Fuzzy Reduced Order Observer-Controller Design For Biomechanical Sit-to-Stand Movement

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Fuzzy Reduced Order Observer-Controller Design for

Biomechanical Sit-to-Stand Movement


Asif M. Mughal Kamran Iqbal
Dept. of Electrical and Computer Engineering Department of Systems Engineering
Center for Advanced Studies in Engineering University of Arkansas at Little Rock
Islamabad, Pakistan Little Rock, USA
asifm@case.edu.pk kxiqbal@ualr.edu

Abstract— Successful execution of the biomechanical sit-to- person’s functional independence and fall prevention [3].
stand (STS) task combines a forward thrust phase with an Physiologists [1-4] have often described STS task in terms of
upward extension phase and stable movement termination. We phases (forward phase, upward extension phase) and
have previously developed a fuzzy dynamic model to analyze the physiological variables such as center of pressure (CoP), center
STS task by joining two local linear models, defined at the of mass (CoM), head position (HP) and ground reaction forces
equilibrium positions, via Gaussian membership functions. The (GRF) etc. The authors in [5] have compared two rules for
local linear models were obtained from a four-segment movement coordination via muscles and body segments; these
biomechanical representation of the human body dynamics in the include threshold position control and minimal interactions.
sagittal plane. Our fuzzy controller model uses an observer to
They also simulated the STS task by using two phases similar
reconstruct velocity data from noisy observation of joint
to the phases discussed in ref [1]. The results reported in ref [6]
positions. In this study, we propose a reduced order observer
with an optimal controller design for the control of STS task. The
confirm that placing the CoM over the foot and following it
fuzzy optimal controller adjusts feedback and feedforward with upward motion makes it easier for humans to attain this
components of joint torques, where the latter are derived from a task through minimal coordination of joints.
reference trajectory. Our movement control strategy employing In our previous studies [5-7], we have used a four-segment
fuzzy reduced order observer with fuzzy controller leads to rigid-body biomechanical simulation model that consists of:
physiologically tractable simulation of the STS movement with foot segment representing the base of support (BoS), lower leg,
results that are superior to those previously obtained with full
upper leg, and the HAT segment. The model has three
order compensators.
rotational degrees of freedom (DOF) at the ankle, knee, and hip
Keywords— Biomechanical Model, sit-to-stand, reduced order joints. We may note that the assembly of the upper three rigid-
observer, fuzzy compensator body links represents a triple inverted pendulum model, which
is intrinsically unstable and consistently needs control effort in
I. INTRODUCTION the shape of joint torques to maintain a stable posture amid
The sit-to-stand (STS) movement is a simple example of a BoS perturbations. Using such models, STS movement can be
coordination task that combines the musculoskeletal structure analyzed as a point-to-point movement from a statically stable
together with the neural control and sensory mechanisms in the sitting position to the statically stable standing position.
human body. Physiologists generally describe the STS Simulation of STS movement requires that reference
movement in terms of different phases and regions of motions. trajectories be supplied at the joints. Furthermore, a control
For example, ref [1] divided the STS motion into two main system must be developed to ensure the stability of motion.
phases: a forward thrust phase and an extension phase. In the Fuzzy modeling provides a useful modeling paradigm for
forward thrust phase lifting off from chair consists of forward the computer simulation of STS movement. In our previous
flexion of head-arm-and-trunk (HAT), forward and upwards studies of STS [5-7], fuzzy model combines local linear
motion of a lower limb, accompanied by maximum flexion and models using membership functions representing the activation
dorsiflexion of hip and ankle joints, respectively. Extension of level of each local model. By using fuzzy models, the STS
hip and knee joints and plantar flexion of the ankle joint finally movement is simulated as gradual transition from one
stabilizes the body in an upright position. Ref [2] described the equilibrium state to another. The fuzzy observer combines two
determinants and different parameters of sit to stand, which are reduced order observers, designed to estimate joint velocities at
important to consider during the task, including height of the the terminal positions, with Gaussian membership functions.
chair and existence of back support, head position, arms, hip, The control effort required to generate the movement is based
knee and ankle. The authors noted the need of control variables upon the fuzzy combination of two different stability regions of
which influence the performance of STS movement. Their sitting and standing position. In order to design Fuzzy
experimental setup included force plates, motion video capture controllers, we linearize the nonlinear model presented in [5] at
techniques, optoelectronics methods and accelerometers. two stable positions i.e. the sitting and standing positions.
The authors in [3] have investigated the STS strategies via These local linear models are then integrated using fuzzy sets
the topology of an associated phase space. The ability of a and Gaussian membership functions for the analysis and
person to stand up from seat plays an important role in an aged computer simulation of biomechanical STS movement.
Fuzzy control systems use a Takagi-Sugeno-Kang (TSK)
type controller model (e.g., [8]) that remains active during the
entire trajectory of the system. Our TSK model combines the
two local linear models using Gaussian membership functions.
As a result, starting the movement from a sitting position, both l3
models stay active until the movement settles in the standing
position. The STS movement is simulated with the nonlinear I3
dynamics, a linear fuzzy controller, and a reduced order fuzzy m3g
observer based on the TSK fuzzy model of the biomechanical θ3
system. In ref [6], we presented full order fuzzy compensator
with active torque and physiological weights for compensation.
τ3 l2
In this paper, we present the design of a feedback controller
using a reduced order fuzzy observer with active (feed- I2
forward) torque component. We simulate the STS movement m2g
in the presence of noise, and obtain physiologically improved θ2
results as evaluated by head displacement, which demonstrates τ2
the applicability of the proposed scheme.
II. MATHEMATICAL MODELING
I1
The nonlinear dynamics of the four-segment biomechanical m1g l1
model (Fig.1) are symbolically described as:
D(θ)θ̈ + H�θ, θ̇�θ̇ + G(θ) = �τ⃑ (1) θ1
τ1
where 𝐷𝐷(𝜃𝜃)θ̈ represents the inertial component of the joint x
moments, 𝐻𝐻�𝜃𝜃, 𝜃𝜃̇ �θ̇ is the Coriolis and nonlinear component, x
and 𝐺𝐺(𝜃𝜃) is the gravitational component included in the rigid Fy Fx
body equations of motion. In this formulation, 𝜃𝜃𝑖𝑖 are the lf
angles from the horizontal and 𝜏𝜏𝑖𝑖 represent the torques at the
ankle, knee and hip joints. The explanation of various terms Figure. 1 Four link rigid body bio-mechanical Model
appearing in the model is given in the Appendix. In Fig. 1, 𝐹𝐹𝑥𝑥
and 𝐹𝐹𝑦𝑦 are ground reaction forces, 𝐼𝐼 is inertia of the segment, 𝑚𝑚 Next, we define a state vector for the linear model:
is the mass, and 𝑙𝑙 is the length of the segment. Equation (1) is
given as vector matrix notation where the individual matrices x�⃑ = [∆θ1 ∆θ2 ∆θ3 ∆θ̇1 ∆θ̇2 ∆θ̇3 ]T
(5)
of inertial, coriolis and gravitation component are given in the x�⃑ = [x1 x2 x3 x4 x5 x6 ]T
Appendix, and the torque vector is defined as:
where ∆θi = θi − θe , and θe is the equilibrium position. Then,
τ�⃑ = [τ1 − τ2 τ2 − τ3 τ3 ]T (2) the state equation are given as:
Each joint torque comprises of feedback and feedforward ẋ 1 = x4 ẋ 2 = x5 ẋ 3 = x6
components, i.e., τ = τfb + τff . Feedforward component τff ẋ 4 x4
ensures static equilibrium of the system and is computed from ẋ −1 x
� 5 � = D ∙ �−H ∙ � 5 � − G + �τ⃑�� (6)
a reference trajectory. Using the joint angles as velocities as ẋ 6 x6 x=θi,2
state variables, the nonlinear system of equations can be �⃑=τ
τ �⃑1,2
written in the state space form as follows: The matrices represented in the above equation are computed
θ̇ = Iθ at their respective equilibrium points. The two local linear
̈θ = D−1 (θ) ∙ �−H�θ, θ̇� − G(θ) + �τ⃑� (3) models are thus derived from the nonlinear four-segment
model of the body. The local models are defined, respectively,
Equation (3) represents a nonlinear system of equations for a at the sitting and standing positions defined by (4). These local
three-DOF biomechanical model. This system can be LTI models are each described in the state space form:
linearized using Taylor series approximation at a stationary
point. For the STS movement, we assume that sitting and ẋ (t) = Ai x(t) + Bi u(t), i = 1,2. (7)
standing positions are both equilibrium positions; so, the 1.1 Fuzzy Modeling
system is linearized at both these positions. The state variables
for linearizing at sitting and standing posture are 𝑥𝑥1𝑒𝑒 and 𝑥𝑥2𝑒𝑒 The fuzzy biomechanical model for STS movement
respectively given as: combines the local linear models developed at the sitting and
standing positions. STS movement can be interpreted as a
π π T
x1e = �2 0 0 0 0� physical system transitioning from one equilibrium state to
2 another. The control effort required to generate the STS
π π π T
(4)
x2e = �2 0 0 0� movement is based upon two different feasible stability regions
2 2
of sitting and standing position. TSK fuzzy model uses
membership functions whose value represents the activation We note that the state weighting 𝑄𝑄 and the input weighting 𝑅𝑅
level of each local model. Fuzzy If-Then rules are used to can be separately chosen for the two component models based
appropriately activate the local models. Applying weights on the requirements for control effort, bandwidth, and the
through membership function assembles all the localized desired steady state response. For example, the local model
models into a single model. computed at standing position is used for the final settling of
the steady state responses, thus we may choose Q 2 > Q1 ,
In the case of STS movement, the IF-Then rules
where Q1 and Q 2 represent the state weighting for sitting and
incorporate the measurement of knee angle 𝑥𝑥2 . It is thus
standing position, respectively, The controller gain matrix for
assumed that only the knee flexion differentiates the terminal
each individual model is given as
positions, whereas the remaining joint angles stay the same as
shown in Fig. 2. TSK fuzzy model [8] of STS movement K i = R−1 T
i Bi Mi , i = 1,2. (11)
changes only when the angular position of knee flexion
changes. The fuzzy control effort for the simulation of STS movement
is given as:
u(t) = −(w1 K1 + w2 K 2 ) ∙ x(t) (12)
A. Fuzzy State Observer
Next, we consider the role of measurement noise as well as
parameter uncertainty in the plant dynamics, arising due to
linearization and modeling errors. A fuzzy linear observer /
(a) Sitting Position (b) Standing Position estimator can be designed to estimate plant states corrupted by
Fig. 2 A Comparison of body posture at the two equilibrium positions noise. The fuzzy model with plant and measurement noise is
given by the following equations:
Let two normal and convex fuzzy sets representing SIT and
STAND position be defined on the range of 𝑥𝑥2 , where the ẋ (t) = ∑2i=1 wi �Ai x(t) + Bi u(t)� + Bv v(t)
(13)
Gaussian membership functions corresponding to the two y(t) = Cx(t) + η(t)
fuzzy sets are given as:
where 𝑣𝑣(𝑡𝑡) and 𝜂𝜂(𝑡𝑡) are uncorrelated process noise and
1− 1� measurement noise vectors. In general we assume stationary
1+exp�−7�x2 −π�4��
w1 (x2 ) = , w2 (x2 ) = 1 − w1 (x2 ), white noise processes where covariance matrices are given as:
1+exp�−7�x2 +π�4��
(8) E�v(t)v T (t + τ)� = Sv δ(τ)
(14)
The fuzzy model for sit-to-stand movement is constructed with E�η(t)ηT (t + τ)� = Sη δ(τ)
the help of following rules: The white noise processes and their corresponding, 𝐵𝐵𝑣𝑣 and C
RULE 1: matrices are the same for each local linear model. Then, using
separation principle, the combined controller-estimator
IF 𝑥𝑥2 is SIT, THEN 𝑥𝑥̇ (𝑡𝑡) = 𝐴𝐴1 𝑥𝑥(𝑡𝑡) + 𝐵𝐵1 𝑢𝑢(𝑡𝑡) equations for the model are given as:
RULE 2: x�̇ (t) = ∑2i=1 wi �Ax� (t) + Bi u(t) + Li y(t)�
(15)
IF 𝑥𝑥2 is STAND THEN 𝑥𝑥̇ (𝑡𝑡) = 𝐴𝐴2 𝑥𝑥(𝑡𝑡) + 𝐵𝐵2 𝑢𝑢(𝑡𝑡) u(t) = −Kx� (t)
The fuzzy model as a function of the membership weights is where the fuzzy observer gains are computed via the following
given as: algebraic Riccati equations:

ẋ (t) = A x(t) + B u(t) Pi Ai + ATi Pi − Pi CSv−1 CMi + FSη F T = 0 (16)


. (9)
= ∑2i=1 �wi �Ai x(t) + Bi u(t)�� Li = Pi C T Sη−1 (17)
B. Tracking Model
III. FUZZY OPTIMAL CONTROLLER
The linear fuzzy model with the regulated control effort
In this section we explore the optimal controllers for the incorporates stable closed-loop dynamics of the system, which
STS transfer of the fuzzy biomechanical model. settle to zero reference (regulation). A reference trajectory is
In the controller design, full state availability is assumed now added to obtain a desired movement and a stable
and closed-loop pole placement is achieved by obtaining a (standing) position. This reference trajectory is not fuzzy as a
solution of a single Riccati equation by principal of separation single reference path is required to reach the desired position.
of variables. Both local linear models (i=1,2) are subjected to The reference input is added to the fuzzy dynamical model as:
linear quadratic regulator (LQR) design for calculation of ẋ (t) = ∑2i=1 wi �Ai x(t) + Bi u(t)� + E ∙ xr (18)
optimal control gains. The accompanying Riccati equations for
the individual models are given as The reference trajectory model for the desired response is
given as:
Mi Ai + ATi Mi − Mi Bi R−1 T
i Bi Mi + Q i = 0, i = 1,2 (10)
ẋ r = Ar xr (19) The measured angles are given as 𝑥𝑥⃑𝑚𝑚 vector and indirectly
observed states (velocities) are given as 𝑥𝑥⃑𝑜𝑜 vector. The state
We note that 𝐴𝐴𝑟𝑟 must be open loop stable matrix. The error in space dynamics for both variables are assumed as:
state variables is defined as 𝑒𝑒(𝑡𝑡) = 𝑥𝑥𝑑𝑑 − 𝑥𝑥(𝑡𝑡). Combining the
reference and the error states, the meta-state model for the ẋ m (t) = Am ∙ �x⃑m (t) + Amo ∙ �x⃑o (t) + Bm ∙ u(t)
system is defined as: (24)
ẋ o (t) = Aom ∙ �x⃑m (t) + Ao ∙ �x⃑o (t) + Bo ∙ u(t)
ė A E e B The standard observation is given as:
� �=� � ∙ � � + � � ∙ u(t) (20)
ẋ r 0 Ar xe 0
y(t) = Cm ∙ �x⃑m (t) (25)
In (20), the state vector 𝑒𝑒 is the error from the instantaneous
position to the desired/final position 𝑥𝑥𝑑𝑑 . In the case of a This implies that
constant reference, the reference trajectory matrix 𝐴𝐴𝑟𝑟 = 0, and
x�⃑m (t) = Cm −1 ∙ y(t) (26)
only the error model can be used to solve the tracking problem.
Stable error dynamics are necessary to control the nonlinear The standard observer will only be designed for states not
plant using linear fuzzy controller, so that the error goes to zero measured and indirectly observed, i.e.,
in moving from initial to the final position. The reference
trajectory also changes the control effort. The additional x�̇ o (t) = Aom ∙ �⃑x� m (t) + Ao ∙ �⃑x� o (t) + Bo ∙ u(t) + Lo y(t)
control effort is also fuzzy for the non-fuzzy reference (27)
trajectory, where the additional gain matrix for the reference
states is computed as: The general structure for estimation will be of form:
K r = w1 K r1 + w2 K r2 �⃑x� o (t) = Lo y(t) + z⃑ (t) (28)
(21)
K ri = −R−1 T −1
i Bi (A i − Bi K i ) Mi E rd
Where 𝑧𝑧⃑(𝑡𝑡) is a 3 order vector for estimation of velocities
The resulting u(t) changes to: 𝑥𝑥�⃑𝑜𝑜 (𝑡𝑡), and the three angle measurements are available by (23).
The 3rd order observer system is defined as
u(t) = − ∑2i=1(wi [K i ∙ x(t) + K ri ∙ xr ]) (22)
This additional effort is not required if the reference trajectory ż⃑ (t) = F ∙ z⃑ (t) + J ∙ y(t) + N ∙ u(t) (29)
matrix 𝐴𝐴𝑟𝑟 = 0 , i.e,, when the reference position is the The error dynamic is only due to the observed states, given as:
stationary position, and only error from the initial and desired
final positions is used for the error state model. The tracking ė o = ẋ����⃑o − x�̇ o (t)
���⃑ (30)
error goes to zero as the plant (biomechanical model) achieves
its final position. The error in observed states goes to zero. Further, we define
the vector 𝑧𝑧⃑(𝑡𝑡) as the observer state variables:
C. Reduced Order Fuzzy Observer
The separation principle also applies to the system when z⃑ (t) = �⃑x� o (t) − Lo y(t)
(31)
full state measurements are not available. In (15), we described = �x⃑o (t) − �e⃑o (t) − Lo Cm ∙ �x⃑m (t)
a full state observer which estimates both angle and velocity Using (31), error dynamics leads to following simplifications
profiles. The fuzzy model is based upon angular differences
and initial and final velocities at both positions are assumed as ���⃑
ė o = F ∙ e�⃑o (t) + (Aom + Lo Cm Am − JCm + FLo Cm ) ∙ �x⃑m (t) +
zeros. In [5-6], we described the forward thrust phase with (Ao + Lo Cm Amo − F) ∙ �x⃑o (t) + (Bo − Lo Cm Bm − N) ∙ u(t)
initial velocities, whereas we now assume that velocities are (32)
not measured in the modeling scheme. In both of these cases as
well as in practical data collection for the biomechanical These matrices are chosen such that the errors in the observed
system, it is difficult to obtain velocity sensor data from the states go to zero and the overall system should be
joints. It is thus necessary that velocity states should be asymptotically stable. This implies the following choices of the
estimated and angle measurement must be used without matrices:
estimation. This scheme will lead to reduced order observer F = Ao + Lo Cm Amo
design of the system and its detailed discussion is given in [9]. N = Bo − L o Cm Bm (33)
This also requires designing two reduced order observer by JCm = Aom + Lo Cm Am + FLo Cm
combining through same membership function.
The LTI models at sitting and standing positions with only
Let us define two sets of state variables, one set which is angular measurements requires reduced order observer
directly measured and a second set, which are observed. For matrices in (33). After calculating these matrices for both
both sitting and standing positions, these variables are sitting and standing position, the results are combined with
independently given as: Gaussian membership function of (8), and are given as:
�x⃑m (t) F = w1 F1 + w2 F2
x1.2 = � … . �
������⃑(t) (23) N = w1 N1 + w2 N2 (34)
�x⃑o (t) J = w1 J1 + w2 J2
IV. STABILITY ANALYSIS settle close to 2 sec at their steady state values, which
The closed-loop stability of the fuzzy biomechanical STS corresponds to the average time for healthy humans from
model depends upon the combined stability of all local linear forward thrust to the termination of STS movement. Velocity
models. The fuzzy biomechanical model is stable if it satisfies estimates similarly reach their zero steady-state values within
the following two conditions for some positive definitive 2sec without reference trajectories.
matrix 𝑃𝑃 [8]. Fig 5 presents the total joint torques, as well as the active
T torque component based on the assumed reference trajectories.
Acl T cl
i P + P A i < 0 i = 1,2 There are higher feedback torques generated in the beginning,
T compared to the active component, indicating greater
G12 P + P T G12 < 0 (35)
contribution from the reflex circuits in the body, but both
where components settle at zero values as desired.
G12 = (A1 − B1 K 2 − L1 C2 ) + (A2 − B2 K1 − L2 C1 ) Fig 6 shows the profiles of normalized x-COM, y-COM at
Acl foot length (lf) and at body length (l1+l2+l3+b) respectively.
i = A i − Bi K i − L i Ci (36)
The second plot shows the profile of ground reaction forces 𝐹𝐹𝑥𝑥
In the above, Acl are the closed-loop systems with controller and 𝐹𝐹𝑦𝑦 , which are seen to settle at zero and the body weight,
and estimator gains, and 𝐺𝐺12 is a cross gain matrix correlating respectively. The head movement profile shows that, during
both local linear models. These stability conditions are also STS, there is an initial dip in the head positon, followed by an
necessary with the reduced estimator design of the system. upward stretch, and finally settling together with the COM.
Thus, F should be stable during the entire trajectory implying
F1 and F2 both should have eigenvalues in the stable region. VI. CONCLUSION
V. SIMULATION OF STS TASK In this paper, we presented a reduced order observer based
optimal controller for TSK fuzzy modeling of biomechanical
We simulated the STS movement using the nonlinear STS movement. This observer model estimates joint velocities
biomechanical model (1), regulated with fuzzy LQR controller from movement data in the presence of measurement noise.
and fuzzy reduced order observer (22),(32). The MATLAB The fuzzy modeling scheme is relevant to physiologists’
simulation scheme built in Simulink is shown in Fig. 3. We studies (e.g., [1]), which describe the STS motion as a
synthesized linear reference trajectories from the reference combination of a forward thrust phase and an upward extension
model (19). The reference trajectories were used to generate phase. The TSK fuzzy model with Gaussian membership
the active (feedforward) input component, which was added to functions further elaborates the physiological relevance of
the feedback component generated from the reduced order biomechanical STS. Finally, the reduced order fuzzy observer
observer and controller gains (22). proposed in this study results in computer simulation of STS
Fig 4 shows the angular profiles (solid blue lines) of ankle, movement that is closer to physiological relevance and
knee and hip joints with comparison of synthesized reference experimental results. The simulation mimics an experimental
trajectories for the joints (red dashed lines). These responses scenario where motion data is collected by attaching reflective
markers to the limbs, but no velocity data are collected.

Figure 3. Simulation of nonlinear model with linear fuzzy LQR and reduced order observer with Gaussian membership weights
coordination task. Additional physiological issues such as
Ankle Knee Hip
Angular Profiles 2 2 2.5 neural transmission delays due to feedback latencies, angular
and velocity limits, shifted center of mass, etc., may be studied
1.5 1 2 in this paradigm and further analyzed. Finally, fuzzy modeling
is proposed for further studies involving analysis of sagittal and
1 0 1.5 frontal plane synergistic movements.
0 1 2 0 1 2 0 1 2
Actual
Ref APPENDIX:
2 10 5
The coefficient matrices appearing in the nonlinear dynamic
Velocity Profiles

0 5 equation (1) of the four-segment model are given as:


0
-2 0 𝑑𝑑11 𝑑𝑑12 cos(𝜃𝜃1 − 𝜃𝜃2 ) 𝑑𝑑13 cos(𝜃𝜃1 − 𝜃𝜃3 )
-4 -5 -5 𝐷𝐷 = �𝑑𝑑12 cos(𝜃𝜃1 − 𝜃𝜃2 ) 𝑑𝑑22 𝑑𝑑23 cos(𝜃𝜃2 − 𝜃𝜃3 )�
0 1 2 0 1 2 0 1 2 𝑑𝑑13 cos(𝜃𝜃1 − 𝜃𝜃3 ) 𝑑𝑑23 cos(𝜃𝜃2 − 𝜃𝜃3 ) 𝑑𝑑33
time time time

0 𝑑𝑑12 𝜃𝜃5 sin(𝜃𝜃1 − 𝜃𝜃2 ) 𝑑𝑑13 𝜃𝜃6 sin(𝜃𝜃1 − 𝜃𝜃3 )


Fig 4. Angular profiles (rads) (blue lines with reference trajectories 𝐻𝐻 = �−𝑑𝑑12 𝜃𝜃4 sin(𝜃𝜃1 − 𝜃𝜃2 ) 0 𝑑𝑑23 𝜃𝜃6 sin(𝜃𝜃2 − 𝜃𝜃3 )�
(red lines) and velocities profile at ankle, knee and hip joints. −𝑑𝑑13 𝜃𝜃4 sin(𝜃𝜃1 − 𝜃𝜃3 ) −𝑑𝑑23 𝜃𝜃6 sin(𝜃𝜃2 − 𝜃𝜃3 ) 0

Ankle Knee Hip 𝐺𝐺 = 𝑔𝑔[𝑓𝑓1 cos 𝜃𝜃1 𝑓𝑓2 cos 𝜃𝜃2 𝑓𝑓3 cos 𝜃𝜃3 ]
800 800 600

700
600 400
The parameter values used for computing the above
600
Total
mathematical model are given below:
500 400 Active
200
TABLE 1
400 Parameter values for biomechanical four-link model terms
Torques

200 0
300 Inertial components [d11 d12 d13 [11.2087 9.6933 5.9652
200 0 -200 (N s2) d22 d23 d33] 9.3593 5.9652 11.8305]
Gravitational
100 [f1 f2 f3] [26.45 22.54 13.87]
-200 -400 components (N m)
0 2
Gravity g (m/sec ) g 9.8
-100
0 1
-400
2 0 1 2
-600
0 1 2 mass of model (kg) m 64.09
time

Fig 5. Total torques (N-m) feedback including both components


feedback and feedforward (blue lines) and feedforward (active) REFERENCES
component (red lines) for stable STS movement. [1] M. W. Rodosky, T. P. Andriachhi, and G. B. Andersson, “The
Influence of chair height on lower limb mechanism during rising.” Journal of
Orthopedic Res, 7, (1989), pp266-271.
CoM
3 Head Position [2] Wim G.M. Janssen, Hans B.J. Bussmann, and Henk J Stam,
Normalized Magnitude

2.5 “Determinants of sit-to-Stand Movement: A Review,” Physical Therapy.


2 x com Volume 82, Number 9, (2002).
y com 2 [3] Patrick D. Roberts and Gin McCollum, “Dynamics of sit-to-stand
1 movement,” Biological Cybernetics. 74, (1996), pp147-157.
1.5 [4] K. Brian, “Evaluation of a generalized model of human postural
0
dynamics and control in the sagittal plane.” Biological Cybernetics. 61,
y- direction

0 1 2
1 (1989).
GRF
2000 [5] Asif Mahmood Mughal and Kamran Iqbal, “Fuzzy Optimal Control of
Fx
0.5 Sit-To-Stand Movement in a Biomechanical Model”, Journal of Intelligent
Fy and Fuzzy Systems, Vol. 25, No. 1, pp247-258, 2013
Magnitude

1000

0 [6] Asif Mahmood Mughal and Kamran Iqbal, “Physiological LQR design
0 for postural control and coordination of sit-to-stand movement,” Cognitive
Computation, Vol. 4, No. 2, pp549-562, 2012
-0.5
-1000
0 1 2
-0.5 0 0.5 [7] Asif Mahmood Mughal, Sunbal Perviaz, and Kamran Iqbal, “LMI
x directon
time (sec) based physiological cost optimization for biomechanical STS transfer,” IEEE
Conf. on Systems, Man and Cybernetics, Anchorage, Alaska, USA, 9~12 Oct
Fig 6. Normalized x-CoM, y-CoM, ground reaction forces Fx Fy and 2011.
head movement for STS movement. [8] Stanislaw. Zak, Systems and Control, Oxford University Press, (2003),
pp421-446.
The fuzzy observer based controller scheme presented here [9] Bernard Friedland, Control System Design McGraw Hill Inc, (1986),
advances the previous studies [5-7], and provides a framework pp276-286.
for physiologically relevant modeling of a movement

You might also like