See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/289261796

Dynamics and Control of the Stewart Platform

Article  in  Doklady Physics · October 2014

DOI: 10.1134/S102833581409002X

CITATIONS READS
5 392

10 authors, including:

D.M. Kostygova Nikolay Vladimirovich Kuznetsov

5 PUBLICATIONS   17 CITATIONS   
Saint Petersburg State University
244 PUBLICATIONS   5,477 CITATIONS   
SEE PROFILE
SEE PROFILE

Petr E. Tovstik
Saint Petersburg State University
204 PUBLICATIONS   653 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chaos Theory and Applications (CHTA) View project

14th International Conference "Dynamical Systems - Theory and Applications" (DSTA 2017) - December 11-14, 2017, Lodz, Poland View project

All content following this page was uploaded by Nikolay Vladimirovich Kuznetsov on 27 December 2017.

The user has requested enhancement of the downloaded file.

ISSN 10283358, Doklady Physics, 2014, Vol. 59, No. 9, pp. 405–410. © Pleiades Publishing, Ltd., 2014.
Original Russian Text © G.A. Leonov, S.A. Zegzhda, S.M. Zuev, B.A. Ershov, D.V. Kazunin, D.M. Kostygova, N.V. Kuznetsov, P.E. Tovstik, T.P. Tovstik, M.P. Yushkov, 2014,
published in Doklady Akademii Nauk, 2014, Vol. 458, No. 1, pp. 36–41.

MECHANICS

Dynamics and Control of the Stewart Platform

Corresponding Member of the RAS G. A. Leonov, S. A. Zegzhda, S. M. Zuev,
B. A. Ershov, D. V. Kazunin, D. M. Kostygova, N. V. Kuznetsov,
P. E. Tovstik, T. P. Tovstik, and M. P. Yushkov
Received April 29, 2014

DOI: 10.1134/S102833581409002X

1. INTRODUCTION The platform orientation is determined by the

position of the point O (the pole)
The Gough–Stewart platform [1] is used in design
ing dynamic stands for teaching and training pilots 0
O 0 O = r ( t ) = x 0 ( t )i 0 + y 0 ( t )j 0 + z 0 ( t )k 0 (1)
and drivers and also for failsafe testing of onboard sys
tems of planes and flying devices. and three consecutive angles of rotation around the
pole through the yaw angle (ψ), the pitch angle (θ),
The kinematics of the Stewart platform was studied
and the roll angle (ϕ). The rotation tensor P(ψ, θ, ϕ) is
in [2–4], and the algorithm of finding the set of
achievable positions of the platform was proposed in ⎛ p p p ⎞
[5]. For investigating the platform dynamics, the ⎜ 11 12 13 ⎟
Newton–Euler formalism was used in [6, 7], the P = ⎜ p 21 p 22 p 23 ⎟
Lagrange formalism was used in [4, 8], and the special ⎜ ⎟
form of the equations of solid motion [10] was used in ⎝ p 31 p 32 p 33 ⎠
(2)
[9]. The problems of stability of the Stewartplatform
motions are discussed in [8, 11]. Because of the insta ⎛ C C –S C +C S S S S +C S C ⎞
⎜ ψ θ ψ ϕ ψ θ ϕ ψ ϕ ψ θ ϕ ⎟
bility of such motions, highgrade operation of the = ⎜ Sψ Cθ Cψ Cϕ + Sψ Sθ Sϕ – Cψ Sϕ + Sψ Sθ Cϕ ⎟,
stand can be provided only by introducing a feedback. ⎜ ⎟
The method of construction for the platform with ⎝ –Sθ Cθ Sϕ Cθ Cϕ ⎠
three rods is specified in [12].
The proposed study is a development of [13]; the
Stewartplatform motion is controlled with the help of (a) (b)
ψ z υk0
six pneumatic cylinders. The effect of a delay in the C υkt0
control is considered. rc
k ekt0
i j
2. STEWARTPLATFORM KINEMATICS x υkn0 0
ekb
O Ak
ϕ ak 0
ekn
We introduce the motionless system of coordinates
O0x0y0z0 with the unit vectors i0, j0, k0, and the mobile
Ak y θ
system of coordinates Oxyz with the unit vectors i, j, k, z0 pI
rigidly connected with the platform (Fig. 1a). Six rods lk Ck
of variable length BkAk, k = 1, …, 6, simulate the pneu r0 lk0
matic cylinders (Fig. 1b) and are attached with the k0
help of spherical hinges by the ends Bk to the motion j0 pII
less base and by the ends Ak (k = 1, …, 6) to the plat Oo y0
form. It is necessary to obtain a set motion of the plat i0 bk0
form due to a change in the lengths of rods. Bk
Bk
x0

St. Petersburg State University, St. Petersburg, 198504 Russia Fig. 1. Kinematics of (a) the stand and (b) the pneumatic
email: leonov@math.spbu.ru, nkuznetsov239@mail.ru cylinder.

405
406 LEONOV et al.

where we designated for brevity Cϕ = cosϕ, Sθ = sinθ, 3. EQUATIONS OF PLATFORM MOTION

etc. We present the Poisson formulas for the derivative The equation of motion of the center of gravity C of
of the tensor P with respect to time [14] the platform and the equation of moments with
respect to the point C in the motionless system of
P· = ω × P, ω = ω i + ω j + ω k
0 0 0 0 0
x 0 y 0 z 0
coordinates have the form
= ω x i + ω y j + ω z k,
m ( r·· + ω· × r c + ω × ( ω × r c ) ) + mgk 0 = F
0 0 0 0 0 0 0
0 ·
ωx = ϕ· cos θ cos ψ – θ sin ψ, (3) 6

∑F e
0
0 · = k kt ,
ω y = ϕ· cos θ sin ψ + θ cos ψ,
k=1
0
ω z = ψ· – ϕ· sin θ, (9)
0
P ⋅ J c ⋅ P ω· + ω × ( J c ⋅ P ω ) = M
T 0 T 0 0

where ω0 is the angular velocity of the platform rota 6

∑ F (a
0 0 0 0
tion. We introduce the vector of generalized coordi = k k – r c ) × e kt , rc = P ⋅ rc ,
nates determining the platform position
k=1
q = { x 0, y 0, z 0, ϕ, θ, ψ } = { q 1, q 2, q 3, q 4, q 5, q 6 }. (4) where m and Jc are the platform mass with a load and
If values (4) are set, the pneumaticcylinder lengths its tensor of inertia with respect to the point C, g is the
0
0
lk = lk(qj) and their directions e kt are determined from acceleration of gravity, r·· is the acceleration of the
point O, and Fk are the forces acting with the platform
explicit formulas
on the part of rods. As F0 and M0, we designate the
0 0 0 0
B k A k = l k = l k e kt = r + P ⋅ a k – b k , main vector of forces Fk and their main moment with
(5) respect to the point C. The set of Eqs. (9) together with
k = 1, …, 6, Eqs. (8) has the 12th order and describes the motion of
0 the platform for the set forces.
where the vectors ak = OA k and b k = O 0 B k are
shown in Fig. 1a. On the contrary, if lk are set, the val If the motion q(t) of the platform is set, the vectors F0
ues of qj are determined from six nonlinear equations and M0 become known, and the forces Fk necessary for
obtaining this motion are found from Eqs. (9), which
l k = l k ( q j ), k, j = 1, …, 6. (6) can be written in the form
T 0 T
After the differentiation of Eqs. (6) with respect to A ⋅F = Ᏺ , F = { F 1, …, F 6 } ,
time, we obtain the set of linear equations relative to (10)
0 0 0 0 0 0 0 T
0
r· , ω0, which we write in the form Ᏺ = { F x, F y , F z , M x, M y , M z } ,

0 0 0 0 0 T
where the matrix A is the same as in Eq. (7).
A ⋅ V = i, V = { x· 0, y· 0, z· 0, ω x, ω y, ω z } ,
(7)
T
l = { l 1, l 2, l 3, l 4, l 5, l 6 } , 4. NUMERICAL EXAMPLE
where the matrix A is composed of the vector lines We consider the Stewart platform, the hinges Ak
0 0 and Bk of which are symmetrically arranged on the
Ak = { e kt , P · ak × e kt }, the superscript T designates
mobile and immobile circles of the radii Ra and Rb,
transposing, and the dot is the derivative with respect
respectively, occurring in the planes Oxy and O0x0y0.
to time.
· Let the minimal distance between the upper and
After solving set (7), the derivatives ϕ· , θ , and ψ· lower hinges be d (for example, A2A3 = B1B2 = d), and
are found from the formulas the angle at the rotation through which we obtain the
0 0 former arrangement of the hinges be 2π/3 (Fig. 2). We
ω y sin ψ + ω x cos ψ · accept that the distance between the planes of the
ϕ· = 
0 0
, θ = ω y cos ψ – ω x sin ψ,
cos θ (8) upper and lower hinges is h. We consider that, in the
initial position, the center of gravity C of the platform
ψ· = ω z + ϕ· sin θ,
0
and a solid fastened on it are above the centers of the
and the values of qi(t) are obtained by the integration. circles, and zc > 0.
The vanishing of the determinant of the matrix Α Let Ra = 0.7608 m, Rb = 1 m, d = 0.2 m, h = 1.0196 m,
testifies to the exit on the boundary of the controllabil zc = 0.8 m, l = 1.255 m, and mg = 104 N. We assume in
ity region. addition that the tensor of inertia of the platform is

DOKLADY PHYSICS Vol. 59 No. 9 2014

 DYNAMICS AND CONTROL OF THE STEWART PLATFORM 407

spherical, Jc = mρ2E, where E is the unit tensor. We y

take ρ = 0.8 m.
B3
In the equations of kinematics and dynamics, we B4
pass to dimensionless variables. For the unit of length,
we accept the value of Rb, and for the unit of force, the
weight mg of the loaded platform. The dimensionless A3
A2
time is introduced as ˜t = ωt ⎛ ω2 = g ⎞ . Further the Ra
⎝ Rb ⎠
B2
sign “~”, as a rule, is omitted. A4
A5 O x
B1
5. LINEARIZATION OF EQUATIONS Rb
OF PLATFORM MOTION A1
It was shown that the position of equilibrium of the A6
platform under the action of the set forces Fk is unsta d
ble. Therefore, below we introduce additional forces
B5
in the form of feedbacks in the displacements and B6
velocities of rods.
We consider small oscillations of a symmetric plat
form under the assumptions made in Section 4. As a Fig. 2. Stewart platform in the projection to the horizontal
plane.
reference position, we take that of equilibrium in
which all generalized coordinates are zero, and the
lengths of rods and the forces in them are constant: where the nonzero elements of the constant matri
q = 0, l k = l 0 = 1.255, ces B, C, and the vector f are equal to
(11)
F k = F 0 = 0.205, k = 1, …, 6. b 11 = b 22 = b 33 = 1, b 44 = b 55 = b 66 = ρ + z c ,
2 2

We introduce the perturbations sk of lengths of rods 2

b 15 = b 51 = ρ z c ,
2
b 24 = b 42 = – ρ z c , (15)
(the displacements of pistons) from the formula lk =
l0 + sk. For small q = (q1, …, q6)T and s = (s1, …, s6)T, c 44 = c 55 = z c , f 3 = 1.
the relation between them is given by the formula
In Eq. (14), the vector f err includes the nonlinear terms
⎧ ∂l k ⎫ of set (9).
s = A 0 ⋅ q, A 0 = ⎨  ⎬ = lim A, (12)
⎩ ∂q j q = 0⎭
q→0 Let the program motion q p(t) be set. From Eq. (12),
we find s p = A0 · q p. Strictly speaking, the value of sp
where the matrix A is the same as in Eq. (7). should be determined as was described in Section 2.
The matrix A0 for the sizes of the stand presented The distinction arising in this case can be included
in Section 4 is as follows: in f err. We determine the program value of F = Fp from
set (14) for f err = 0 and q = qp. Now we can write
A0
p
B ⋅ q·· – C ⋅ q + f = A 0 ⋅ F .
p T p
(16)
⎛ – 0.424 – 0.401 0.812 – 0.490 – 0.377 – 0.442 ⎞
⎜ ⎟
⎜ – 0.424 0.401 0.812 0.490 – 0.377 0.442 ⎟ Below it is shown that the accuracy of setting Fp
⎜ 0.559 – 0.166 0.812 0.571 – 0.236 – 0.442 ⎟ poorly affects the resulting motion. Moreover, it is
= ⎜ ⎟. possible to consider that Fp = 0 if certain restrictions
⎜ – 0.136 – 0.568 0.812 0.081 0.612 0.442 ⎟
⎜ ⎟ on the control are fulfilled.
⎜ – 0.136 0.568 0.812 – 0.081 0.612 – 0.442 ⎟
⎝ 0.559 0.166 0.812 – 0.571 – 0.236 0.442 ⎠ With the purpose of stabilization of motion, we add
(13) the controlling forces Fp to the forces Fc putting in
For small q, we write the equations of motion of the Eq. (14) that F = Fp + Fc and introduce the differences
platform singling out the linear terms in them: of the real and program motions
p p
B ⋅ q·· – C ⋅ q + f + f = Ᏺ = A ⋅ F,
err 0 T
(14) δq = q – q , δs = s – s , δs = A 0 ⋅ δq. (17)

DOKLADY PHYSICS Vol. 59 No. 9 2014

408 LEONOV et al.

The control forces Fc are introduced from the for We consider the effect of a delay in the control writ
mula ing Eq. (20) and introducing the constant delay τ in
the control:
F = – Gδs – G f δs· ,
c
(18)
δs··( t ) – C * ⋅ δs ( t ) + A * ⋅ ( Gδs ( t – τ ) + G f δs· ( t – τ ) )
where the constants G ≥ 0 and Gf ≥ 0 are subject to (26)
err
choice from the condition of minimization of the vec + A err ⋅ f = 0.
tor δs. The difference of Eqs. (14) and (16) gives The delay is naturally related to the time, which is nec
–1 –1 essary for the pneumatic cylinders for formation of the
B ⋅ ( A 0 ) ⋅ δs·· – C ⋅ ( A 0 ) ⋅ δs
(19) control pressure corresponding to Eq. (18).
+ A 0 ⋅ ( Gδs + G f δs· ) + f = 0.
T err
Considering the delay as small, we introduce into
Eq. (26) the expansion
2
τ
6. INVESTIGATION OF Eq. (19) δs ( t – τ ) = δs ( t ) – τδs· ( t ) + δs··( t ). (27)
2
We rewrite Eq. (19) as
Then the characteristic equation for Eq. (26) takes the
δs·· – C * ⋅ δs + A * ⋅ ( Gδs + G f δs· ) + A err ⋅ f = 0, (20)
err
form
2
where det ( λ ( E – kA * ) + λA * ( G f – τG ) + GA * – C ) = 0, (28)
2
–1
C* = A0 ⋅ B ⋅ C ⋅ ( A0 ) ,
–1 –1 T
A* = A0 ⋅ B ⋅ A0 , τ G . It suffices to fulfill three condi
where k = τGf – 
(21) 2
–1
A err = A 0 ⋅ B . tions for the asymptotic stability: the previously estab
lished condition G > 0.706, the inequalities Gf > τG,
For f err = 0, the zero solution of Eq. (20) is asymp and the positive definiteness of the matrix E – kA∗.
totically stable if all roots λ the characteristic equation
The last condition is fulfilled at
2
det ( λ E – C * + A * ( G + λG f ) ) = 0 (22) 1 ,
k <  (29)
have negative material parts. ρ max
The matrix C∗ is symmetric positive, and the where ρmax is the largest eigenvalue of the matrix A∗.
matrix A∗ is symmetric and positively defined. From For the parameters under consideration, ρmax = 3.956.
here at once it follows that the zero solution is unstable For the coefficient Gf, we obtain a bilateral estimate
in the absence of a control (G = Gf = 0), is stable at a τG 1
reasonably large G, and is asymptotically stable at Gf > 0. τG < G f <  + . (30)
2 τρ max
We present the numerical results for the platform with
the above parameters. The lefthand side of inequality (30) is less than the
righthand side if
At G = Gf = 0, Eq. (22) has an eightfold zero root
and two pairs of roots λ = ±0.887, which points to an 2 .
G <  (31)
2
instability. The stability comes at G > 0.706, and the τ ρ max
mode of undamped oscillations takes place at Gf = 0.
The last inequality imposes the restriction on the coef
Let G Ⰷ 1 and Gf ~ 1. Then the main terms in ficient G in the presence of a delay.
Eq. (19) are
T
GA 0 ⋅ δs + f
err
= 0. (23) 7. CERTAIN NUMERICAL RESULTS
We consider the dynamics of a symmetric con
From here it follows that the estimate is valid in the trolled platform with the parameters described in Sec
absence of resonances: tion 4. The effect of the delay is disregarded. We con
err sider the model of complex motion of a solid partici
f
δs ∼  , (24) pating in one translational and two rotational motions.
G As the program motion, we take
where the vector norm is determined by the relation p t p
q 1 ( t ) = a sin νt, q 5 ( t ) = a sin νt, q 6 ( t ) = a sin νt,
s = max s k ( t ) . (25) p p p (32)
t, k q 2 ( t ) = q 3 ( t ) = q 4 ( t ) = 0,

DOKLADY PHYSICS Vol. 59 No. 9 2014

 DYNAMICS AND CONTROL OF THE STEWART PLATFORM 409

Table 1. Dependence η(ν)

No. a G Gf ν=1 2 3 4 5
1 0.1 100 5 0.010 0.016 0.027 0.042 0.061
2 0.1 30 2 0.034 0.055 0.094 0.152 0.231
3 0.1 10 1 0.107 0.183 0.331 0.556 0.911
4 0.1 3 0.3 0.476 1.043 2.660 2.377 1.811
5 0.2 100 5 0.010 0.016 0.027 0.042 0.062
6 0.2 30 2 0.033 0.054 0.094 0.155 0.239
7 0.2 10 1 0.107 0.183 0.341 0.581 0.962
8 0.2 3 0.3 0.482 1.094 3.074 2.417 1.865
9 0.2 30 1 0.033 0.055 0.096 0.163 0.266
10 0.2 30 5 0.033 0.052 0.084 0.126 0.174

Table 2. Dependence η(ν) at zero program forces

No. a G Gf ν=1 2 3 4 5
5* 0.2 100 5 0.015 0.019 0.029 0.045 0.065
6* 0.2 30 2 0.049 0.065 0.105 0.167 0.252
7* 0.2 10 1 0.153 0.225 0.391 0.643 1.028
8* 0.2 3 0.3 0.781 1.6329 3.316 2.500 1.892

where a is the set amplitude of linear and angular According to approximate formula (24), the func
oscillations (for simplicity, we consider the amplitudes tion η(ν) decreases approximately as 1/G with increas
of linear and angular oscillations in dimensionless ing G. At G ~ 3, the accepted way of control is inappli
variables as identical). cable (a more exact set of forces Fp is required).
We integrate the set of Eqs. (14) calculating the vec The function η(ν) considerably increases with fre
tor of the program displacements sp of pistons from quency ν.
exact formulas (5) and the program forces Fp in rods As the comparison of lines 1–4 and 5–8 shows, for
from approximate set (16) in which only linear terms relatively small values of η(ν) (for example, η(ν) <
are retained. 0.5), this function depends only weakly on the ampli
We estimate the quality of the control with the rel tude a. The dependence manifests itself only in the
ative error region where the described way of control is inapplica
ble.
δq ,
η =  δq = q – q ,
p
q
p
= a, (33)
p
The dependence on the parameter Gf follows from
q the comparison of line 6 with lines 9 and 10. With
where the norm of the vector is determined by Eq. (25). increasing Gf, the function η(ν) decreases, and is more
considerable for higher frequencies ν.
We investigate the quality of the control in the
dependence of the feedback parameters G and Gf on The data in Table 1 are obtained under the assump
the oscillation amplitude a of the program motion and tion that the initial conditions for the real and program
on the oscillation frequency ν ≤ 5 (in the dimensional motions coincide, i.e.,
time, the frequency is limited by 2.5 Hz). First, we
q k = 0, k = 1, …, 6, q· 1 = q· 5 = q· 6 = νa,
consider that the initial conditions for the real and (34)
program motions coincide. q· 2 = q· 3 = q· 4 = 0 for t = 0,
In Table 1 for two values of the amplitude a and five
values of the oscillation frequency ν, we listed the val while the program forces in rods are calculated by the
ues of the function of quality of the control in depen method described.
dence on the feedback parameters G and Gf. The data However, it is technically difficult to provide the
in Table 1 enable us to make the following conclusions. implementation of the nonuniform initial conditions

DOKLADY PHYSICS Vol. 59 No. 9 2014

410 LEONOV et al.

q1 inverse problem of the dynamics is unstable. For

0.2 q1 implementing the program motion, it is necessary to
q1p introduce feedbacks. The control on displacements
and velocities of rods, which results in a steady
motion, was considered. For obtaining a sufficiently
exact program motion of the platform, there are two
possibilities, either to set precisely the forces in rods or
to increase strongly the feedback factor G in the dis
placements of rods. When reducing the requirements
for the accuracy, it is possible to obtain comprehensi
ble results in the case of setting the approximate
expressions for the program forces in the rods or even
equating them to zero. Indeed, it is necessary to
0 1.5
increase the factor G in this case.
t
The presented material is useful when analyzing a
Fig. 3. Program and real motions. more complex system including the dynamics of
pneumatic cylinders.

in velocity. Therefore, we take homogeneous condi

tions for the real motion: ACKNOWLEDGMENTS
This work was supported by the St. Petersburg State
q k = q· k = 0, k = 1, …, 6, for t = 0. (35) University, a group of Transas, and by the RF Presi
dent grant for supporting the leading scientific schools
The obtained solution consists of the transient process NSH3384.2014.1
for 0 ≤ t ≤ t∗, which tends to the periodic solution with
the frequency ν with time. Therefore, when calculat REFERENCES
ing the norm of this solution from Eq. (25), it is neces
1. D. Stewart, Proc. Inst. Mech. Eng. 180 (15), 371 (1965).
sary to consider that t > t∗. It is natural that the
2. P. Nanua, K. J. Waldron, and V. Murthy, IEEE Trans.
obtained results coincide with those listed in Table 1. Robotics and Automation 6 (4), 438 (1990).
The point concerns only the duration of the transient 3. K. Harib and K. Srinivasan, Robotica 21 (05), 541 (2003).
process, which proved to be less than half of the period 4. Z. Geng, L. S. Haynes, J. D. Lee, and R. L. Carroll,
according to the calculations carried out (see also Robotics and Autonom. Syst. 9, 237 (1992).
Fig. 3). 5. F. A. Adkins and E. J. Haug, Trans. ASME J. Mech.
It is possible to simplify the set of program forces Des. 31 (368), 330 (1997).
p 6. Z. Ji, Proc. IEEE Conf. Robotics and Automation 1,
assuming that F k = 0, k = 1, …, 6. In Table 2, we listed 212 (1993).
the results obtained upon integration. It can be seen 7. B. Dasgupta and T. S. Mruthyunjaya, Mech. and
that the results changed insignificantly in comparison Machine Theory 33 (8), 1135 (1998).
with those in Table 1 at least for G = 100 and G = 30. 8. G. Lebret, K. Liu, and F. L. Lewis, J. Robotic Systems
For illustrating the approach of the real and pro 10 (5), 629 (1993).
gram motions with the discrepancy of the initial con 9. S. A. Zegzhda, Sh. Kh. Soltakhanov, and M. P. Yushkov,
Nonholonomic Mechanics. Theory and Applications
ditions and the incorrectness in setting the program (Nauka, Moscow, 2009) [in Russian].
forces in rods, we show the plots of the functions q1(t) 10. N. N. Polyakhov, S. A. Zegzhda, and M. P. Yushkov,
p
and q 1 (t) in Fig. 3. We took the parameter values a = Dokl. Akad. Nauk 309 (4), 805 (1989).
0.2, G = 30, Gf = 2, ν = 2, and considered the case with 11. V. V. Aleksandrov, Dokl. Akad. Nauk 299 (2), 296 (1988).
p 12. V. V. Aleksandrov, B. Ya. Lokshin, E. L. Gomes, and
F k = 0, while the functions qk(t) satisfy zero initial I. Kh. Salazar, Fundament. i Prikl. Matematika 11 (7),
conditions (35). 97 (2005).
13. G. A. Leonov, S. A. Zegzhda, N. V. Kuznetsov,
P. E. Tovstik, T. P. Tovstik, and M. P. Yushkov, Dokl.
CONCLUSIONS Phys. 59 (3), 153 (2014).
The equations of dynamics of the Stewart platform 14. P. A. Zhilin, Vectors and Tensors of the Second Rank in
ThreeDimensional Space (Nestor, St. Petersburg,
under the action of the forces applied to rods are 2001) [in Russian].
derived. The direct and inverse problems of the
dynamics were solved. As is known, the solution of the Translated by V. Bukhanov

DOKLADY PHYSICS Vol. 59 No. 9 2014

View publication stats