Professional Documents
Culture Documents
Abstract. Compliant actuators are widely used in industrial robots due to the fact that the compliant
elements have the capacity to absorb excessive collision force and guarantee the robot safety.
Moreover, the compliant elements can be also employed as force/torque sensors in the control loops.
Therefore, the research on the compliant actuators is of importance in the theory and application
sides. Specifically, it is required that the robot systems should have a low stiffness when the
subjected collision force is greater than the prescribed injury tolerance. But, in the normal status, the
systems maintain the required capability. In this paper, the system stiffness of a compliant actuator
is defined as a ‘mechanical impedance’ (the ratio between output force and output shaft angle). A
cascade controller is well designed by following the results on analyzing the characteristic of
compliant actuator impedance stability and collision. At the end, simulation results show the
effectiveness of the analytic results and designed controller.
Introduction
For traditional robot systems especially those used in industrial robot arms, high
impedance-position controlled actuators are widely used in order to get a high force/torque
bandwidth and trajectory tracking control fidelity. However, at the safety side, it is required that the
actuator systems should be with compliant elements. That is, there is a conflict between the
requirements on the bandwidth and on the safety. In order to accommodate these two requirements,
Pratt, Robinson and Williamson[1~6] proposed a concept called series elastic actuator (SEA). The
main idea is to use a compliant element in the flexible actuator system such that the collision energy
can be stored and the shocks can be absorbed. In other words, a system with a compliant element
leads to a low impedance which can force the controlled actuation to improve the system safety. In
the literature, we can see that this kind of actuators were widely used in walking robots[7],
prosthetic and orthotic leg systems[8], and force-sensing robot arms[9].
Compared with a stiff robot system, the manipulator stability of a compliant system cannot be
guaranteed when it contacts an unknown environment. Thus, the stability analysis of the impedance
is necessary. According to the Colgate and Hogan’s principle, Williamson[2] and Heike Vallery[10]
designed controllers to ensure the passivity and the stability of the actuators. But further research on
the system stability, mechanical impedance, collision power is necessary for a better performance.
For walking robots, such as prosthetic and orthotic leg systems and force-sensing robot arms,
these tasks run at a low speed but a high force density such that a transmission or a gear reduction is
required. Transmissions introduce the friction and backlash which are undesirable effects. The
backlash between every part of the actuation creates noise and instability. In order to conquer such
drawbacks, cascade scheme based on torque control in main loop and velocity control in the inner
loop is adopted. Gordon Wyeth[11] developed a new variant of SEA that minimizes the effect of
the system backlash, and overcomes the stiction effect. Heike Vallery[10] employed a cascade
scheme which can ensure passivity and while counteracting static errors. In his work, he
emphasized on the analysis of the parameters boundaries. Reza Ghorbani’s[12] cascade scheme
used saturation operators in the system to ensure minimize the saturation effects. The characteristics
of the system based on cascade control scheme need a further research, which motivate the work in
our paper.
All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications, www.ttp.net. (ID: 142.103.160.110, University of British Columbia, Kelowna, Canada-14/07/15,17:36:23)
24 History of Mechanical Technology and Mechanical Design 2012
System Model
In this section, the cascade control strategy with inner velocity loop will be described. The
compliant actuator consists of a DC motor coupled to the load via series elasticity. As shown in
Fig.1, the cascade control strategy has a torque control main loop and a velocity inner loop. The
velocity feedback from an encoder forms a velocity loop for controlling the motor and
compensating for the gearbox nonlinear and error. The PI compensator in the velocity loop is used
to remove the steady state error.
Motor Model.
The Eqs of motor dynamics are
1 d 2θ g dθ g
Tm − Tl = nJ m 2
+ nBm
n dt dt (1)
1
Tm = K i I a − Tl
n (2)
dθ m
Ria = V − K b (3)
dt
where θ m , θ g and θl denote the motor angle, gearbox angle and load angle, respectively. J m is the
motor rotor moment of inertia. Bm is the viscous damping ratio. Tm is the motor torque. Tl is the load
torque. K s is the compliant element spring constant. n is the gearbox ratio. K i is the torque
constant. I a is armature current. K b is the back EMF constant.
In the Laplace domain, these Eqs can be rewritten as
RI a ( s ) = V ( s ) − K bωm ( s ) (4)
( J m s + Bm )ωm ( s ) = K i I a ( s ) − Tl ( s ) / n (5)
θ s = θ g − θl
(6)
The position of the load is determined by the output torque applied to the load Tl . The torque
applied to the load Tl is due to the compliant element deflection θ s .
1 1
θl = Tl 2
= K sθ s (7)
Jl s Jl s2
Tl = K s (θ g − θ l )
(8)
The open loop transfer function from gearbox position to SEA output torque by combining the
Eqs (6)~(8) can be rewritten as:
Tl Ks s2 (9)
=
θg K
s2 + s
Jl
SEA Controller Model.
The cascade control strategy including two PI compensators. The two integral compensators are
used in order to remove steady state error. Compared with Fig.1, Fig.2 shows the cascade control
details with the desired velocity and desired voltage
Applied Mechanics and Materials Vol. 163 25
Kii K
GPIi = K PIi + , GPIo = K PIo + io (12)
s s
TDesired ωDesired ωm
ωl
ωg
Tl
Fig. 1 The cascade control strategy sketch map with a torque main control loop and a velocity
inner loop
1/ n
Td ωd Tm − 1 Tl
GPIo GPIi Ki / R 1/ n 1/ s Ks
− J m s + Bm −
− − θl
1
Kb Jl s2
Stability Analysis
The most important problem in linear control systems concerns stability. Compared with a stiff
actuator, besides the stability analysis of the cascade control system, the output impedance stability
of SEA should also be identified. We use two PI compensators but not two PID compensators in
order to simplify the cascade control model for getting an analytical stability boundary.
System stability.
The linear closed-loop system has closed-loop transfer function in the form:
The closed loop transfer function provides the characteristic polynomial equation. Calculating
the polynomial by Routh’s stability criterion, the stability of the system can be checked.
26 History of Mechanical Technology and Mechanical Design 2012
Impedance stability.
For an SEA, the impedance is defined as the ratio between output torque Tl and output shaft
angle θl . Output impedance is measure of the system stiffness for different load motion requirement.
For robot motion, low impedance is recommended. Not only the impedance is a measure whether a
robot system will be stable while in contact with an environment, but also the impedance is a
symbol of collision power. The impedance closed-loop transfer function as below
K i K b K i K pi K KK K i K s K pi K po
where a20 = − K S J m , a21 = − K S Bm , b20 = J m , b21 = Bm + + , b22 = 2s + i ii +
R R n R nR
Ki K s KK K K
, b23 = ( K po K ii + K pi K io ) , b24 = i s ii io
nR nR
Colgate[13] and Hogan[14] proved that a system will be stable when contact with unknown
environment if the impedance Z ( s ) obeys two rules:
1, Z ( s ) has no poles in the right half plane ( Z ( s ) is stable);
2, The imaginary part of Z ( jω ) is negative for all frequencies ω .
The first rule can be identified by the Routh’s stability criterion. According to the second rule,
RJ m K pi
pi J m po J m
K K
K ii RBm K io K b K io
> ∧ > ∧ K po > max( K ( RB + K K + K K ) , + ) (16)
K
ii B m K
io B m RJ K K K
pi m i b i pi
− m Ti1 i ii ii
K ii K io K io
Fig.3. shows the response of the actuator to a desired torque signals in period of 1 second. The
desired sinusoid torque signal is with amplitude 2Nm and frequency 50rad/s. The desired velocity
signal and response are also shown in Fig.4. As shown in the Fig.3 and Fig.4, the actuator can
successfully follow the desired torque signal and desired velocity signal.
Applied Mechanics and Materials Vol. 163 27
2 500
1.5
1
Torque(Nm)
Speed(rad/s)
0.5
0 0
-0.5
-1
-1.5
-2 -500
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time(s) Time(s)
Fig. 3. Commanded and actual torque signal Fig. 4. Commanded and actual velocity signal
0
increasing Kpo
-20
Magnitude (dB)
-40
-60
-80
-100
0
-45
Phase (deg)
-90
-135
-180 -2 0 2 4 6
10 10 10 10 10
Frequency (Hz)
Fig. 5 The change of system bandwidth with increasing the outer-loop gain
A compliant element in the joint will decrease the bandwidth of the system, but the bandwidth is
not absolutely dependent on the spring constant. As can be seen in Fig5, in order to maintain a high
closed-loop bandwidth, increasing the control gain could compensate the decreasing of the system
bandwidth.
Conclusion
In this paper, a cascade controller for a compliant actuator has been designed. There are two
sub-controllers in the cascade control strategy who provide four parameters to regulate the
characteristic. By analyzing the stability of the control system model and output impedance, simple
boundaries for the control system parameters are derived.
References
[1] D.W. ROBINSON: Design and Analysis of Series Elasticity in Closed-loop Actuator Force
Control ( PhD Thesis, Department of Mechanical Engineering, Massachusetts Institute of
Technology, 2000).
[2] M. M. WILLIAMSON: Series Elastic Actuators (MS Thesis, Department of Electrical
Engineering and Computer Science, Massachusetts Institute of Technology, 1995).
[3] G. A. PRATT and M. M. WILLIAMSON: Series Elastic Actuators, IEEE International
Conference on Intelligent Robots and Systems, 1995, vol. 1, pp. 399-406.
[4] G. A. PRATT: Stiffness isn’t everything. In Fourth International Symposium on Experimental
Robotics, ISER 95, (Stanford, California, June 30-July 2.1995).
[5] D. W. ROBINSON, J. E. PRATT, D. J. PALUSKA and G. A. PRATT: Series elastic actuator
development for a biomimetic walking robot. In 1999 IEEE/ASME International Conference on
Advanced Intelligent Mechatronics, (Atlanta, USA. September 19-23, 1999). pp:561-568.
[6] G. A. PRATT, P. WILLIAMSON, C. BOLTON and A. HOFMAN: Late Motor Processing in
Low-Impedance Robots: Impedance Control of Series Elastic Actuators, Proc of American
Control Conference (Boston, Massachusetts, June 30-July 2, 2004). pp: 3245-3251.
28 History of Mechanical Technology and Mechanical Design 2012