Content uploaded by Arash Arami
Author content
All content in this area was uploaded by Arash Arami on May 14, 2015
Content may be subject to copyright.
International Journal of Innovative
Computing, Information and Control ICIC International c
°2005 ISSN 1349-4198
Volume x, Number 0x, x 2007 pp. 0–0
REAL TIME EMOTIONAL CONTROL FOR ANTI-SWING AND
POSITIONING CONTROL OF SIMO OVERHEAD TRAVELING
CRANE
M. R. Jamali, A. Arami, B. Hosseini, B. Moshiri, and C. Lucas
Control and Intelligent Processing Centre of Excellence
Department of Electrical and Computer Engineering
University of Tehran
School of Electronic and Computer Engineering, Department of Engineering
Shahid Rajaee University
m.jamali@ece.ut.ac.ir; a.arami@ece.ut.ac.ir ; b.hosseini@ece.ut.ac.ir,b.moshiri@ut.ac.ir,c.lucas@ipm.ir
Abstract. Research in artificial intelligence and bioinspired algorithms is still being ac-
tively pursued in different fields of engineering. In this work, Brain Emotional Learning
Based Intelligent Controller (BELBIC) is applied for real time positioning of laboratorial
overhead travelling crane. This controller is based on biologically motivated algorithm
originating from emotional processes in the limbic system of the mammalian brain. Sim-
ulations show that learning capability, adaptation, robustness and other control concerns
of this controller are comparable with conventional techniques which lead to better per-
formance in many cases. Two objectives, tracking desired position and keeping pendulum
vertically, must be considered simultaneously. The bottom up strategy was utilized for
designing the controllers. First separated BELBIC was designed for each control task.
Next, in order to suppose the real coupling between control tasks, the objective of each
control tasks was considered in the stress signal of the other one. Obtained results in
real tracking applications are also comparable with other conventional and intelligent ap-
proaches such as hierarchical fuzzy control (HFLC) and confirm the simulation results.
Learning capability, model free control algorithm, robustness and fast response are main
characteristics of this controller and designer can define emotional stress signal based on
control application objectives.
Keywords: BELBIC, Model Free Control, HFLC, PID Controller
1. Introduction. Brain Emotional Learning Based Intelligent Controller (BELBIC) is
an example of bioinspired control methods which is based on limbic system of mammalian
brains. This controller is based on emotional behaviours in biological systems. Emotion
is an emergent behaviour in biological systems for fast decision making in complex envi-
ronments. The advantages of this behaviour cannot be neglected in creature surveillance.
Several attempts have been made to model the emotional behavior of human brain [1],
[2], and [3]. In [2] the computational models of Amygdala and context processing were in-
troduced. Based on the cognitively motivated open loop model, BELBIC was introduced
in [4] and after that this controller was utilized in several applications. Applying BELBIC
in Speed control of an interior permanent magnet synchronous motor was shown in [6]. In
[5] a modified version of BELBIC was applied to heating, ventilating and air conditioning
(HVAC) system which is multivariable, nonlinear and non minimum phase. In [7] this
controller used for controlling identified washing machine and in [8] this controller was
1
2 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
tuned for washing machine with multi objectives constraints based on evolutionary ap-
proaches. A trade-off between energy consummation and other control objectives should
be considered by designer. Because of recurring applications of BELBIC in new contexts,
describing the BELBIC in pattern format was done in [9]. Pattern describes a problem,
which occurs over and over again in our environment, and then describes the core of the
solution to that problem [10]. Reusability, extendibility and implementation concerns in
different platforms were described in this pattern and designer can reuse the BELBIC
framework easily in desired control applications.
All the above applications of BELBIC were done in software simulation environments.
For the first time, the real-time implementation of the BELBIC for interior permanent
magnet synchronous motor (IPMSM) drives was presented in [11]. The controller was
successfully implemented in real-time using a digital signal processor board ds1102 for
a laboratory 1-hp IPMSM. Results show superior control characteristics, especially very
fast response, simple implementation and robustness with respect to disturbances and
manufacturing imperfections. These features make the BELBIC useful in different tasks
which fast tuning should be a necessity. In this paper this controller is implemented
for real time and model free control of overhead travelling crane which is a laboratorial
set and is a small model of bridge crane. The plant is nonlinear, one input-two output
and multi objectives. The benefit of model free control is eliminating the identification
task. However many methods are presented to identify and model the complex nonlinear
systems and with respect to successful applications to model unknown nonlinear system
by a set of fuzzy rules [12], these procedures are too costly and time consuming. Also when
the identifying procedure leads to very complex system, it is hard to design controllers
even define proper rule for fuzzy controllers [13]. Two BELBICs based on control strategy
cooperate together for controlling the mentioned system. In emotional stress signal of each
controller, the objective of the other one is employed and the emotional signal is the fusion
of control objectives. In other words, each of these controllers controls an objective of
whole system which firstly assumed to be decoupled and next the emotional stress signals
play the role of coupling these two controllers. Unlike classic controllers, this controller
has the learning capability and initial tuning of parameters is not very critical.
This paper consists of five sections: In section 2, mathematical description of BELBIC
is demonstrated. Practical overhead crane is presented in section 3. Experimental result
and comparison of this controller with other classical and intelligent methods is shown in
section 4 and finally section 5 contains concluding remarks.
2. Brain Emotional Learning Based Intelligent Controller. BELBIC is a simple
computational model of Amygdala and Orbitofrontal cortex of brain. In this section, a
short description about each part of BELBIC is given. The structure of BELBIC is shown
in Figure 1 [2].
Thalamus is a simple model of real thalamus in mammalians’ brain. Some simple pre-
processing on sensory input signals such as noise reduction or filtering can be done in
this part. Sensory Cortex receives its inputs from thalamus, and it is assumed that this
part is responsible for subdividing and discrimination of the coarse input from thalamus
[2]. Orbitofrontal Cortex is supposed to inhibit inappropriate responses from Amygdala
based on the context given by the hippocampus [2]. Amygdala is a small structure in
the medial temporal lobe of brain that is thought to be responsible for the emotional
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 3
Figure 1. Structure of BELBIC [Moren00]
evaluation of stimuli [2]. This evaluation is in turn used as a basis for emotional states,
emotional reactions and is used for attention signal and laying down long-term memories
[2].
BELBIC receives Sensory Input signals via Thalamus. After pre-processing in Sensory
Input, processed input signal will be sent to Amygdala and Sensory Cortex. Amygdala
and Orbitofrontal compute their outputs based on Emotional Signal received from envi-
ronment. Final output is calculated by subtracting Amygdala’s output from Orbitofrontal
Cortex’s output. Through this section, functionality of these parts and learning algorithm
is discussed based on [2].
The Thalamic connection is calculated as the maximum overall Sensory Input S and
becomes another input to Amygdala as described in equation (1). Unlike other inputs to
Amygdala, the thalamic input is not projected into the Orbitofrontal part and can not
be inhibited by itself.
Ath = max
i(Si) (1)
For each Anode in Amygdala, there is a plastic connection weight V. Any input is
multiplied by this weight to provide the output of the node. The Onodes behave anal-
ogously, with a connection weight Wapplied to the input signal to create an output.
The connection weights Viare adjusted proportionally to the difference between the emo-
tional stress and the activation of the Anodes. The αterm is a constant that is used
to adjust the learning speed. In formula (2) Amygdala learning rule is presented. This
4 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
is an instance of a simple associative learning system. The real difference between this
system and similar associative learning systems is the fact that this weight-adjusting rule
is monotonic, i.e., the weights Vcannot decrease. At first glance, this may seems like a
fairly substantial drawback; however, there are good reasons for this design choice. Once
an emotional reaction has been learned, this should be permanent. It is the task of the
Orbitofrontal part to inhibit this reaction when it is inappropriate [2].
∆Vi=α(Simax(0, stress −XAj)) (2)
The reinforcement signal for the Onodes is calculated as the difference between the
previous output Eand the reinforcing signal stress . In other words, the Onodes compare
the expected and received reinforcement signals; therefore, inhibiting the output of the
model should be a mismatch. In (3), the learning rule in the Orbitofrontal Cortex is
presented.
∆Wi=β(SiX(Oj−stress)) (3)
The Orbitofrontal learning rule is very similar to the Amygdala rule. The only difference
is that the Orbitofrontal connection weight can either increase or decrease as needed to
track the required inhibition. Parameter βis another learning rate constant. The A
nodes produce their outputs proportionally to their contribution in predicting the reward
or stress signal stress, while the Onodes inhibit the output of Ewhen necessary.
Ai=SiVi
Oi=SiWi
E=PAi−POi
(4)
Equation (4) presents the Model output expression.
3. Practical Overhead Crane. As it is shown in Figure 2, the overhead traveling crane
consists of a trolley, a rope, and a load. The load is regarded as a material particle with a
mass of m. The rope is considered as an inflexible rod with a length of l, which its mass is
negligible in comparison with the load mass. Having a mass of M, the trolley moves on a
straight rail. It is supposed that no friction exists in the system. Therefore the dynamic
of the overhead traveling crane can be described by equations (5) and (6) [15].
With more detail these assumption can be made for simplifying the model: (a) The
dynamics and nonlinearities of driving motor are neglected. (b) The trolley moves along
the track without friction or slip. (c) The rope has no mass and elasticity. (d) There is
no damping of pendulum. (e) The load is regarded as a point mass.
cos(θ)..
x+l..
θ+gsin(θ) = 0 (5)
(M+m)..
x+ml cos(θ)−ml(.
θ)2sin(θ) = F(6)
As shown above, it is obvious that the state equations are nonlinear. The cart friction
is a non-linear function of its velocity as shown in Figure3 [14].
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 5
Figure 2. A simplified model for overhead crane
Figure 3. Friction model of crane [Feedback02]
The control task is tracking reference signal along with and avoiding of pendulum load
swings. Block diagram of plant with BELBIC controllers is presented in Figure 4.
Figure 4 shows that two BELBICs work together to control the plant. Sensory input
of the first one is position error and sensory input of the other one is angle error. The
6 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
Figure 4. Block Diagram of Control System
emotional signals are created based on designer objectives. In addition, designer can
fuse objectives of one controller by the other one. In the other words, decoupling of
controllers is compensated by fusing the objectives in reward or stress functions. It must
be noticed that any of above equation and knowledge about friction in procedure of
proposed controller design is not used in this paper. The control strategy which is used
in this paper is model free.
4. Experimental Results. In this section BELBIC performance is compared with other
classic and intelligent controllers in controlling a crane laboratorial set.1
The first controller in this experiment is originally supplied sample controller, which
consists of two PID controllers and a nonlinear friction compensator. This controller is a
model based controller and designed by Feedback Instruments Corporation. It has to be
said that double PID controller is designed to act on sinusoid reference signal and don’t
lead to desire response when the position reference signal is in different type for example
step signals.
The second controller is a HFLC that was implemented on crane and it can track
different type of reference signals with minimal dependency to the model. It is a model
free controller which has been designed based on very simple knowledge about behaviour
of system.
Finally the proposed BELBIC controller is implemented that have learning capability
and works in completely model free manner. Reference signal is sinusoidal which could
be assumed similar to real reference signal of industrial overhead crane. The experi-
mental results are gathered without effect of external disturbance and in the presence of
disturbance.
4.1. Results without Effect of Disturbance. This section demonstrates obtained
results without effect of disturbance. Figure 5 shows the reference sinusoidal and tracked
1The Digital Pendulum Control System, crane system, manufactured by Feedback Instruments Lim-
ited, England.
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 7
signal for each controller. The reference is dashed line and the tracked signal is complete
one.
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4 Tracking Response
PID
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
HFLC
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
BELBIC
Time(s)
Figure 5. Tracking of sinusoid reference signal employing different controllers
It is obvious from Figure 5 that the BELBIC learning capability enables it to learn
tracking task and its performance in comparison with double PID controller is admirable.
The performance of proposed controller is dramatically better than HFLC in accurate
tracking reference signal.
As mentioned before, the second control task is fixing the angle about zero degree
during tracking. Figure 6 shows that the double PID controller does the control task
better than others but the performance of BELBIC is still comparable with the HFLC.
In industrial applications, there is direct relation between energy consummation and
control effort. In this experience, control signal of BELBIC is acceptable and better than
other controllers as shown in Figure 7.
4.2. Results in Presence of Disturbance. Like the previous sub-section, the plant
must track the sinusoidal reference signal but in the presence of disturbance between 30th
and 75th seconds. The results are shown in Figures 8, 9 and, 10 for tracking, angle and
control signal respectively.
8 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5 Pendulum Swing
PID
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5
HFLC
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5
BELBIC
Time(s)
Figure 6. Pendulum angle using different controllers
As it is demonstrated in Figure 8, BELBIC controller is more robust than double PID
and HFLC and it can compensate disturbance effect in an acceptable manner for tracking
task.
In the presence of disturbance, the performance of BELBIC in fixing the pendulum is
better than other controllers and it is because of the capability of proposed controller to
reject the disturbances quickly as shown in Figure 9, however according to Figure 6, the
performance of double PID controller was better than others in the absence of disturbance.
The control signal for each controller in presence of disturbance is presented in Figure
10.
In Table 1, Integral Squared Error(ISE) of pendulum angle in absence and presence of
disturbance is demonstrated. Based on results which are depicted on Table 1, in absence
of disturbances the BELBIC did not learn to damp pendulum swing properly. As demon-
strated in Figure 6, at the beginning of learning procedure because of small emotional
stresses, BELBIC learn to damp load swings slowly and some oscillation is observed at this
period. After continuing learning procedure the result of BELBIC improved extremely.
Due to above expression in the absence of disturbance,the ISE of pendulum angle in orig-
inal sample controller and HFLC controllers are better than the BELBIC one. But when
disturbance implies to the system, BELBIC learn faster and leads to better performance.
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 9
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5
1Control Effort
PID
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5
1
HFLC
0 10 20 30 40 50 60 70 80 90 100
−0.5
0
0.5
1
BELBIC
Time(s)
Figure 7. Control effort using different controllers
Table 1. Integral Squared Error of Angle
Applied Controller Without Disturbance In Presence of Disturbance
Original controller (PID) 0.0717 rad 0.1410 rad
HFLC 0.1084 rad 0.1493 rad
BELBIC 0.1221 rad 0.1236 rad
In Table 2 Integral Squared Error (ISE) of trolley position tracking in absence and
presence of disturbance is presented. The indexes values are calculated considering the
results from 20th second till 100th second. This consideration is rational because of
eliminating delayed response arise from initial learning phase.
Table 2. Integral Squared Error of Position
Applied Controller Without Disturbance In Presence of Disturbance
Original controller (PID) 0.0465 0.1020
HFLC 1.0546 1.0232
BELBIC 0.0514 0.0774
10 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4 Tracking Response
PID
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
HFLC
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
BELBIC
Time(s)
Figure 8. Tracking of sinusoid reference signal using PID, HFLC and
BELBIC controllers in presence of disturbance in 30th and 75th seconds.
Based on Table 2, in absence of disturbance state, the result of original sample con-
troller and BELBIC are approximately equal but employing double PID controller leads
to slightly better result in tracking reference signal. BELBIC learn how to compensate the
tracking error while system operation this capability represents itself particularly in pres-
ence of disturbance. In presence of disturbance comparing the results show the superiority
of BELBIC.
In Figure 11, real execution of BELBIC on plant is shown.
5. Coclusion. Previous simulations show that brain emotional learning based controller
(BELBIC) is comparable with classical controllers and even leads to better performance
in some cases. The learning capability, robustness and fast response are the main features
of this controller that has been reported in the previous simulation works. Also the initial
parameters are not very critical and learning capability enables it to obtain reasonable
gains. In this paper, BELBIC was applied to real overhead traveling crane and similar
behaviour to previous simulation in robustness, fast response and learning capability is
seen. Due to control objectives, the emotional signals could be defined properly and
because of learning capability of this controller, the initial parameters are not important.
Obtained results showed that this controller was comparable with double PID controller
and disturbance rejection of this controller is better than classical approaches. Also results
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 11
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4 Pendulum Swing
PID
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
HFLC
0 10 20 30 40 50 60 70 80 90 100
−0.4
−0.2
0
0.2
0.4
BELBIC
Time(s)
Figure 9. Pendulum angle using different controllers in presence of dis-
turbance in 30th and 75th seconds.
of BELBIC were better than HFLC as an intelligent and model free method. This work
is the essential step towards enclosing BELBIC to an appropriate industrial controller.
REFERENCES
[1] Balkenius, C. and J. Moren, A Computational Model of Emotional conditioning in the Brain, work-
shop on Grounding Emotions in Adaptive Systems, Zurich,Germany, 1998.
[2] Moren,J. and C. Balkenius, A Computational Modelof Emotional Learning in the Amygdala: From
animals to animals , 6th International Conference on the Simulation of Adaptive Behavior, Cam-
bridge, MIT Press, pp.383-391, 2000.
[3] Moren, J.,Emotion and Learning: A computational model of the amygdale, PhD thesis, Lund uni-
versity, Lund, Sweden, 2002.
[4] Lucas, C., D. Shahmirzadi, N. Sheikholeslami,Introducing BELBIC: Brain Emotional Learning Based
Intelligent Controller, International Journal of Intelligent Automation and Soft Computing,pp. 11-
22, 2004.
[5] Sheikholeslami,N., D. Shahmirzadi, E. Semsar, C. Lucas, and M. J. Yazdanpanah, Applying Brain
Emotional Learning Algorithm for Multivariable Control of HVAC Systems,International Journal of
Intelligent and Fuzzy Systems, 17 (1), pp 35- 46, 2006.
[6] Milasi, R. M., C. Lucas, and B. N. Araabi,Speed Control of an Interior Permanent Magnet Synchro-
nous Motor Using BELBIC (Brain Emotional Learning Based Intelligent Controller), Special Session
on Emotional Learning and Decision Fusion in Satisficing Control and Information Processing, Min-
isymposium on Satisficing, Multiagent, and Cyberlearning Systems, 5th International Symposium on
12 M. R. JAMALI, A. ARAMI, B. HOSSEINI, B. MOSHIRI, AND C. LUCAS
0 10 20 30 40 50 60 70 80 90 100
−2
−1
0
1Control Effort
PID
0 10 20 30 40 50 60 70 80 90 100
−2
−1
0
1
HFLC
0 10 20 30 40 50 60 70 80 90 100
−2
−1
0
1
BELBIC
Time(s)
Figure 10. Control efforts belong to different controllers in presence of
disturbance in 30th and 75th seconds.
Intelligent Automation and Control, World Automation Congress, WAC 2004, Seville, Spain, June
28- July1, ISIAC 116, 2004.
[7] Milasi, R. M., C. Lucas, and B. N. Araabi, Intelligent Modeling and Control of Washing Machine
Using Locally Linear Neuro- Fuzzy (LLNF) Modeling and Modified Brain Emotional Learning Based
Intelligent Controller (BELBIC), SAsian Journal of Control, 8 (4), 2005.
[8] Milasi, R. M., M. R. Jamali, and C. Lucas,Intelligent Washing Machine: A Bioinspired and Multi-
Objective Approach, Accepted in International Journal of Control, Automation, and Systems, 2007.
[9] Jamali, M. R.,A. Pedram, M. R. Milasy, C. Lucas,Design and Implementation of BELBIC Pattern,
Proceeding of 14th Iranian Conference on Electrical Engineering,ICEE2006,Tehran,Iran, 2006.
[10] Alexander, C., The timeless way of building, Oxford University Press, 1979.
[11] Milasi, R. M., C. Lucas, B. N. Arrabi, T. S. Radwan, and M. A. Rahman, Implementation of
Emotional Controller for Interior Permanent Magnet Synchronous Motor Drive, IEEE / IAS 41st
Annual Meeting: Industry Applications, Tampa, Florida, USA, October 8 - 12, 2006.
[12] Tong, S. and Y. Li ,”Direct Addaptive Fuzzy Backstepping control for a class of nonlinear systems,
International Journal of Innovative Computing, Information and Control,Vol. 3, No. 4, 2007.
[13] Fujimoto, K., H. Sasaki, Y. Masaoka and R. Q. Yang,An implementation of the neuro-fuzzy inference
circuit,International Journal of Innovative Computing, Information and Control ,Vol. 3, No. 4, 2007.
[14] Feedback Instrument Ltd, Digital Pendulum Control Experiments, Manual: 33-935/936-1V60, 2002.
[15] Yager,R. R., An alternative procedure for the calculation of fuzzy logic controller values,IJ. Jpn.
Soc. Fuzzy Technol, SOFT 3,pp. 736-746, 1991.
REAL TIME EMOTIONAL CONTROL FOR OVERHEAD TRAVELING CRANE 13
Figure 11. Digital Pendulum Controller with BELBIC.