is improved, but the convergence performance of the algorithm is not good, the local optimization ability and search speed need to be improved.⁷ It is concluded that human posture is a multivariable nonlinear control system affected by small disturbance piecewise linear error in physical training. The traditional GA for motion attitude simulation control appears with steady tracking error in the boundary layer.

In view of the above problems, this article presents a physical training process control model based on inverse integral and fuzzy control. The complex problem of high-dimensional spatial motion planning involved in physical training is decomposed into a series of sub-problems in low-dimensional space, and the inertial attitude parameter fusion is carried out according to the position and pose state of the human body in the end of the workspace during the process of physical training. The multi-dimensional control of physical training process is carried out by fuzzy constraint and inverse integral control, and the attitude parameters of human body are adjusted by means of mechanical analysis model and inertial parameter analysis method. The inversion simulation control is realized in the process of physical training, and the effect of physical training is improved. Finally, the performance test is carried out through the simulation experiment, which shows the superior performance of improving the sports posture simulation and control under the method of this article.

Design of body posture environment model and analysis of constraint parameters in physical training

Human posture simulation model for physical training

In this article, we first simulate and construct the human posture simulation model in sports training, using bottom-up agent-based modeling method, and abstract the human posture spatial coordinate into the virtual world of genetic population.^8,9 The human posture distributed structure model of physical training is shown in Figure 1.

Figure 1.

Distributed structure model of human posture in physical training.

In physical training, the movement space of human body can be expressed by a plane right angle coordinate system, which is organically connected with macro and micro. The optimization of human body’s path and the control of posture simulation in physical training are generally based on the multi-information fusion model of N decision variables. Several constraints of human posture simulation control in physical training are obtained as follows

\sin β = \cos θ [\cos γ \sin (ψ - ψ_{V}) + \sin ϑ \sin γ \cos (ψ - ψ_{V})] - \sin θ \cos ϑ \sin γ

(1)

\sin α = \frac{{\cos θ [\sin ϑ \cos γ \cos (ψ - ψ_{V}) - \sin γ \sin (ψ - ψ_{V})] - \sin θ \cos ϑ \cos γ}}{\cos β}

(2)

\sin γ_{V} = \frac{(\cos α \sin β \sin ϑ - \sin α \sin β \cos γ \cos ϑ + \cos β \sin γ \cos ϑ)}{\cos θ}

(3)

Considering the nonlinear characteristics and uncertainty of the model, the two-position motion space of the human body in physical training is obtained

R_{p} = R_{TAN} + \frac{R_{HAC} ψ_{H}}{\cos 30^{°}} - X_{HAC}

(4)

Every grid of human posture distribution in physical training can be described by point S; through the above description,^9,10 the complex problem of high-dimensional spatial motion planning involved in physical training is decomposed into a series of sub-problems in low-dimensional space. According to the position and attitude status of the human body in the end of the workspace during physical training, the inertial attitude parameters are obtained as follows

R_{p} = R_{TURN} + \frac{R_{HAC} ψ_{H}}{\cos 30^{°}} - X_{HAC}

(5)

where $R_{TURN} = (R_{TAC} | Δ ψ_{HAC} |) / \cos ϕ + \sqrt{A^{2} + B^{2}}$ , in which $R_{TAC} = v_{2} \cos γ / g \tan ϕ$ . Combined with the method of attitude parameter fusion, the adjustment and inverse control of human posture and pose parameters in the course of physical training are carried out.^11,12 The working process of sensor protocol for information via negotiation (SPIN) protocol is shown in Figure 2.

Figure 2.

Working process of SPIN protocol.

Description of human body posture simulation control parameters in physical training

On the basis of constructing the grid structure model of human body motion space in physical training, this article analyzes the environment parameter model of human body posture simulation control in sports training, and according to the classical GA. In the process of sports posture simulation, the human body finds the target point and moves through the way of genetic evolution.^13,14 The sensor collects a certain amount of external data, and it carries on the information characteristic analysis and the control. The process function of human body simulation control in physical training is obtained as follows

ϕ_{COM} = sgn (γ) [30 + G_{R} Δ \overset{\cdot}{R} + GR Δ \overset{\cdot}{R}]

(6)

Suppose $x_{t}$ is the weight vector of t moment in the longitudinal motion process of the human body in the course of sports training, and time $d_{0, \dots, t}$ is the measured linear disturbance data value. The measured human body in physical training and the attitude control equation of human body in physical training are expressed in discrete form. For $t = 0, 1, \dots, k$ , the corresponding genetic evolution measurement value is $Z_{k} = {z_{0}, z_{1}, \dots, z_{k}}$ , and the zero potential energy surface is chosen as the plane of human body motion in two sports training. In physical training, the kinetic energy T and the potential energy V of the human body control system are

\begin{matrix} T = \frac{1}{2} M_{RL} {\overset{\cdot}{X}}_{RL}^{2} + \frac{1}{2} M_{RR} {\overset{\cdot}{X}}_{RR}^{2} + \frac{1}{2} J_{RL} {\overset{\cdot}{θ}}_{RL}^{2} \\ + \frac{1}{2} J_{RR} {\overset{\cdot}{θ}}_{RR}^{2} \\ + \frac{1}{2} M_{P} [({\overset{\cdot}{θ}}_{P} L \cos θ_{P} + {\overset{\cdot}{X}}_{RM})^{2} + (- θ_{P} L \cos {\overset{\cdot}{θ}}_{P})^{2}] + \frac{1}{2} J_{P θ} {\overset{\cdot}{θ}}_{P}^{2} \\ + \frac{1}{2} J_{P δ} {\overset{\cdot}{δ}}^{2} \end{matrix}

(7)

V = M_{P} gL \cos_{θ_{P}}

(8)

where $x_{k}$ is used to express the displacement of the center of human body position and posture in physical training, that is, ${x_{k}, y_{k}, θ_{k}}$ .^15,16 Combining with the constraint conditions, the parameters of human body motion state in physical training are described as follows

{\overset{\cdot}{V}}_{1} = e_{1} e_{2} - c_{1} {e_{1}}^{2} - e_{1} λ_{1} ζ_{1}

(9)

where $u_{k}$ is the moment of inertia of the human body around the center of mass and parallel to the Z-axis in the physical training, and $w_{k}$ is the measurement error in the process

e_{2} = ω_{2} - ω_{2 r} = ω_{2} + c_{1} e_{1} + λ_{1} ζ_{1} - {\overset{\cdot}{ϑ}}_{r}

(10)

By describing the parameters of human posture simulation control in sports training, this article provides a data basis for human posture simulation control in sports training.

Algorithm improvement design and implementation

Design of improved genetic control algorithm

According to the above analysis, it can be seen that the posture of human body is affected by the small disturbance piecewise linear error in sports training, and the human body attitude control system is a multivariable nonlinear control system in sports training. The traditional GA is used to simulate the motion and attitude control and the steady tracking error occurs in the boundary layer. In view of the disadvantages and defects of the traditional methods,¹⁷ this article presents an improved GA based on the human body posture simulation control algorithm in sports training. First, three state-space equations of human body in the virtual world of genetic space in physical training are constructed as follows

Y_{α} (u) = \sqrt{\frac{1 - j \cot α}{c^{2} - j \cot α}} e^{j \frac{u^{2}}{2} \cot α (1 - \frac{\cos^{2} α}{\cos^{2} β})} X_{β} (u \frac{\sin β}{c \sin α})

(11)

where

β = acr \tan (c^{2} \tan α)

(12)

The expression of confidence in moving position of human body in physical training is obtained as follows

P (Y) = \frac{\exp {- β \sum_{c \subset C} V_{c} (Y)}}{\sum_{Y} \exp {- β \sum_{c \subset C} V_{c} (Y)}}

(13)

According to Bayesian theorem, in order to satisfy the matching condition of GA and the error caused by parameter perturbation in physical training,¹⁸ it is necessary to design an integral term of tracking error on the synovial surface of human posture in physical training. The genetic evolutionary switching control law of human posture in physical training can be written as $u_{sw} = - K sgn (S (t) / μ)$ on the thickness of the boundary layer. In physical training, the human body system carries on the posture simulation control under the control law function and obtains the human body movement azimuth confidence in the sports training as the recursive computation expression¹⁹

\begin{matrix} P (y_{w_{3}} | x_{w_{3}}, θ, β) \propto P (y_{w_{3}} | x_{w_{3}}, θ) (y_{w_{3}} | β_{i}) \\ \propto Π_{k = 1}^{K} α_{k} \frac{1}{\sqrt{2 π σ_{k}^{2}}} \exp {- \frac{{(x_{i} - μ_{k})}^{2}}{2 σ_{k}^{2}}} \cdot \\ \frac{1}{Z (β_{i})} \exp (- \sum_{c \subset C} V_{c} (Y, β_{i})) \\ \propto Π_{k = 1}^{K} \frac{α_{k}}{Z (β_{i}) \sqrt{2 π σ_{k}^{2}}} \cdot \\ \exp {- [\sum_{k = 1}^{K} \frac{{(x_{i} - μ_{k})}^{2}}{2 σ_{k}^{2}} + \sum_{c \subset C} V_{c} (Y, β_{i})]} \end{matrix}

(14)

In physical training, human posture produces new individuals according to the quantum probability amplitude of quantum chromosomes and carries out quantum coding.²⁰ The measured data N are independent of their past measurements $z_{t}$ , and are only related to $x_{t}$ , that is

g (x_{i}, y_{j} | μ_{k}, σ_{k}^{2}) = Π_{k = 1}^{K} α_{k} \frac{1}{\sqrt{2 π σ_{k}^{2}}} \exp {- \frac{{(x_{i} - μ_{k})}^{2}}{2 σ_{k}^{2}}}

(15)

Then

\begin{matrix} P (y_{w_{4}} | x_{w_{4}}, θ, β) \propto Π_{s = 1}^{4} w (d_{ij}) P_{s} (y_{w_{3}} | x_{w_{3}}, θ, β) \\ \propto Π_{s = 1}^{4} w (d_{ij}) Π_{k = 1}^{K} \frac{α_{k}}{Z (β_{i}) \sqrt{2 π σ_{k}^{2}}} \cdot \\ \exp {- [\sum_{k = 1}^{K} \frac{{(x_{i} - μ_{k})}^{2}}{2 σ_{k}^{2}} + \sum_{c \subset C} V_{c} (Y, β_{i})]} \end{matrix}

(16)

The control law of the body posture simulation control in the physical training is obtained as

u = u_{eqx} + u_{eq θ} + u_{sw}

(17)

The control error of the whole system is

P (x_{w_{3}}, y_{w_{3}} | Θ) = \underset{x_{i} \in w_{3}}{Π} Π_{k = 1}^{K} α_{k} g (x_{ij}, y_{ij} | μ_{k}, σ_{k}^{2})

(18)

It can be seen from the above analysis that the conventional sliding mode control can overcome the shortcomings of high error and no stability in motion trajectory control using the improved genetic evolutionary algorithm.

Retrieval integral control

According to the inertial attitude parameter fusion of the human body in the position and pose state of the end effect in the working space during physical training, the observation sequence of the human body in the course of physical training is $O = {O^{1}, O^{2}, \dots, O^{k}}$ , in which $O^{k}$ is the kth observation sequence. Take $π = {π_{i}}$ as the initial state probability distribution, where $π_{i} = P (q_{1} = S_{i})$ , 1 ≤ i ≤ N, and the orientation is

\begin{matrix} {{}^{4}T}_{7} = Π_{i = 5}^{7} {{}^{i - 1}T}_{i} (q_{i}) = Π_{i = 4}^{1} {{}^{i - 1}T}_{i}^{- 1} (q_{i}) \cdot^{0} \\ T_{4} = [\begin{matrix} n_{e} & o_{e} & a_{e} & p_{e} \\ 0 & 0 & 0 & 1 \end{matrix}] \end{matrix}

(19)

The total number of degrees of freedom $n$ is 10, and its configuration can be expressed as $θ = [q_{0}^{T}, q_{1}^{T}]^{T} \equiv [θ_{1}, \dots, θ_{10}]^{T}$ , so that $q_{1} = [\begin{matrix} q_{1}, & . . ., & q_{7} \end{matrix}]^{T}$ , and $\sin q_{i}$ and $\cos q_{i}$ are re-recorded as $s_{q_{i}}$ and $c_{q_{i}}$ , respectively. The initial configuration of human body is $[\begin{matrix} θ_{1 start}, & . . ., & θ_{10 start} \end{matrix}]^{T}$ and the pose of target is $p_{obj} \in R^{6}$ in the course of physical training. Of which

\begin{matrix} a_{4} = \frac{r_{w} - p_{4}}{‖ r_{w} - p_{4} ‖}, r_{w} = p - l_{h} \times n, \\ o_{4} = \frac{a_{4} \times (r_{s} - p_{4})}{‖ a_{4} \times (r_{s} - p_{4}) ‖}, n_{4} = o_{4} \times a_{4} \end{matrix}

(20)

Given the initial configuration of the human body in the course of physical training, the initial posture of the human body $θ_{start} \in C_{free}$ , and the feasible attitude set $p_{obj}$ , the target configuration $θ_{goal}$ is unknown. According to the inertial attitude parameter fusion processing in the end effect position and pose state of human body in the working space during the course of physical training, the fuzzy constraint and inverse integral control method are used to carry out the multi-dimensional control of the physical training process, according to the above analysis. The inverse simulation of physical training is realized.

Simulation experiment and result analysis

In order to verify the application performance of this method in the process control of physical training, the simulation experiment is carried out, the model of human attitude parameter acquisition and characteristic analysis is constructed, and the signal acquisition sensor is designed. The route diagram by node resume is shown in Figure 3.

Figure 3.

Route diagram by node resume.

Three data are obtained for each sensor. The simulation scene of human body motion space in physical training is in a two-dimensional (2D) plane of 300. The starting point of human attitude distribution coordinate is [20, 20], [250,250]. We adopt the international standard data set opportunity, Activity recognition data set, which includes a large number of acceleration and radio frequency identification (RFID) tags data, a total of 250 attributes. Two accelerometer data (six acceleration attributes and one state label attribute) are selected. According to the above simulation parameters, the analysis results of human posture simulation parameters in the course of physical training are shown in Figure 4. The number of nodes detected by the network is shown in Figure 5.

Figure 4.

Simulated results of human posture during physical training.

Figure 5.

Number of nodes detected by the network.

Figure 4 shows that the human posture performance of this method is better in the process of physical training, and the accuracy of inversion control in the process of physical training is tested by different methods. The comparison of control performance is shown in Figure 6, and the analysis in Figure 4 shows that this method has better convergence, higher control quality, and better inverse simulation performance in the process of physical training. The average energy consumption for transmitting a data message is shown in Figure 7.

Figure 6.

Performance comparison of inversion simulation control.

Figure 7.

Average energy consumption for transmitting a data message.

Conclusion

In this article, a physical training process control model based on inverse integral and fuzzy control is proposed. The complex problem of high-dimensional spatial motion planning involved in physical training is decomposed into a series of sub-problems in low-dimensional space, and the inertial attitude parameter fusion is carried out according to the position and pose state of the human body in the end of the workspace during the process of physical training. The multi-dimensional control of physical training process is carried out by fuzzy constraint and inverse integral control, and the attitude parameters of human body are adjusted by means of mechanical analysis model and inertial parameter analysis method. The inversion simulation control is realized in the process of physical training, and the effect of physical training is improved. The simulation results show that the inversion simulation control has better convergence, higher control quality, and better inverse simulation performance in the process of physical training, which can effectively guide physical training and improve the effect of physical training. This model has good application value in process control and posture simulation of sports training.

Footnotes

Handling Editor: Mohamed Elhoseny

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Tao Shifang

References

Cai

. Scheduled offline model predictive control based on multiple LPV models. Contr Decis 2016; 31(8): 1468–1474.

Wan

Kothare

. Efficient scheduled stabilizing output feedback model predictive control for constrained nonlinear system. IEEE Trans Automat Contr 2004; 49(7): 1172–1177.

. PID tuning for optimal closed-loop performance with specified gain and phase margins. IEEE T Contr Syst Technol 2013; 21(3): 1024–1030.

Wang

Shao

. PID auto-tuner based on sensitivity specification. Chem Eng Res Des 2000; 78(2): 312–316.

Zhang

Sun

Zhang

, et al. Adaptive sliding mode control for re-entry attitude of near space hypersonic vehicle based on backstepping design. IEEE/CAA J Automat Sin 2015; 2(1): 94–101.

Peng

Bai

Qiao

. Static anti-windup compensation control of yaw movement for a coaxial eight-rotor UAV. Robot 2018; 40(2): 240–248.

Shi

Zhang

Yao

. A CPG-based control method for the multi-mode locomotion of a desert spider robot. Robot 2018; 40(2): 146–157.

Gao

Wang

, et al. Gait simulation of snake robot based on CPG method. J Syst Simul 2015; 27(6): 1374–1380.

Zhang

Gong

Yao

. Effects of head and tail as swinging appendages on the dynamic walking performance of a quadruped robot. Robotica 2016; 34(12): 2878–2891.

10.

Tunc

Shaw

. Experimental study on investigation of dynamics of hexapod robot for mobile machining. Int J Adv Manuf Tech 2016; 84(5–8): 817–830.

11.

Wang

Gao

Zhao

. CPG-inspired locomotion control for a snake robot basing on nonlinear oscillators. J Intell Robot Syst 2017; 85(2): 209–227.

12.

Wang

Tang

, et al. Tip-following path planning and its performance analysis for hyper-redundant manipulators. J Mech Eng 2018; 54(3): 18–25.

13.

Tang

Wang

Zheng

, et al. Design of a cable-driven hyper-redundant robot with experimental validation. Int J Adv Robot Syst 2017; 14(5): 1–12.

14.

Huang

Zhu

, et al. An efficient group recommendation model with multiattention-based neural networks. IEEE T Neur Net Lear Syst. Epub ahead of print 15 January 2020. DOI: 10.1109/TNNLS.2019.2955567.

15.

Wei

Song

, et al. Gradient-driven parking navigation using a continuous information potential field based on wireless sensor network. Inf Sci 2017; 408(2): 100–114.

16.

Dong

Guo

. Speech corpora subset selection based on time-continuous utterances features. J Comb Optim 2019; 37(4): 1237–1248.

17.

Ofodile

Turner

. Anti-windup design for input-coupled double integrator systems with application to quadrotor UAV’s. Eur J Control 2017; 38(4): 22–31.

18.

Cheng

Yip

, et al. Fuzzy logic based dynamic decision-making system for intelligent navigation strategy within inland traffic separation schemes. Ocean Eng 2020; 197: 106909.

19.

Sun

Zhang

Meng

, et al. Primary resonance analysis and vibration suppression for the harmonically excited nonlinear suspension system using a pair of symmetric viscoelastic buffers. Nonlin Dynam 2018; 94(2): 1243–1265.

20.

Huang

, et al. Multimodal representation learning for recommendation in Internet of Things. IEEE Int Things J 2019; 6(6): 10675–10685.

Research on inverse simulation of physical training process based on wireless sensor network

Abstract

Keywords

Introduction

Design of body posture environment model and analysis of constraint parameters in physical training

Human posture simulation model for physical training

Description of human body posture simulation control parameters in physical training

Algorithm improvement design and implementation

Design of improved genetic control algorithm

Retrieval integral control

Simulation experiment and result analysis

Conclusion

Footnotes

Declaration of conflicting interests

Funding

ORCID iD

References