538 IEEE TRANSACTIONS ON EDUCATION, VOL. 55, NO.

4, NOVEMBER 2012

Mechatronic Aeropendulum: Demonstration of

Linear and Nonlinear Feedback Control Principles
With MATLAB/Simulink Real-Time Windows Target
Eniko T. Enikov, Member, IEEE, and Giampiero Campa, Member, IEEE

Abstract—This paper presents a low-cost hands-on experiment and equipment time. Hands-on experience, on the other hand, is
for a classical undergraduate controls course for non-electrical invaluable for active and sensory learning styles, which are the
engineering majors. The setup consists of a small dc electrical predominant types of learning styles exhibited by undergrad-
motor attached to one of the ends of a light rod. The motor drives
a 2-in propeller and allows the rod to swing. Angular position uate students [9]–[12]. This paper describes the development
is measured by a potentiometer attached to the pivot point. A and testing of a new low-cost portable laboratory module,
custom-designed circuit board produces the controlled voltage designed to supplement the experience of students taking their
input to the motor. The target board is powered and communi- first course in controls system design.
cates with the PC through its USB port using a virtual RS-232 While there are many turn-key desktop systems designed
port. A simple MATLAB/Simulink module has been created to
read the pendulum angle and send a command signal to the motor. to illustrate controls systems courses, portable kits (such as
The module is based on Real-time Windows Target software, Arduino) are primarily designed for mechatronics and em-
which allows a sampling rate of up to 200 Hz. Students are able bedded computing courses [13], [14]. As such, they require
to design and test classical PID and phase lead-lag controllers, as programming environments, installation of additional software,
well as modern controllers, including state-space controller design and additional plug-in modules for operating dc motors and
combined with feedback linearization. A semester-long series of
assignments is described that can be carried out without the need other actuators. Furthermore, unless advanced circuit boards
for a specialized laboratory or teaching assistants. The project and processors are used, implementing a PID or other dy-
was tested in a classical control systems design class of senior-level namic compensators is cumbersome and requires training in
mechanical engineering students. Student feedback and survey digital control and programming. With the emergence of the
data on the effectiveness of the modules are also presented. MATLAB Simulink graphical programming environment,
Index Terms—Feedback linearization, linear feedback control, modeling and simulation of various plants and controllers can
real-time control, real-time windows target, Simulink. be accomplished quite easily by students who might not have
extensive training in digital control and numerical methods.
However, practical implementation of such controllers remains
I. INTRODUCTION
illusive for most undergraduate students outside of electrical
and mechatronics engineering programs. Therefore, the objec-

H ANDS-ON laboratories have always been an integral tive of this project was to develop a simple physical plant that
part of the engineering curriculum. Their importance has can be used seamlessly with the Simulink Real-Time Windows
been recognized by the Accreditation Board of Engineering Target environment to allow students who are not in Electrical
Technology (ABET) and its predecessors by creating criteria Engineering programs to implement and test real-time con-
requiring adequate laboratory practice for students [1]–[4]. trollers using drag-and-drop-style graphical programming.
During the last three decades, engineering laboratories have In the Department of Aerospace and Mechanical Engineering
become more complex, including simulation tools and com- of The University of Arizona, Tucson, it is not unusual for the
puter-controlled test and measurement equipment [5], [6]. Control System Design course to have enrollments of over 100
This increased sophistication has also led to more expensive students. This makes offering a laboratory section within the
equipment [7], [8]. The inclusion of such laboratory courses in course nearly impossible. The experiment described here was
the undergraduate curriculum is challenging, due to the large developed primarily as a way to provide some practical experi-
numbers of students and the increased demands for instruction ence for students, using an inexpensive and portable setup that
can be taken home. The experiment was developed following
the principles of the variational theory of learning developed by
Manuscript received September 29, 2011; revised January 27, 2012; accepted
March 30, 2012. Date of publication May 01, 2012; date of current version Oc- Marton and coworkers [15], [16] and the approach of guided
tober 26, 2012. This work was supported by the NSF under Grants 0856761 and discovery/interactive-engagement labs characteristic of several
0927661
well-known labs, such as the Modeling Workshop Project [17],
E. T. Enikov is with the Department of Aerospace and Mechanical Engi-
neering, College of Engineering, University of Arizona, Tucson, AZ 85721 Socratic Dialogue Inducing Labs [18], Real Time Physics [19],
USA (e-mail: [email protected]). and Tools for Scientific Thinking [20]. The portability and low
G. Campa is with Mathworks, Inc., El Segundo, CA 90245 USA.
cost of the setup allowed the students to conduct experiments
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. during the entire semester and use the device to complete a term
Digital Object Identifier 10.1109/TE.2012.2195496 project. In addition to significantly reducing the cost of offering

0018-9359/$31.00 © 2012 IEEE

ENIKOV AND CAMPA: MECHATRONIC AEROPENDULUM 539

Fig. 1. Aeropendulum: dc-motor/propeller-driven pendulum.

Fig. 3. Adjustable pendulum with counterweight.

pendulum. These functions are implemented using a Freescale

MC9S08JM16 microcontroller. The apparatus communicates
with the controlling computer (PC, Mac, or Linux) using the
USB protocol, eliminating the need for an increasingly rare
serial port. The device is powered by two USB ports that are ca-
Fig. 2. Target board.
pable of pulling a total of 600 mA from the host computer. The
microcontroller is commanded to apply various PWM signals
to appropriate sides of the H-bridge IC drives (two P-MOS,
an experimental component, the experimental module provided two N-MOS, ZETEX ZXMHC10A07T8TA), depending on the
an opportunity to demonstrate concepts from system identifica- desired direction. When queried, the microprocessor returns
tion, nonlinear feedback control, and digital control. the average of several 12-bit analog-to-digital conversions to
MATLAB, which is then correlated through a proportionality
II. HARDWARE DESCRIPTION constant to the angle of the pendulum.
In its latest implementation, the aeropendulum kit has been
A. Aeropendulum modified to include a movable counterweight and variable
The setup consists of a small dc electric motor driven by a 5-V length rod, as shown in Fig. 3. This simple modification allows
pulse-width modulated (PWM) signal. The motor is attached to the instructor to adjust the inertia of the pendulum, as well as
the free end of a light carbon rod, while the other end of the rod the required feedback linearization action, and thus produce an
is connected to the shaft of a low-friction potentiometer. The individualized setup for each student.
potentiometer is fixed on a plastic stand at a height where the
pendulum can swing freely (see Fig. 1). A 2-in propeller (model C. Simunlink Environment
U-80) is attached to the motor shaft to produce a thrust force in The experiment also illustrates the use of the MATLAB/
order to control the angular position of the pendulum. A self- Simulink Real-Time Windows Target (RTW) environment (see
calibrating step during the initialization allows the system to Fig. 4). The RTW module performs classical control experi-
automatically find the vertical position (origin of the coordinate ments using hardware-in-the-loop simulations. Using RTW,
system). the sampling time was reduced by an order of magnitude to
5 ms. This was achieved by a built-in functionality of RTW
B. Target Board that compiles the Simulink model down to C or C++ code, and
A custom-designed circuit board produces the controlled then builds a native executable file. Removing the need for an
voltage supply for the motor via PWM with a resolution of interpreter greatly improves the efficiency of the simulation.
0.05 V (see Fig. 2). It also reads the voltage on the poten- Packet-In and Packet-Out blocks are used in the RTW model
tiometer, which is proportional to the angular position of the to communicate with the microcontroller. The Packet-Out
540 IEEE TRANSACTIONS ON EDUCATION, VOL. 55, NO. 4, NOVEMBER 2012

Fig. 4. Simulink Real-Time Windows controller.

block is used to send a request for data to the microcontroller. Fig. 5. Open-loop response.
Once the data are ready, the microcontroller sends the data to
the Packet-In block. A step function with a period of 10 ms
and a duty cycle of 50% is used to generate a query to the
microprocessor every 10 ms.

III. STUDENT EXPERIMENTS

A. Modeling
The hardware described here has been tested by senior-year
mechanical and aerospace engineering students taking their first
course in controls system design. Prior to this experiment, this
course had been a lecture-only class, so the experiment had to be
conducted as part of the regular homework assignments. Conse-
quently, the students had to be given a detailed manual on how
to install the kit.1 The project assignment asks the students to
develop a nonlinear mathematical model of the pendulum and
identify its physical parameters. The students focus on the dy-
Fig. 6. Input signal (PWM) versus sine of steady-state angle.
namics of the pendulum. (The dynamics of the electronic com-
ponents and the dc motor are assumed fast and negligible for the
sake of this experiment.) Typically, the students arrive at value of the parameter K/mg by noting that the steady-state
angle and the steady-state input are related through
(1)
(4)
where is the weight of the motor, is the length of the
pendulum, is the viscous friction coefficient, and is the thrust
force from the propeller. The students then model the resulting
thrust force as a linear function of the applied voltage, , using B. Feedback Linearization
a thrust coefficent
In their second assignment, students are asked to carry out ex-
periments and verify (4) experimentally. A typical experimental
(2)
plot is shown in Fig. 5. Using multiple input values, students
obtain pairs of steady-state angles and input voltages, as shown
resulting in a modified model
in Fig. 6. This plot presents an opportunity to observe a non-
ideal behavior associated with real systems. The motor exhibits
a dead-band, i.e., for voltages below approximately 1 V (20%
(3)
of 5 V), the propeller does not turn due to friction. The modified
model for the thrust force is therefore
Students are next asked to examine the steady-state behavior of
the plant (3). Quick analysis leads to a method for extracting the if
1http://nano.arizona.edu/pages/education/micro-mechatronics-and-con- if (5)
trols-experiment1.php if
ENIKOV AND CAMPA: MECHATRONIC AEROPENDULUM 541

Fig. 7. Implementation of feedback linearization according to (6) and (8).

Coefficients and correspond to positive and negative

input voltages and are different due to the asymmetry of the
propeller operation. Using this model and experimenting with
positive and negative inputs, the students determine approxi-
mate values of these coefficients extracted from a bidirectional Fig. 8. Feedback-linearized pendulum (“weightless”) response.
version of the plot in Fig. 6

where the factor 2 reflects the experimentally observed fact that

the propeller is twice as efficient in the forward direction as in
the reverse direction . The next step is to eliminate Fig. 9. Closed-loop system.
the observed dead-band. To this end, the students are instructed
to use a discontinuous control law due to feedback linearization. It is easy to show that under such
if conditions is a solution to (9). Therefore, if the
(6) pendulum is manually pushed to a particular angle, it remains
if
there due to the automatic adjustment of power according to (8).
which has the effect of canceling the dead-band region. As is Fig. 8 demonstrates the corresponding response. If the feedback
evident from Fig. 6, the approximate value of is 22. The linearization is not perfect due to errors in the parameter esti-
resulting model in terms of the new input variable is mation, the pendulum will have the tendency to droop or rise,
which is easy to observe visually.
(7)
C. System Identification of Linearized Plant
Finally, the feedback linearization is achieved through the use
of a nonlinear feedback in the form As a third assignment, the students are asked to identify the
system type of the linearized plant (9). Generally, steady-state
(8) errors and associated system types are covered early in the
semester, therefore this step tends to serve as a reminder and
which cancels out the nonlinear term. The resulting linear illustration of a known material. Typically, students recog-
system has a simple transfer function with two real poles at 0 nize the type-1 behavior and correctly expect to see a zero
and , respectively or negligible steady-state error. The suggested approach is to
implement a unit-feedback control of (9) by setting
(9)
(10)
Implementation of the nonlinear feedback linearization repre-
with . The resulting block diagram is shown in Fig. 9.
sented by (6) and (8) is shown in Fig. 7. Since the propeller
Students then use the step response of this system to identify
is more efficient in the forward direction, only positive com-
the values of the parameters K/mL and by deriving
mand signals are considered, and elimination of the dead zone
the following formulas relating the natural frequency and
is reduced to addition of the offset . The cancellation of the
damping of a second-order system to the physical parameters
effect of gravity is accomplished by the addition of .
of the model:
The resulting linearized system is a classical one (presented in
all textbooks on controls) and is easy to relate to the associated
lectures.
The final step in developing the plant model is to identify the (11)
parameters of (9). Due to the integrator term (pole at zero), the
system is of type 1 and will produce an unbounded response A sample experimental plot of the step response for a com-
if tested with a step input. Another interesting observation stu- mand input of 40 degrees is shown in Fig. 10. Through this ex-
dents make is that when , the pendulum is “weightless” periment, the students are able to apply the classical formulas
542 IEEE TRANSACTIONS ON EDUCATION, VOL. 55, NO. 4, NOVEMBER 2012

Fig. 10. Step response of closed-loop system. Fig. 11. Unstable response for .

presented in the course lectures to relate the plant parameters to where is the resistance of the motor coils, is the back emf
system response parameters such as overshoot and rise-time constant, and is the inductance of the motor coil. Similarly,
the rotor dynamics is governed by

(14)
(12)
where is the motor torque constant, is the aerodynamic
In a more recent implementation, the length of the rod of the torque (drag) coefficient, is the air density, and is the di-
pendulum can be adjusted. When working with shorter lengths, ameter of the propeller [22]. Solving (14) for and substituting
the inertia is reduced, and the friction in the potentiometer be- in (13), it is easy to see that that and ! are related through
comes more dominant. The unit-feedback response of such pen- a second-order nonlinear differential equation for . Conse-
dulums has no overshoot. In such cases, students can use higher quently, the input signal and the thrust force of the propeller
values of the feedback gain, e.g., , leading to an under- are also subject to the dynamics of the associated poles. With
damped response, and follow the same identification method. this explanation, students are able to examine the effect of these
Alternatively, MATLAB’s System Identification Toolbox pem two additional poles by assuming that they are approximately
command can be used to automatically estimate the parameters equal (critically damped pole pair) and reexamine the root locus
of the system. An example of its use is available online [21]. of a fourth-order system

D. Model Fidelity (15)

With the parameters fully identified, it is now possible to ex- A quick estimate of the effect of these poles can be made by
amine the validity of the model. To this end, the students are assuming that they are close to each other and by examining a
asked to apply the root locus design method and determine the system with a critically damped pole-pair
behavior of the system (9)–(10) under different values of the
feedback gain . Using the root locus method, students pre- (16)
dict that the system is stable under all values of . However,
when asked to run experiments for they re- where the actual parameters of the transfer function (9) have
alize that the response is stable for low values of the gain, and been used.
that the system loses its stability and the pendulum undergoes The students are then asked to vary the value of the time con-
unstable oscillations, beyond a certain critical gain, as shown stant until the critical gain predicted from the root-locus
in Fig. 11. This experience leads to the fourth task, where the agrees with the value obtained from experimentation .
students are asked to examine their model and propose possible The corresponding plot is shown in Fig. 12. The resulting time
reasons for the observed discrepancy. Interestingly, the majority constant is ms, which matches tachometer measure-
of the students (over 66%) provide inaccurate or superficial an- ments of the propeller [22]. This experience was well received,
swers (discussed in Section IV). Upon submission of their re- in particular by students who had struggled to come up with
sponses, students are provided with a guided solution directing an explanation for the shortcomings of the second-order model.
them to refine the model of the motor/propeller model by incor- It also illustrates the use of the variational theory of learning
porating the dynamics of the motor current and propeller rota- in the design of the experiment. According to this theory, we
tion. This results in the addition of a voltage-current equation learn through the experience of difference, rather than recog-
nizing similarities. As Marton and coworkers state [16]: “What
(13) we believe is that variation enables learners to experience the
ENIKOV AND CAMPA: MECHATRONIC AEROPENDULUM 543

Fig. 13. Design of dynamic compensator (lag) in SISOTool.

Fig. 12. Root locus of a fourth-order model incorporating motor-propeller

dynamics.

features that are critical for a particular learning as well as for

the development of certain capabilities. In other words, these
features must be experienced as dimensions of variations.”

E. Design of Dynamic Compensator

The final assignment associated with the pendulum is to de-
sign a dynamic compensator. Typical performance specifica-
tions include a steady-state error of not more than 0.5 degrees
and a setting time of not more than 3 s. Students are given the
freedom of choosing a design method and compensator type,
however most prefer using the frequency response techniques
(Bode plots). It is worth noting that when the System Identifica- Fig. 14. Step response and disturbance rejection using phase lag compensation.
tion Toolbox’s pem command is used, the extracted character-
istic polynomial usually contains a small nonzero constant term
(0.0033 in this case), which reflects the quality of the feedback of the USB. Given the slow speed of the apparatus, we utilize a
linearization 10-ms update rate. The corresponding conversion command is
` ' , resulting in
(17)
(19)
Following classroom examples, they import the plant’s transfer
function in MATLAB’s SISOTool and use its Bode plot to Similarly, students perform a phase lead (proportional deriva-
examine gain crossover frequency and stability margins. The tive) controller design where they place the gain cross-over fre-
SISOTool allows them to graphically place the poles and zeros quency around 20–30 rad/s. As expected, the resulting response
of their compensator and adjust the gain. An example of a is much faster, but with increased noise and larger overshoot.
phase lag (proportional integral) compensator with a transfer The response of the dynamically compensated plant with phase
function lag and lead controllers under multiple manually induced dis-
turbances is shown in Figs. 14 and 15, respectively.
(18)
IV. EVALUATION
is shown in Fig. 13. Since a Real-Time Windows Target oper- During the 2008–2011 academic years, the project was of-
ates in the discrete domain, testing the performance of the com- fered to three cohorts with different instructors. The impact of
pensator (18) requires a conversion of the compensator to a dig- the project was assessed through student surveys conducted at
ital (discrete) transfer function using MATLAB’s func- the end of the course following the protocol approved by the
tion. The sampling time of the target board is determined by Institutional Review Board. Additional data were drawn from
a function generator block inside the Simulink model. Since student reports. The data reported here are from a section not
the target board uses a USB-to-serial driver, the actual data taught by any of the authors; instead, the instructional mate-
transfer speed and hence the limit of the sampling rate is that rials and hardware were provided to a different instructor and his
544 IEEE TRANSACTIONS ON EDUCATION, VOL. 55, NO. 4, NOVEMBER 2012

and gain, the importance of transfer functions in capturing the

physical models, followed by the ability to deal with nonlinear
systems and time delay. Interestingly, the highest gains were
achieved in understanding of the relationship of stability and
gain. When asked to comment on the discrepancy observed be-
tween the theory and experiment, 33% of the students correctly
identified the missing rotor dynamics as a possible cause, while
56% felt that the feedback linearization somehow masked the
unstable modes or was imperfect, leading to loss of stability.
Another 11% looked for physical limitations in the system or
faulty components. It appears that the large number of miscon-
ceptions, paired with challenging the students’ confidence in
their ability to model the plant, along with providing a plausible
solution to the problem, could explain the highest gains in this
category. Further case studies would be required to confirm this
observation. Among the least understood topics was the use of
Fig. 15. Step response and disturbance rejection using phase lead Bode plots, perhaps due to the fact that it was covered at the very
compensation. end of the semester, leaving little time for practice and explo-
ration. The portability and convenience of the implementation
TABLE I of the experiment was evaluated through a second set of ques-
STUDENT FEEDBACK ON HOW WELL THE PROJECT ILLUSTRATES TECHNICAL tions, where 42.9% of the students reported that they did not
CONCEPTS (PERCENTAGE; NUMBER OF RESPONSES IN PARENTHESES) need a permanent lab, and another 42.9% had to use a teaching
assistant consultation for not more than 1 h. Only 3.6% indi-
cated that more consultation was needed, while 10.7% wanted
to have a permanent lab space dedicated to the project. The av-
erage duration for completion of the project was 7.78 h.

V. CONCLUSION
An inexpensive portable experimental setup has been de-
scribed for use as a hands-on experience for undergraduate
students taking senior-level classical control system design
courses. The project requires minimal or no supervision
without the need for a specialized laboratory space. In 10 out of
11 topics, students self-reported above average learning gains.
Highest gains were achieved through a problem that challenges
the student’s trust and belief in the theory when confronted
with an apparent contradiction with experimental observations.
Evolution of the project has shown that presenting it as a series
of short assignments allows the instructor to provide guidance
to the students without sacrificing the ability to encourage
individual experimentation. The project is particularly aimed
at helping students whose major is not electrical engineering
become familiar with the modern developments in implemen-
tation of real-time control systems. While simple, the hardware
allows demonstration of advanced concepts such as feedback
linearization. Evaluation data show that the project is well re-
ceived by students, and it can be completed independently over
an average of 8 h. Parameter variation through modification
teaching assistant. However, the results from surveying the au- of the configuration of the pendulum allows the instructor to
thors’ sections agree to within 5%–8% in most categories of the individualize each kit.
data shown here. As part of the evaluation, students were asked
questions about the technical content, as well as the implemen- REFERENCES
tation and impact of the portable experiment. Table I shows the [1] Commission on Engineering Education, “New directions in laboratory
evaluation of technical content. As expected, the majority of the instruction for engineering students,” J. Eng. Educ., vol. 58, no. 3, pp.
students found the system quite useful in illustrating the princi- 191–195, 1967.
[2] American Society of Engineering Education, Washington, DC, “The
ples of control system design. The highest benefits are derived undergraduate engineering laboratory: Final report of the Quality of
from better understanding of the relationships between stability Engineering Education Project (QEEP),” 1986.
ENIKOV AND CAMPA: MECHATRONIC AEROPENDULUM 545

[3] E. W. Ernst, “The undergraduate engineering laboratory: 1983 En- [21] E. T. Enikov, “Aeropendulum project,” 2011 [Online]. Available:
gineering Foundation Conferences, New England College, Henniker, http://aeropendulum.arizona.edu/
New Hampshire, July 24–29, 1983 Engineering Foundation, 1983. [22] V. Glowacka, “Modeling of nonlinearities in a propeller-driven pen-
[4] ABET, Baltimore, MD, “Engineering criteria 2000,” 2002. dulum,” Master’s thesis, University of Arizona, Tucson, AZ, 2009.
[5] M. Aburdene and M. El-Sharkawy, “Integrated engineering worksta-
tions in electrical engineering laboratories,” IEEE Trans. Educ., vol.
32, no. 3, pp. 404–408, Aug. 1989.
[6] J. Kadlowec, P. Lockette, E. Constans, B. Sukumaran, and G. Cleary,
“Visual beams: Tools for statics and solid mechanics,” in Proc. 32nd Eniko T. Enikov (M’10) received the M.S. degree from the Technical Univer-
ASEE/IEEE Frontiers Educ. Conf., Boston, MA, Nov. 2002, pp. sity of Budapest, Budapest, Hungary, in 1993, and the Ph.D. degree from the
T4D-7–T4D-10. University of Illinois at Chicago in 1998, both in mechanical engineering.
[7] P. Horàcek, “Laboratory experiments for control theory courses: A As a Postdoctoral Associate with the University of Minnesota, Minneapolis,
survey,” Annu. Rev. Control, vol. 24, pp. 151–162, 2000. he has worked on several projects in the area of microassembly, capacitive force
[8] R. V. de Molengraft, R. Steinbuch, and B. Kraker, “Integrating experi- sensing. Currently, he is an Associate Professor with the Aerospace and Me-
mentation into control courses,” IEEE Control Syst., vol. 25, no. 1, pp. chanical Engineering Department, University of Arizona, Tucson, where he
40–44, Feb. 2005. established the Advanced Micro- and Nanosystems Laboratory. His research
[9] R. M. Felder, “Reaching the second tier-learning and teaching styles is focused on the design and fabrication of microelectromechanical systems
in college science education,” J. Coll. Sci. Teaching, vol. 23, no. 5, pp. (MEMS) as well as developing theoretical models of multifunctional materials
286–290, 1993. used in MEMS. His group at the University of Arizona has an ongoing re-
[10] R. Dunn, “Learning styles and its relation to exceptionality at both ends search program on tactile displays, electrostatic micro-grippers for assembly of
of the spectrum,” Exceptional Children, vol. 49, pp. 496–506, 1983. MEMS, and nanoassembly of macromolecules using electrostatic fields, as well
[11] S. Papert, Mindstorms: Children, Computers, and Powerful Ideas. as development of MEMS-compatible wireless sensing platforms with biomed-
New York: Basic Books, 1980. ical applications.
[12] I. Harel and S. Papert, Constructionism. Norwood, NJ: Ablex, 1991. Dr. Enikov is a member of the professional societies of ASME and ASEE.
[13] J. Sarik and I. Kymissis, “Lab kits using the Arduino prototyping plat-
form,” in Proc. IEEE FIE, Oct. 2010, pp. T3C-1–T3C-5.
[14] P. J. Gawthrop and E. McGookin, “A Lego-based control experiment,”
IEEE Control Syst., vol. 24, no. 5, pp. 43–56, Oct. 2004. Giampiero Campa (M’00) received the Laurea degree in electrical engineering
[15] F. Marton and A. S. Booth, Learning and Awareness. Mahwah, NJ: and Ph.D. degree in robotics and automation from the University of Pisa, Pisa,
Erlbaum, 1997. Italy, in 1996 and 2000, respectively.
[16] F. Marton and A. Tsui, Classroom Discourse and the Space of He has also worked with the Industrial Control Centre, Strathclyde Uni-
Learning. Mahwah, NJ: Erlbaum, 2004. versity, Glasgow, U.K., in 1995, and with the Department of Aerospace
[17] I. Halloun and D. Hestenes, “Modeling instruction in mechanics,” Engineering, Georgia Institute of Technology, Atlanta, in 1999. From 2000
Amer. J. Phys., vol. 55, pp. 455–462, 1987. to 2008, he has served as faculty with the Flight Control Group, Department
[18] R. Hake, “Promoting student crossover to the newtonian world,” Amer. of Aerospace Engineering, West Virginia University (WVU), Morgantown.
J. Phys., vol. 55, pp. 878–884, 1987. His research at WVU involved system identification, adaptive and nonlinear
[19] D. Sokoloff, R. K. Thornton, and P. Laws, Real Time Physics Module control, fault-tolerant systems, machine vision, and sensor fusion, especially
1: Mechanics. New York: Wiley, 1998. applied to unmanned aerial vehicles (UAVs). Since January 2009, he has
[20] R. Thornton, “Tools for scientific thinking: Microcomputer-based labo- worked with MathWorks, Inc., as the Technical Evangelist for the US West
ratories for teaching physics,” Phys. Educ., vol. 22, pp. 230–238, 1987. Coast area.