Velocity control of a two-wheeled inverted pendulum mobile robot: a fuzzy mod...
Research.Essay_Chien-Chih_Weng_v3_by Prof. Karkoub
1. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 1 / 10
Statement of Research Interest
Chien-Chih Weng, PhD
My general research interest is focused on fuzzy, neural network, and integral sliding
mode control for nonlinear MIMO systems with system uncertainties and external
disturbances. In particular, I am interested in the dynamic modeling and numerical analysis of
robotic systems. My future interests are directed towards using theoretical and experimental
validation of complex dynamic systems such as drilling systems and robotic systems.
Highlights of Past Research
My previous research work reflects my theoretical background in the field of nonlinear
MIMO systems. My M.Sc. thesis research focused on the control of nonlinear MIMO systems with
matched time-varying uncertainties using an adaptive fuzzy terminal sliding mode controller
design. During the course of my doctoral work, I focused on physical nonlinear systems, such
as parallel manipulators and rolling robots with uncertainties and external disturbances.
Highlights of Relevant Research
An adaptive fuzzy/neural network was developed to control a nonlinear MIMO systems
with uncertainties and external disturbances. My current research focuses on:
1. Solving nonlinear MIMO systems with uncertainties and external disturbances.
2. Developing algorithms for nonlinear MIMO systems with uncertainties and external
disturbances.
3. Verify the model through numerical analysis and apply the developed algorithms to practical
nonlinear MIMO/SIMO systems such as a train of self-balancing vehicles and drilling systems.
Research Experience
Since I started my Ph.D work in 2007, I have been involved with several research
projects at Tatung University in Taiwan and Texas A&M University at Qatar. As a result of my
research, two journal articles have been published and three journal articles submitted to top
journals in systems and control engineering. Following is a brief description of each project that
I actively participated in.
2. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 2 / 10
Project 1: An observer-based adaptive neural network tracking control of robotic
systems, 2013
Systems such limbs/links and joints undergoing rotation and sliding motion with and without fixed base
are very complex dynamical systems, which are highly nonlinear. These problems studied by many
researchers, and several models were introduced. Only a few models have considered limb-to-limb
coupling uncertainties (e.g., serial or parallel manipulators) and multiple time-delayed state
uncertainties (e.g., snake-like or rolling-like robots). These highly nonlinear phenomena, in which
coupling is likely to occur, are due to the geometric and construction properties of robotic systems.
Understanding the physics of an inherently nonlinear problem is also important in applications such as
motion planning of robots. In this project, we developed a new approach to solve the system
uncertainties, multiple time-delayed state uncertainties, and external disturbances (see Fig. 1). The
methodology is observer-based adaptive wavelet neural network (OBAWNN) tracking control method.
The approach uses the recurrent adaptive wavelet neural network model to approximate the dynamics
of the robotic system. The method also uses an observer-based adaptive control scheme to stabilize
system. The advantage of employing adaptive wavelet neural dynamics is that we can utilize the neuron
information by activation functions to on-line tune the hidden to-output weights, and the adaptation
parameters to estimate the robot parameters and the bounds of the gains of delay states directly using
linear analytical results. Based on the simulation results, it was clear that the proposed controller was
effective in controlling the highly nonlinear system (see Figs. 2-5). The work resulted in one publication
[1].
Figure 1 The rolling robot with three-links model Figure 2 The trajectories of the error states of the
distances for the first joint
Figure 3 The trajectories of the error states of the
distances for the second joint
Figure 4 The trajectories of the error states of the
distances for the third joint
3. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 3 / 10
Figure 5 The trajectories of the estimates of the
joints
Project 2: 𝑯∞ tracking adaptive fuzzy integral sliding mode control for parallel
manipulators, 2014
In the first part of this project, we derived the inverse kinematics, and mathematical model of a 2-dof
parallel robot (see Fig. 6). Two controllers were used, 𝐻∞ tracking adaptive integral sliding mode
control (HTAFISC) and adaptive integral sliding mode control (AFISC) methods. We compared the
trajectories of the joint states from the two methods and the results showed that the HTAFISC method
is better than the AFISC (see Figs. 7-10). In the second part of this project, we compared the responses
of the tracking errors for joints 1, 2, 3, and 4 with time-varying exponent p(t) and constant exponent p.
From Figs. 11-14, it is seen that time-varying exponent p(t) can have better performance than that
constant exponent p. The work resulted in one publications [2-5].
Figure 6 The parallel manipulator model Figure 7 The trajectories of the joint states (Angles)
using AFISC
Figure 8 The trajectories of the joint states (Angles)
using HTAFISC
Figure 9 The trajectories of the joint states (velocities)
using AFISC
4. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 4 / 10
Figure 10 The trajectories of the joint states
(velocities) using HTAFISC
Figure 11 Phase-plane trajectories of the tracking
errors for joint 1 using constant exponent and time-
varying exponent from the same initial joint
Figure 12 Phase-plane trajectories of the tracking
errors for joint 2 using constant exponent and time-
varying exponent from the same initial joint
Figure 13 Phase-plane trajectories of the tracking
errors for joint 3 using constant exponent and time-
varying exponent from the same initial joint
Figure 14 Phase-plane trajectories of the tracking
errors for joint 4 using constant exponent and time-
varying exponent from the same initial joint
5. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 5 / 10
Project 3: Adaptive Integral Sliding Mode Recurrent Neural Tracking Control for
Nonlinear Interconnected Systems, 2014
In this project, an 𝐻∞ tracking control problem for nonlinear interconnected systems is proposed via
adaptive, integral sliding mode, recurrent neural control (AISMRNC) scheme. The dynamics of the
nonlinear interconnected system is first formulated as an error dynamics according to a specified
reference model. Then, a recurrent neural technique with on-line estimation schemes is developed to
approximate the dynamics of the nonlinear interconnected systems. The on-line estimation schemes in
the neurons are developed to overcome the uncertainties and identify the gains of the unknown
interconnected ones, simultaneously. The advantage of employing a recurrent neural interconnected
system is the use of linear analytical results instead of estimating nonlinear uncertain functions in
dynamics with interconnection. Using the concept of parallel distributed compensation (PDC), the
adaptive integral sliding mode control scheme is synthesized from the recurrent neural interconnected
system to tackle the system with the interconnected uncertainties, and external disturbances. Finally, in
the simulations, we present some comparison of responses of the trajectories of the joint states for
parallel manipulators using AISMRNCS and HTAFISMCS [2] (see Figs. 15-18). The work resulted in one
publications [6-7].
Figure 15 Trajectories of the joint states and desired
joint states using AISMRNCS (blue line) and
HTAFISMCS (red line)
Figure 16 Trajectories of the joint states and
desired joint states using AISMRNCS (blue line)
and HTAFISMCS (red line)
Figure 17 Trajectories of the joint states and desired
joint states using AISMRNCS (blue line) and
HTAFISMCS (red line)
Figure 18 Trajectories of the joint states and
desired joint states using AISMRNCS (blue line)
and HTAFISMCS (red line)
6. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 6 / 10
Project 4: 𝑯∞ Tracking Adaptive Fuzzy Integral Sliding Mode Control for a Train
of Self-Balancing Vehicles, 2014
Underactuated mechanical systems arise in many practical applications. These systems are
inherently highly nonlinear which makes the tracking task very challenging. For example, commercial
underactuated products, such as the SEGWAY and the Honda U3-X unicycle, have been well recognized
as powerful personal transportation vehicles. In addition, at the National Institute for research in
automation and informatics (INRIA) in France, researchers have been working on a different version of
the two-wheel vehicle which they named the B2. Unlike the SEGWAY which is a one person vehicle
system, the B2 vehicle can carry two passengers. Moreover, the B2 passengers can be seated to
maneuver the vehicle through a joystick whereas the SEGWAY passengers must stand up. In a way, the
B2 vehicle is more like a two-wheel car rather than a scooter, i.e. it represents a potential rival to the
available two and three-wheels vehicles. In order to expand the practicality and applicability of the B2
vehicle, in this project, we design a self-balancing-train system composed of n-vehicles connected by
linear springs (see Fig. 19). In this project, an 𝐻∞ tracking adaptive fuzzy integral sliding mode control
scheme is proposed for n-self-balancing interconnected vehicles system where uncertainties and
disturbances are included. Using the concept of parallel distributed compensation (PDC), the adaptive
fuzzy scheme combined with the integral sliding mode control scheme is synthesized to tackle the
system uncertainties and the external disturbances such that 𝐻∞ tracking performance is achieved.
Simulation results for a 2-self-balancing interconnected vehicles system are presented in Figs. 20-21.
The work resulted in one publications [8].
Figure 19 Schematic diagram of the n-vehicle self-balancing-train model
Figure 20 The responses of the trajectories of C.M.
of the first vehicle using time-varying exponent
with the road inclination 9.96 deg. from the initial
position to desired position
Figure 21 The responses of the trajectories of C.M. of
the second vehicle using time-varying exponent with
the road inclination 9.96 deg. from the initial position to
desired position
7. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 7 / 10
Project 5: Trajectory Tracking Control for Rotary Drilling Systems via Integral
Sliding Mode Control Approach, 2014
In this project, the trajectory tracking of a rotary drilling system via integral sliding mode control
(ISMC) scheme was developed. The rotary drilling system is composed of the rotary table, gear box,
motor, drill pipe, drill collar, and drillbit (see Fig. 22). The control scheme with time-varying exponent
was proposed to deal with the stick-slip and bit-bounce of the rotary drilling system so that the tracking
objective of the rotary table velocity of drilling components is achieved. The simulation results showed
that the ISMC law is capable of accurately controlling the bit speed. It also leads to faster dynamic
responses and suppresses torsional stick-slip induced vibrations, axial bit-bounce, and external
disturbances affecting the rotary drilling system (see Figs. 23-26) . The work resulted in one
publications [9].
Figure 22 Mechanism of rotary drilling model Figure 23 Time responses of the bit (blue line) and
rotary table (red line) speeds
Figure 24 Time response of the WOB Figure 25 Time response of the bit-bounce
Figure 26 Time responses of the TOB (blue line)
and top torque u (red line)
8. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 8 / 10
Project 6: Trajectory Tracking Control for the Self-Excited Vibrations in Deep
Drilling Systems via Adaptive Fuzzy Integral Sliding Mode Control
Approach, 2015
Stick-slip and bit-bounce are destructive dynamic phenomena encountered during rotary
drilling of oil-wells, but their exact origins and interplay are far from obvious. Consequently, the system
running output must be controlled accurately to avoid serious damage to the drilling equipment. In this
paper, a novel adaptive fuzzy integral sliding mode (AFISM) controller is designed to suppress the self-
excited axial and torsional vibrations of deep drilling systems subject to lumped uncertainty, including
system parameter variation and external disturbance. The proposed approach effectively combines the
design techniques of adaptive fuzzy control and integral sliding mode (ISM) control in order to
overcome the system nonlinearity and to provide the boundedness of the tracking errors in the closed-
loop system. Moreover, the proposed adaptive fuzzy control strategy does not depend on accurate
mathematical models, in other words, without having a priori knowledge of the upper bound of the
lumped uncertainty. Therefore, fuzzy control technique can be used to approximate the lumped
uncertainty as well as the upper bound of the uncertainty on line caused by the fuzzy modeling error. In
addition, the salient feature of the proposed control scheme with time-varying exponent is that the
controlled system has a integral sliding motion without a reaching phase where the conventional
sliding-mode control (SMC) has. The proposed technique offers superior properties such as faster finite
time convergence than with constant exponent. Therefore, an innovative intelligent control method
combining adaptive fuzzy with integral sliding mode control is developed for suppressing the self-
excited axial and torsional vibrations of deep drilling systems. The control law is derived with the strict
theoretical proof of the Lyapunov stability. The simulation results (see Figs. 27-34) show that the
AFISMC law is capable of: (1) accurately controlling the bit speed (2) suppressing the torsion stick-slip
induced vibration and axial bit-bounce with fast response time and (3) minimizing the affect of the
lumped uncertainty and modeling error on the rotary drilling system response. The work resulted in
one publications [10].
Figure 27 Simplified lumped model of the drilling
system
Figure 28 Time responses of the bit (red line) and
rotary table (blue line) speeds.
9. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 9 / 10
Figure 29 Time response of the WOB Figure 30 Time response of the TOB
Figure 31 Time response of the axial position of the
BHA
Figure 32 Time response of total depth of cutting
process d
Figure 33 Time response of the minimum
approximation error
Figure 34 Time response of the top torque
10. April 2015
Statement of Research Statement Chien-Chih Weng, Ph.D Page 10 / 10
Publications:
[1] Wen-Shyong Yu and Chien-Chih Weng, (2013) An observer-based adaptive neural network
tracking control of robotic systems, Applied Soft Computing, 13 (12), pp. 4645-4658 (SCI)
[2] Chien-Chih Weng and Wen-Shyong Yu, (2008) Adaptive Fuzzy Sliding Mode Control for Linear
Time-Varying Uncertain Systems. Proceedings of IEEE International Conference on fuzzy systems
[3] Chien-Chih Weng and Wen-Shyong Yu, (2010) Tracking Adaptive Fuzzy Integral Sliding Mode
Control for Parallel Manipulators. Proceedings of IEEE International Conference on Systems Man
and Cybernetics
[4] Chien-Chih Weng and Wen-Shyong Yu, (2012) 𝐻∞ Tarcking Adaptive Fuzzy Integral Sliding
Mode Control for Parallel Manipulators. Proceedings of IEEE International Conference on Fuzzy
Systems
[5] Wen-Shyong Yu and Chien-Chih Weng, (2014) 𝐻∞ Tracking Adaptive Fuzzy Integral Sliding Mode
Control for Parallel Manipulators, Fuzzy Sets and Systems, 248, pp. 1-38. (SCI)
[6] Wen-Shyong Yu and Chien-Chih Weng, (2014) Neural-Based Adaptive Integral Sliding Mode
Tracking Control for Nonlinear Interconnected Systems. Proceedings of IEEE International Joint
Conference on Neural Networks (IJCNN)
Under Review:
[7] Chien-Chih Weng, Mansour Karkoub, Hsuan-Yi Chen, Wen-Shyong, and Ming-Guo Her, Adaptive
Integral Sliding Mode Recurrent Neural Tracking Control for Nonlinear Interconnected Systems,
Submitted to IEEE Transactions on Neural Network and Learning Systems
[8] Chien-Chih Weng, Mansour Karkoub, Tzu-Sung Wu, Wen-Shyong, and Ming-Guo Her, 𝐻∞ Tracking
Adaptive Fuzzy Integral Sliding Mode Control for a Train of Self-Balancing Vehicles, Submitted to
ASME Journal of Dynamic Systems, Measurement and Control
[9] Chien-Chih Weng, Mansour Karkoub, Hsuan-Yi Chen, Wen-Shyong, and Ming-Guo Her, Trajectory
Tracking Control for Rotary Drilling Systems via Integral Sliding Mode Control Approach,
Submitted to IEEE Transactions on Automation Science and Engineering
[10] Chien-Chih Weng, Mansour Karkoub, Hsuan-Yi Chen, Wen-Shyong, and Ming-Guo Her,
Trajectory Tracking Control for Rotary Drilling Systems via Integral Sliding Mode Control
Approach, Submitted to IEEE Transactions on Fuzzy Systems