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

net/publication/200744655

An autonomous educational mobile robot mediator

Article in Autonomous Robots · November 2008

DOI: 10.1007/s10514-008-9101-z

CITATIONS READS
144 1,519

3 authors, including:

Rubén Mitnik
Pontifical Catholic University of Chile
6 PUBLICATIONS 396 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rubén Mitnik on 15 January 2015.

The user has requested enhancement of the downloaded file.

Noname manuscript No.
(will be inserted by the editor)

An Autonomous Educational Mobile Robot Mediator

Ruben Mitnik · Miguel Nussbaum · Alvaro Soto

Received: date / Accepted: date

Abstract So far, most of the applications of robotic a collaborative and constructivist learning approach;
technology to education have mainly focused on sup- iii) The implementation and testing of these ideas in
porting the teaching of subjects that are closely related a real scenario, working with students at local schools.
to the Robotics field, such as robot programming, robot
Keywords Robots in education · Mobile robots ·
construction, or mechatronics. Moreover, most of the
Autonomous robot · Robot-human interaction.
applications have used the robot as an end or a passive
tool of the learning activity, where the robot has been
constructed or programmed. In this paper, we present a 1 Introduction
novel application of robotic technologies to education,
where we use the real world situatedness of a robot to With the emergence of new digital devices along with
teach non-robotic related subjects, such as math and faster and more powerful computers, new scopes have
physics. Furthermore, we also provide the robot with opened in computer assisted educational activities. As
a suitable degree of autonomy to actively guide and an example, computer simulations have offered new ed-
mediate in the development of the educational activ- ucational environments, supplying a great variety of
ity. We present our approach as an educational frame- opportunities for modeling concepts and processes. At
work based on a collaborative and constructivist learn- different educational levels, computer simulations have
ing environment, where the robot is able to act as an been successfully applied to support teaching a great
interaction mediator capable of managing the interac- variety of subjects, such as mechanics, biology, chem-
tions occurring among the working students. We illus- istry, and physics, among others, where they have proved
trate the use of this framework by a 4-step methodology to be efficient teaching tools [39,42,14,38,25].
that is used to implement two educational activities. In the area of Artificial Intelligence, intelligent tu-
These activities were tested at local schools with en- toring systems or ITSs have received great attention,
couraging results. Accordingly, the main contributions as another relevant way of using computer based tech-
of this work are: i) A novel use of a mobile robot to nology as an effective educational tool. Originally in-
illustrate and teach relevant concepts and properties of troduced in [6] and [12], an ITS tracks the work of the
the real world; ii) A novel use of robots as mediators student with the goal of identifying his main strengths
that autonomously guide an educational activity using and weaknesses. This knowledge is then used to sug-
gest suitable additional work. Using this scheme, an ITS
R. Mitnik
is able to adapt its behavior according to the student
Department of Computer Science needs. See [29] for a in-depth review of ITSs.
Pontificia Universidad Catolica de Chile With the advancement of robotic technology, the
E-mail: [email protected] scope of simulations and ITSs has extended to the phys-
M. Nussbaum ical world, allowing new educational activities to take
E-mail: [email protected] place in a real environment. Supported by sensors and
A. Soto actuators, robots are capable of exploring and interact-
E-mail: [email protected] ing with the real world. Based on these capabilities, a
2

series of educational activities have been developed to students. As an interaction mediator, an autonomous
aid and foster learning of relevant topics. As an exam- robot can define roles and supervise the appropriate and
ple, teachers have been able to increase the motivation fluent development of the educational activity. More-
and performance of their students by means of diverse over, regarding motivation, an environment based on
robot building activities, such as robotic competitions an active robot mediator may help students to develop
and algorithm testing, to name just a few [28, 31]. affective bonds with the robot [35], helping to develop a
Among current educational activities that include stronger situational interest in the educational activity
robotic technologies, it is possible to distinguish two [13,11], a form of externally controlled motivation.
relevant common factors. First, most of the approaches Based on the previous two ideas, this paper presents
aim to teach subjects closely related to the Robotics a novel application of robotic technology to primary
field [34,18, 15]. Examples of these subjects are robot and secondary school-level education. The application
programming, robot construction, artificial intelligence, consists of an autonomous mobile robot that helps stu-
algorithm development, and mechatronics [45, 46, 3, 15, dents in the creation of abstract models of relevant con-
1]. One of the notable exceptions to the previous cases, cepts and properties of the real world by physically il-
it is the early work of Papert [33], who began with the lustrating them. Furthermore, by acting as a situated
idea of using robots, and mainly simulations, to teach mediator of the educational activity, and using a collab-
subjects such as planar geometry. Secondly, most of the orative and constructivist learning approach, the robot
approaches use the robot in the educational activity is able to guide the activity, playing a key role to in-
just as a passive tool. As an example, in the case of crease the motivation and social bonds among the stu-
robot construction, the robot is the result of the as- dents. As far as we know, this is the first time that
sembling process of the students, while in the case of robotic technologies are been used in such educational
robot programming, the robot is just a physical ma- setting, opening a new paradigm to apply robots in ed-
chine that executes a set of instructions and algorithms ucation.
provided by the students. In both cases, the robot has Accordingly, the main contributions of this work
a passive role in the learning process of the students. are: i) A novel use of a mobile robot to illustrate and
While the previous approaches have proved to be teach relevant concepts and properties of the real world;
effective, research in this area has mostly left aside the ii) A novel use of robots as mediators that autonomously
possibility of using robotic technologies to teach non- guide an educational activity using a collaborative and
robotics related subjects and to use an autonomous constructivist learning approach; iii) The implementa-
robot as an active mediator of the educational activity. tion and testing of these ideas in a real scenario working
These two key ideas are at the core of the Autonomous with students at local schools.
Educational Mobile Robot Mediator presented in this This paper is organized as follows. Section 2 presents
paper. a review of previous works related to the use of robotic
In terms of teaching non-robotic related subjects, technologies in education and the use of autonomous
the embodied nature of a robot provides a way to depart robots as activity mediators. We also review some pre-
from the traditional abstract teaching scheme based on vious projects regarding the use of collaborative envi-
a blackboard to a teaching model in which the student ronments in education. Section 3 describes the details
can learn by directly observing the actions of a mobile of our approach, including two concrete examples that
robot. This opens a new teaching paradigm that, with illustrate the use of our educational framework. Section
the help of an embodied robot, can provide a more nat- 4 analyzes the results of testing our approach in local
ural setting to teach subjects such as math and physics. schools. Finally, Section 5 presents the main conclusions
For example, a mobile robot can use its rotational ca- of this work.
pabilities to illustrate angular relations, or it can use its
acceleration and velocity capabilities to illustrate rele- 2 Bibliographic review
vant kinematic principles.
In terms of the role of the robot in the educational There is an extensive list of educational initiatives that
activity, an autonomous robot provides a way to ac- use robotic technologies as pedagogical tools. As we
tively guide the development of the activity. In particu- noted before, most of these initiatives are oriented to
lar, the robot can act as an interaction mediator. Inter- teach subjects directly related to the Robotics field. In
action mediators, mainly used in collaborative and con- [37] and [41], the authors describe educational activi-
structivist learning environments [16], focus not only on ties mainly focused on robot construction using toolk-
the content, but also on the management of the virtual its, such as commercial Lego bricks. Similarly, the work
and physical interactions established among groups of in [31] uses a robot kit that allows students to assemble
 3

their own robots. As the main purpose of these activ- Social robots, such as Pearl [36] and Valerie [10],
ities is to motivate students with technology, particu- are also examples of a new generation of robots able to
larly the Robotics field, the robots are usually provided interact with people and to play the role of mediators
with appealing locomotion capabilities able to execute to specific knowledge sources. Pearl is intended to as-
engaging behaviors but under constrained situations. sist elderly individuals with mild cognitive and physical
Robot competition is another area where there have impairments, as well as support nurses in their daily ac-
been lots of activities at all educational levels [28]. In tivities. Valerie currently operates as a robot reception-
particular, large scale robot competitions, such as First ist for Newell-Simon Hall at Carnegie Mellon Univer-
1
or RobotCup 2 , have become highly popular, being sity, helping people by providing information about uni-
the focus of attention of an extensive list of educational versity members, campus directions, or retrieving data
activities, such as complete courses or summer camp from the web. Although these robots perform specific
programs. tasks, they are designed to operate in highly general
In the case of robot programming, the work in [19] environments.
describes a course oriented to program robots to achieve In order to operate in highly unconstrained envi-
high-level tasks in a structured world, leaving aside ronments, museum tour guides and social robots are
hardware and low-level issues. In the same way, in [40], provided with powerful and expensive sensors, actua-
we present our own experience teaching a mobile robotic tors, and processing units. In contrast, as we describe
class that includes the implementation of low-level robot in the next section, our educational robot mediator is
behaviors performed in a real world environment, as designed to be a low cost robotic platform, equipped
well as the implementation of high-level behaviors per- with limited hardware resources that provide enough
formed in a structured world. robustness to successfully deal with the complexity of
As we pointed out, in all the previous cases the robot a structured and specific educational activity.
plays a passive role, being the end or a tool of the ed- In terms of collaborative educational environments,
ucational activity. Furthermore, to keep the complex- computer-based technologies have been successfully used
ity and cost of the robots to a manageable level for a to foster learning, motivation, and social bonds among
medium size class, the robots are constrained to exe- students [17,8,16]. In this type of educational envi-
cute specific tasks in structured environments. In our ronments, students build knowledge by actively inter-
approach, we also constrain the robot to execute a spe- acting and assuming asymmetric roles. In particular,
cific educational activity in a structured environment, [47] demonstrated that in these collaborative settings
however, this work is distinguished from the previous a technological system able to interact independently
mentioned works by focusing on teaching non-robotic with each student can effectively control their interac-
related subjects and using the robot as an active medi- tions, supervise the educational activity, regulate tasks,
ator. rules, and roles among the students, and mediate in the
students acquisition of new knowledge. This is exactly
Regarding robots as activity mediators, some of the
the type of mediation that we envision for our robot in
most relevant robotic initiatives have focused on mu-
the educational activities presented in this work. There-
seum robots that guide their human counterparts through
fore, our robot is able to interact not only with the stu-
the museum, explaining to them the relevant aspects of
dents as a group, but also with each student indepen-
the different expositions and halls. This is the case of
dently. As we describe in Section 3, this dual flexibility
robots such as Rhino [5], Minerva [43], Sage [30], and
in the interaction with the students is a key feature of
Joe Historybot [32]. As an example, Rhino was the first
our approach.
museum tour-guide robot, installed in mid-1997 at a
museum in Germany. Rhino was responsible for greet-
ing visitors and guiding them through a fixed set of 3 Our approach: The Autonomous Educational
museum attractions. As another example, Sage, later Mobile Robot Mediator (AEMRM)
renamed as Chips, was a robot that operated in the Di-
nosaur Hall at Carnegie Museum of Natural History, As we pointed out before, the main novelties of our ap-
USA, providing tours and presenting audiovisual in- proach are twofold. On one hand, we exploit the mobil-
formation regarding bone collections. All the previous ity of our robot to illustrate relevant concepts of school
robots proved to be helpful to guide and teach museum subjects, such as physics and geometry, with the goal
visitors [9]. of helping students in the process of creating abstract
models of reality. On the other hand, we exploit the
1 http://www.usfirst.org/ autonomy and situatedness of our robot to mediate in
2 http://www. robocup.org/ the development of the educational activities, with the
4

goal of improving the motivation and social interac- used by the robot to physically illustrate relevant con-
tions among the students. This is performed using a cepts of school subjects.
collaborative and constructivist learning environment. Third, by distributing different tasks among the stu-
Next, we describe the generic aspects of the educational dents, the robot can grant unique roles to each of them.
framework that we develop to apply our ideas. Then, In this way, the robot can assign a specific software tool
we present a methodology to implement educational ac- to each student or it can allow each student to con-
tivities using this framework. Afterwards, we illustrate tribute with only part of the solution to a proposed
the use of this methodology by describing the imple- problem. By using such strategies, the robot can re-
mentation of two practical cases. quire every student to collaborate and agree with the
collectively assembled solution before continuing with
the activity, thus compelling the group of students to
3.1 General educational framework
reach a consensus. This form of consensus is the key
element that helps our framework to support a collab-
Figure 1 shows the general setting of our educational
orative and constructivist learning environment.
framework. As shown, we extend the local capabilities
of our mobile robot by providing it with remote wireless In the next section, we present a methodology that
interfaces. These remote interfaces correspond to hand- can be used to implement educational activities under
held devices distributed to each of the students. Using the proposed educational framework.
these devices, the robot can either, individually interact
with each student by sending an exclusive message to
the corresponding handheld device, or it can commu- 3.2 Methodology for robot mediation
nicate with the group as a whole, by sending a shared
message to all the students. To further specify the role of the robot as a mediator
The general educational framework presented in Fig- of the educational activities, we propose the following
ure 1 provides the robot with 3 relevant features to sup- 4-step methodology, as shown in Figure 2.
port our goals. First, the use of handheld devices, as the 1)Initialization: in the first step, the robot assigns
main communication channel between the robot and roles and supply relevant data to each of the students
each student, makes possible an effective mediation of (Figure 2a). This is achieved by sending suitable mes-
the robot in the educational activity. In effect, by using sages to the handheld device of each student. The roles
a suitable graphical interface on the touch screen dis- and data being distributed depend on the concepts be-
play of each handheld device, the robot is able to con- ing taught under the current educational activity.
trol the progress of the educational activity. Although, 2)New Problem: the next step is the formulation of
there exist more flexible communication modalities for the new problem. To achieve this, the robot perceives
human-robot interaction, such as verbal communication its surrounding and performs a set of actions that phys-
through a speech recognizer and synthesizer, or gestu- ically complete the information required to fully speci-
ral communication through a vision based gesture rec- fies the new problem (Figure 2b).
ognizer; these technologies are still not robust enough 3)Deliberation: after the new problem is presented,
to be deployed in the intended application, particularly the robot compels the students to deliberate, argue, and
if the goal is to use a low cost robotics platform. In this to collaboratively construct a common answer, thus,
sense, the use of remote handheld devices provides the forcing explanations, discussions, and negotiations among
robot with a highly structured communication chan- them (Figure 2c). In the eventual case that the group
nel that facilitates its work as mediator of the educa- does not achieve consensus, the robot repeats the set of
tional activity. Furthermore, given their reduced size actions of the previous step, providing additional infor-
and wireless communication, the handheld devices do mation that can be helpful in the deliberation process
not interfere with the students mobility and face-to-face of the students.
communication. 4)Solution: finally, when the group has achieved a
Second, besides the individual and shared commu- consensus, the robot evaluates the proposed answer. If
nication through the handheld devices, the robot can the answer is correct, the robot displays it in each re-
also communicate visually with the students as a group mote interface; otherwise, the robot physically executes
by simply performing actions in the real world. In ef- the correct answer, displaying on each handheld device
fect, since the robot and the students reside in the same suitable feedback to guide the learning process of the
physical space, any physical action that the robot per- students (Figure 2d).
forms is by default an interaction with the group of To exemplify the usage of the proposed educational
students. This visual communication is the key element framework and the methodology for robot mediation
 5

Fig. 1 General setting of the proposed educational framework. a) The robot is augmented with remote interfaces. b) The robot
can communicate independently with each student by handheld devices. c) The robot can communicate visually with the group of
students.

Fig. 2 Methodology to implement educational activities using the AEMRM. a)Initialization: robot distributes roles and relevant
data through the handheld devices. b)New Problem: robot perceives its surrounding and performs a set of actions that fully specify the
new proposed problem. c)Deliberation: robot compels the students to deliberate and to construct an answer in a collaborative manner,
requiring every group member to agree with it. d)Solution: after consensus has been achieved, the robot evaluates the proposed solution
providing suitable feedback to the students.

presented above, the following sections describe the im- In pedagogical terms, the goal of the See-You-There
plementation of two real cases used to teach concepts learning activity is to teach basic geometrical concepts
related to geometry and physics. in a more natural way, facilitating the process of cre-
ating mental abstractions of the real world. In particu-
lar, the activity aims to help in the development of the
3.3 See-You-There learning activity
“measurement sense” or “mental ruler”, defined as the
ability to estimate lengths and draw lines of a given
See-You-There is an activity aimed to teach and rein-
size [7]. This is a highly important concept that usu-
force, in a problem solving approach, geometric con-
ally current teaching techniques are not addressing in a
cepts, such as lengths, relative positions, angles, and
proper way. As an example, it was determined that in
vectors [27]. In this activity, the group of students must
USA more than 50% of students of seventh grade can-
help the robot to solve a path planning task to arrive
not measure the length of a segment when this is not
to a predefined goal location. The activity is oriented
aligned with the beginning of the ruler [7].
to second-grade school children who work in groups of
three students to complete the activity.
Figure 3a shows a schematic view of the playground 3.3.1 Robot architecture
area of the activity. This area is delimited by three arti-
ficial visual landmarks that are freely allocated by each In terms of robot architecture, one of our main goals is
of the students. The goal of the students is to select and to achieve robust autonomous navigation using a sim-
order a sequence of motions proposed by the robot, such ple and low cost robotic platform. We achieve this goal
that the robot successfully moves from an initial to a by constraining the environment of the educational ac-
goal location. If the students answer correctly, the robot tivity to be a small playground, assumed free of ob-
successfully navigate to the predefined goal, as shown stacles, and surrounded by a set of bright color arti-
in Figure 3b; otherwise, the robot end up in a different, ficial landmarks. For our application this turns to be
erroneous location, as shown in Figure 3c. a highly reasonable scenario, mainly, because we em-
6

Fig. 3 a)Initially the robot detects the landmarks and generates a problem according to the current configuration. In this case the
goal is to arrive to the green landmark. b) Path followed by the robot when the correct answer is selected by the students. c) Path
followed by the robot when an erroneous answer is selected.

bed the landmark positioning phase as a playful part of gorithms and control architecture designed to support
the activity. By using these simplifications, we achieve the See-You-There learning activity see [27].
autonomous navigation by using low cost sensors and
simple perception algorithms.
3.3.2 Robot mediation
Figure 4 shows a picture of the robot used in the
See-You-There learning activity and a diagram of its
1)Initialization: in its role as a mediator of the See-
hardware architecture. As the basic mobile platform,
You-There learning activity, the first task of the robot is
we base the robot on the Palm Pilot Robot Kit 3 . We
to assign roles and relevant data. In this case, the robot
chose this platform due to its holonomic motion capa-
assigns to each student the task of freely allocating one
bility. This feature facilitates the robot motion planning
of the visual landmarks in the playground area. This is
scheme and also produces an attractive impression on
performed by sending a message to the corresponding
the students. Mounted on the robot, the processing de-
handheld device of each student.
vice consists of a 206 MHz PocketPC equipped with
2)New Problem: after the students have positioned
wireless communication. This processing unit provides
the landmarks, the robot spins in place using its vi-
real time processing of sensors and actuators, being also
sual and range perception to detect the position of each
the main brain of the robot functionalities as a media-
landmark. Using this information, the robot generates a
tor of the educational activity.
new problem, consisting of a sequence of three straight
In terms of sensing, the robot is able to measure
motions that, if executed in the right order, will move
distance using an off-the-shelf Polaroid sonar. Vision
the robot right next to one of the landmarks, as in
algorithms are accomplished using a CMUcam, with a
the case of Figure 3b. Figure 5a shows the screenshot
resolution of 80 x 143 pixels. As this camera transmits
used to present the new problem to the students. The
data using serial communication at 115.200 bauds, real
problem is presented as a target goal location and four
time color video is not possible. Nevertheless, this cam-
suggested motions, three correct ones plus a distracter.
era has on-board color processing capabilities that allow
Each suggested motion has three parameters: the dis-
us to define color ranges to sequentially retrieve binary
tance to travel, a referential landmark, and the direc-
images only of the relevant colors of the visual land-
tion of the motion (towards or away from the referential
marks, at a rate of 17 frames per second with a final
landmark). In this way, the task of the students is to se-
resolution of 80 x 44 pixels. In this way, we developed
lect the appropriate motions, order them correctly, and
a set of single-color vision algorithms that, when com-
assign to each motion a corresponding reference land-
bined, can efficiently detect and differentiate the set of
mark. It is important to note that by using an holo-
landmarks. For further details regarding the vision al-
nomic motion base, the robot can go everywhere in the
3 http: //www.cs.cmu.edu/ pprk playground using just straight motions.
 7

Fig. 4 a) Picture of the physical robot used in the See-You-There activity, displaying the camera, sonar, and on-board PocketPC.
b) Hardware architecture of the robot.

3)Deliberation: to foster collaboration and construc-

tivism among the students, each student is allowed to
choose only one of the four possible motions suggested
by the robot. Hence, each student alone cannot con-
struct a solution, yet the group as a whole can, by
means of each student constructing a part of the so-
lution. Figure 5b shows a screenshot of a case where a
student has already selected one of the suggested mo- Fig. 6 Graph-Plotter activity scenario. The robot moves
tions and he/she has allocated this motion as the second through a straight path while a group of students observes and
one in the sequence. measures relevant data to build a 2D kinematics graph according
4)Solution: finally, once every student has selected to the requirements of the posed problem.

a motion, the robot asks the group to confirm the con-

structed answer, as shown in Figure 5c. If there is agree-
ment, the robot physically executes the selected answer;
otherwise, the students have to discuss the proposed so-
the virtual set of coordinate axes. Additionally, a phys-
lution and eventually select new motions. The discus-
ical measuring tape is provided to measure relevant dis-
sion process repeats until the group reaches a consensus.
tances. Depending on group demand, the robot may
repeat the performed motion so that the students are
able to gather all the required data to construct their
3.4 Graph-Plotter learning activity plots.

Graph-Plotter is an activity aimed to help students in As in the case of the See-You-There learning ac-
the development of skills for the construction and in- tivity, the pedagogical goal of the Graph-Plotter ac-
terpretation of 2D graphs while also reinforcing diverse tivity is to teach school subjects in a more natural
kinematics concepts. In this activity the problem posed way, in this case, graph representations and kinemat-
to a group of students is to graph, in a blank set of ics. Regarding graph representations, research shows
coordinate axes, different linear motions performed by that despite the amount of experience working with
a mobile robot. The activity is oriented to secondary- graphs, students of all ages have difficulties compre-
school teenagers who must work in groups of three to hending them [24,20]. Regarding kinematics, studies
complete the goals of the activity. have shown that students often emerge from traditional
A schematic view of the playground area of the ac- physics courses with serious misconceptions about kine-
tivity is shown in Figure 6. The robot moves follow- matics [2,22,23,44]. Even students who show good un-
ing a straight path, varying its speed and acceleration. derstanding of kinematic concepts often experience a
Students are requested to plot either position versus series of difficulties when making connections between
time or velocity versus time graphs. In their handheld graphs and both, physical concepts and the real world
devices, the students are supplied with different soft- [24,21]. These difficulties and connections are the main
ware tools that enable them to measure time, perform problems that the Graph-Plotter learning activity seeks
math calculations, take notes, and plot the graph in to overcome.
8

Fig. 5 a) A new problem along with the assigned landmark are shown in the handheld device of each student. b) A student has
selected one of the available motions and this motion disappears from the available list, c) After all the students have selected a motion,
the robot asks the group to confirm the constructed answer.

3.4.1 Robot architecture meter traveled. This accuracy level satisfies our practi-
cal purpose.
The key requirements for the Graph-Plotter learning Mounted on the robot, the overall system-controlling-
activity are motion accuracy and clock-synchronization device is a 1.4 GHz laptop computer supplied with wire-
between the robot and the remote handheld devices. less communication and a High-Resolution Hardware
Regarding motion accuracy, as the students are required Counter (HRHC). This HRHC has a resolution of one
to precisely graph the motions of the robot, these must microsecond. Using this counter and similar HRHCs
be executed with minimum acceleration, speed, distance, provided in each of the handheld devices, we implement
or time errors. Hence, fast and rigorous wheel motor a time synchronization scheme between the controlling
control is needed to successfully develop the activity. laptop and each handheld device. Under this scheme,
Regarding clock-synchronization, inaccurate time syn- we use the laptop clock as the activity master-clock. At
chronization between the handheld devices and the robot the moment of transmitting this master-clock to each of
would translate into erroneous or delayed student mea- the handheld devices, we also consider the latency time
surements. Hence, an accurate clock-synchronization al- of the wireless link. We measure this latency by inde-
gorithm is needed to overcome the variable latency-time pendently estimating the mean message-sending delay
and communication delays inherent to wireless network between the laptop and each handheld device. In this
connections. We explain next how we face these prob- way, we are able to achieve a synchronization error of
lems. less than 100 milliseconds, which is suitable for our ap-
Figure 7 shows a picture of the robot used in the plication.
Graph-Plotter learning activity and a diagram of its
hardware architecture. As the basic mobile platform, 3.4.2 Robot mediation
we use an ER1 robot 4 . We chose this platform, mainly,
because of its affordable price and accurate motions. 1)Initialization: at the beginning of the activity, and
The ER1s are differential drive robots, equipped with in order to regulate the solving process, the robot dis-
two driven wheels powered by stepper motors controlled tributes two roles among the students. While one stu-
by microsteps. High precision in the generation of the dent is assigned to be the grapher (Figure 8a), the rest
microsteps provides an accurate control of the kinemat- of the students are assigned to be data collectors (Fig-
ics of the wheels, especially regarding their angular po- ure 8b-d). The grapher is responsible of plotting the
sition, velocity, and acceleration. The microsteps are answer, while the data collectors are responsible of sup-
generated by a dedicated hardware controller, achiev- plying him with the needed data. Each data collector
ing displacement errors of less than one centimeter per can switch among the different tools, which are a blank
pad (Figure 8b), a calculator (Figure 8c), and an inter-
4 http://www.evolution.com/ active chronometer (Figure 8d).
 9

Fig. 7 a) Picture of the physical robot used in the Graph-Plotter learning activity. b) Hardware architecture of the robot.

Fig. 8 Screenshots of Graph-Plotter roles: a) Grapher has coordinate axes in which to graph the motion of the robot. Data collectors,
on the other hand, can switch among different software tools: b) A blank pad, c) A calculator, and d) A chronometer.

2)New Problem: the new problem consists of pre- 4)Solution: when a graph proposal has finally been
senting a new robot motion to the students. Accord- accepted by the whole group, the robot evaluates the
ingly, after the roles have been assigned, the robot be- graph determining whether it is correct or wrong. This
gins its motion while synchronously starting the chronome- is done by analyzing its similarity with respect to the
ters of the grapher and the data collectors. While the exact graph of the true motion. If the robot determines
robot moves, the grapher is entitled to plot the robot that the proposed graph is correct, each remote inter-
motion in his coordinate axes, where the passing of time face shows the group answer superimposed to the exact
is shown visually as a widening shade (Figures 9a-b). solution (Figure 10c) giving the students the opportu-
After the robot motion ends, the grapher is entitled to nity to analyze their accuracy and precision. In case
edit his graph (Figure 9c), to submit it, or to ask for that the robot determines that the answer is wrong, it
a motion repetition. In this last case, the robot returns repeats its motion once again, but this time the correct
to its starting position, the chronometers reset, and the graph is plotted synchronously in every device, super-
motion is repeated (Figure 9d). imposed over the plot constructed by the group (Figure
3)Deliberation: after the grapher submits his pro- 10d). In this way, the students can not only visualize
posal answer, each data collector must consent with their mistakes, but they can also observe, in real-time,
it (Figure 10a). In case that a data collector rejects how the motion is represented on the graph. This fa-
the proposed graph, the robot prompts for agreement, cilitates the understanding of the relationship between
displaying a “Come to an agreement” message in each reality and the graphical abstract representation.
handheld device (Figure 10b). Afterwards, the system
returns to the previous editing stage (Figure 9c).
10

Fig. 9 Screenshots of Graph-Plotter answer construction: a)Blank coordinate axes. Elapsed time is represented by a widening shade,
b) Grapher can plot during the motion of the robot, c) After the robot ends its motion, the Grapher can edit the plot, d) Grapher
can ask for repetition of robot motion, starting a new elapsed time.

Fig. 10 Screenshots of Graph-Plotter answer evaluation: a) Proposed graph needs to be accepted/rejected by the other group
members, b) If any member rejects the proposal, the system prompts for an agreement confirmation. c) Proposed answer is correct
and is shown over the exact answer. d) Proposed answer is wrong and the system plots in real-time the true motion of the robot.

4 Results the participating students were divided into two groups,

an experimental one which used the AEMRM, and a
In this section, we present the main results of testing control group that did not.
the See-You-There and Graph-Plotter learning activi- In the case of See-You-There learning activity, the
ties at local schools. We start the section by provid- control group worked with a paper based version of the
ing details about the testing scenarios and experimental activity performed by the experimental group, as shown
conditions. Afterwards, we describe our major findings, in Figure 12a-b. Figure 12-a shows an example of the
focusing on 4 main aspects of our proposed educational answer sheet used by the students. Each answer sheet
framework: a) Robot Autonomy, b) Learning of tar- indicates a landmark (small circle) assigned to the stu-
get concepts, c) Collaboration and motivation among dent. As in the experimental case, the students also
the students, and d) Social interactions among the stu- worked in groups of three to solve each problem (Figure
dents. 12-b). The solution to each problem consisted in select-
ing and sorting a set of three robot motions, where each
student must indicate a motion relative to its assigned
4.1 Testing scenario and experimental conditions landmark.
In Graph-Plotter learning activity, the control group
Figure 11 shows two of the testing scenarios. They cor- worked with a computer based simulation of the activ-
respond to regular classrooms at local schools. To ana- ity performed by the experimental group. This simula-
lyze the educational impact of the proposed framework, tion consisted of a virtual robot presented on a com-
 11

Fig. 11 Students using AEMRMs: a) See-You-There activity, b) Graph-Plotter activity.

puter screen, as shown in Figure 12-c. Besides the vir- tional activities. We organize the analysis in terms of
tual representation of the robot, the activity remained the three traditional modules of a robotic system: ac-
the same. Each student had a handheld device wire- tion, perception, and reasoning.
lessly connected to the computer running the simula-
tion (Figure 12-d). All the main features of the original In terms of action, the robotic platforms used in
Graph Plotter activity (motion repetition, face-to-face both educational activities were highly robust. In the
collaboration, consensus, etc.) were also present in the case of See-You-There activity, as expected, the holo-
simulation. Furthermore, as in the experimental case, nomic motion of the robot made a very attractive im-
the simulation also provided a virtual scaled measuring- pression on the students. Furthermore, the straight mo-
tape underneath the animation. The set of daily exer- tions of the robot presented an accuracy suitable to
cises was common for both groups. In this sense, the conduct the activity. The main limitation of the robot
main difference between the activity performed by the was given by the lifespan of its batteries. In particu-
experimental and the control groups was the presence of lar, when the batteries of the robot were slightly dis-
an embedded robot able to illustrate the relevant con- charged, the power consumption of the motors caused
cepts in the real world instead of a computer screen. a voltage drop when the AEMRM was in motion. This
The See-You-There learning activity was tested with in turn affected the quality of the video images ac-
twelve 7th-grade students at a local public school, as- quired by the robot, introducing important errors in
signing another six students to be part of the control its perception modules, particularly, the landmark de-
group. The students worked with the activity during 6 tection algorithms. In practice, we solved this problem
sessions of 30 minutes each. The Graph-Plotter learn- by changing the set of batteries at the middle of each
ing activity was tested at two local schools, one public session. In the case of the Graph-Plotter activity, both
and one private, being used by a total of 26 10th-grade the accuracy of the robot motion and the time syn-
students. In this case, the students in the public school chronization scheme were highly robust during all the
worked with the activity during 6 sessions of 30 minutes experimental sessions. Also, the battery charge of the
each, while the students in the private school worked robot lasted for a complete educational session with-
during 4 sections of 60 minutes each. Table 1 shows out causing problems, therefore each robot was able to
further details about the testing conditions of both ac- guide the complete activity without interruptions.
tivities.
It is important to note that in the case of the See- In terms of perception, the AEMRM used in the
You-There learning activity, there was just 1 robot avail- See-You-There activity presented serious problems at
able, therefore the different groups worked sequentially. the beginning of the testing period. These problems oc-
In the case of Graph-Plotter, there were 5 robots avail- curred as the testing location provided by the school
able, then all the experimental groups worked in paral- had walls and floor painted in colors similar to those
lel in the same room. of the landmarks. As a consequence the landmark de-
tection algorithm worked poorly. In practice, the prob-
lems were solved by placing a white carpet over the
4.2 Robot autonomy playground floor and situating a black cardboard be-
hind the landmarks (Figure 11a). An alternative solu-
In this section, we analyze the performance of the AEM- tion is to provide the robot with a more diverse set of
RMs in terms of their autonomy to conduct the educa- landmarks. In the case of Graph-Plotter, the perception
12

Fig. 12 Activities performed by the control groups. See-You-There: a) Answer sheet used by the students, b) Students working in
the activity. Graph-Plotter: c) Simulation used by the students, d) Students working in the activity.

Table 1 Descriptive information about the experimental conditions of the AEMRM activities.

Activity Type of Grade Participating Experimental Control Num. Session Overall

school students group size group size Sessions length [min] length [min]

See-You-There Public 7th 18 12 6 6 30 180

Graph-Plotter Public 10th 29 14 15 6 30 180
Graph-Plotter Private 10th 23 12 11 4 60 240

modules, mainly based on the micro-steps counters and 4.3 Learning of target concepts
HRHCs did not present any practical problem.
Regarding academic issues, by means of a pretest-posttest
scheme, we were able to determine how the proposed ac-
tivities foster learning. The tests used for each activity
In terms of reasoning, in both learning activities the aimed to determine the knowledge and understanding
AEMRMs were able to execute their planning strategies of the students regarding the subject matter taught by
without inconveniences. Two supervisors were present the activity. As we detailed before, See-You-There fo-
during each experimental session, however, the AEM- cused on teaching distances and angles, while Graph-
RMs did not need major assistance being able to au- Plotter focused on teaching kinematics and graph con-
tonomously guide the complete sessions. In this way, the struction and interpretation.
main role of the supervisors was to clarify conceptual In the case of See-You-There, the instrument used
doubts of the students, such as the difference between to measure the proficiency of the students was designed
position and velocity graphs or relations between vari- by our group according to the target concepts of the ac-
ables, among others. On average, during each session tivity. An example question of this test can be seen in
just one of the AEMRMs required assistance, and only Figure 13-a. As can be seen, this question is very simi-
once throughout the session. This assistance was mainly lar to the paper-based exercises, with the difference that
required because of problems with the wireless network, no particular landmark (small circle) is assigned to the
such as the disconnection of a handheld device. Given student, since each student must provide a complete
that the AEMRM software was provided with recovery answer. In the case of Graph-Plotter, the proficiency of
mechanisms, after the assistance, the AEMRMs were the students was measured by the Test of Understand-
capable of retaking the activity at the point the failure ing Graphs in Kinematics (TUG-K) [4], a test whose re-
occurred, neither losing any data nor needing to restart liability and internal consistency have been determined
the educational activity. by means of the “KR-20” coefficient, the point-biserial
 13

coefficient, and the Ferguson’s delta, among others [4]. Table 4 Post-activity survey: Motivation and Collaboration
An example question of this test is shown in Figure mean results. Scale ranges from -2 to 2.
13-b. Group Collaboration Motivation
To determine and compare the impact of the exper-
Experimental 1.27 1.45
imental and control treatments, we used an ANCOVA Control 1.04 1.22
analysis. The required normality assumption was sup-
ported by the results obtained from the Kolmogorov-
Smirnov test. The treatment F-test value was found to Table 5 Variation of the number of students assigned to each
be 6.169 (p-value = 0.022) at a significance level of 95% social-appreciation category before and after the activities.
(Table 2), indicating that there exist significant differ-
Control Experimental
ences between the posttest scores of the experimental
and control groups. The estimated marginal means of I like him/her very much 7 11
these analysis allow us to analyze which of these groups I like him/her 22 36
I am indifferent of him/her 7 10
attained a greater score increase (Table 3). It can be I dislike him/her 3 -2
seen that, based on a pretest score of 8.04, the experi- I dislike him/her very much 0 -1
mental group improved its score in 4.354 (54.2%) cor-
rect answers while the control group accomplished an
increase of 2.166 (26.9%) correct answers. Thus, the ef- motivating” versus a 26% of the students of the con-
fect of the experimental treatment outperformed the trol groups. Regarding collaboration, every student of
control one, doubling the effect of this last. the experimental group believed to have collaborated
Regarding gender, the analysis showed that there either “more than usual” (55%) or “very much” (45%)
were no significant differences between the learning ac- (Figure 14).
complished by boys and girls throughout the exper-
iments. Regarding previous knowledge, pretest scores
were found to be not correlated with the improvement 4.5 Social interactions among the students
accomplished after the treatment [26]. Thus, the AEMRM
was similarly effective among the participating students, Finally, social interactions were measured using sociograms
independent of their gender or previous subject under- in which each student scored his/her social apprecia-
standing. tion of each of his/her classmates. When comparing the
results of the sociograms, completed before and after
the activities, it could be seen that the enhancement of
4.4 Collaboration and motivation among the students the social bonds of the experimental students surpassed
that of the control students.
Collaboration and motivation were analyzed based on Table 5 shows the variation of the number of stu-
qualitative in-site observations and quantitative results dents assigned to each social-appreciation category be-
obtained from a post activity survey. Comparing the fore and after the activities. While in the control groups
experimental and control groups, a greater amount of the number of students positively classified (top two
collaborative interactions were observed in the students categories) increased in 29 students (87.8%), the same
of the experimental groups. Moreover, in these groups increase in the experimental groups was of 47 students
it was common to see all members involved in dis- (235%). Moreover, in the experimental groups the num-
cussions and explanations, situations not often seen in ber of students evaluated negatively (bottom two cat-
the control groups where discussions were less frequent egories) diminished in 3 students (out of 4), while in
and usually between just two group members; the third the control groups this number increased from 0 to 3
member usually had a completely passive role. Regard- students.
ing motivation, during the last session, students of the
experimental groups usually expressed their wish to
continue working with this kind of activities. On the 5 Conclusions
contrary, students of the control groups usually showed
and verbally expressed their boredom after two activity The Autonomous Educational Robot Mediator has shown
sessions. to be a powerful educational tool capable of becoming
The previous observations were consistent with the an active actor in the development of educational ac-
results of the post activity surveys, shown in Table 4 tivities. After testing the AEMRMs in three different
and Figure 14. As it can be seen, 42% of the experi- schools, they successfully proved to be capable of au-
mental group students found the activities to be “very tonomously guide educational activities, foster the cre-
14

Fig. 13 a) See-You-There pre-post test exercise example. b) Graph-Plotter pre-post test question example.

Table 2 Results from ANCOVA: Dependent variable = (Posttest score); Factor = (Treatment group); Covariates =(Pretest score,
Assistance).
Type III
Source Sum of Squares df Mean Square F Sig.
Corrected Model 352.526a 3 117.509 32.503 .000
Intercept 13.501 1 13.501 3.734 .068
Assistance 25.330 1 25.330 7.006 .016
Pretest 340.215 1 340.215 94.103 .000
Group 22.400 1 22.400 6.196 .022
Error 68.692 19 3.615
Total 3383.000 23
Corrected Total 421.217 22

a R Squared = 0.837 (Adjusted R Squared = 0.811)

Table 3 Estimated marginal means of ANCOVA.

95% Confidence Interval
Group Mean Diff. Std. Error
Lower Bound Upper Bound
Robot 12.394b 4.354b .578 11.185 13.604
Simulation 10.206b 2.166b .606 8.937 11.475

b Evaluated at covariates appeared in the model: Pretest score = 8.04, Assistance = 3.83.

ation of abstract representations of relevant concepts of developing free-riding behaviors, compelling them to
the real world, and effectively mediate in the develop- work as a team. An example of this interaction rules
ment of the proposed activities. was the consensus requirement implemented in both ac-
In relation to its teaching potential, it has been tivities, which forced the students to explain the answer
shown than the proposed educational framework can to any non-convinced group member. In the same way,
help students to understand different physical and math- other rules may be implemented, such as the order in
ematical concepts, such as geometry and kinematics. which the different students must answer, or the pos-
Students who worked with the AEMRMs experienced sibility for other group members to edit or correct a
a substantial increase in their understanding of relevant teammate answer.
target concepts. When compared with students who Regarding collaboration, teamwork among students
worked with similar but non robotic-based activities, was mainly achieved based on the wireless remote inter-
the AEMRM users significantly outperformed them, at- faces, since it was using these devices that the robot was
taining an increase in their test scores of nearly twice capable of distributing roles among the students, and of
the increase observed in the non-AEMRM users. providing different software tools and hints about the
Regarding mediation, since the students handed the proposed problems. In this way, the robot was not only
activity solution through the AEMRM remote inter- able to guide the team of students to pursue a com-
faces, the robot was able to mediate the interactions mon goal, but it was also able to provide unique ca-
among the students by establishing interaction rules pabilities to each student, fostering collaboration and
and guaranteeing their fulfillment. With these interac- inhibiting free-riding behaviors. This establishes a need
tion rules, the robot could also prevent students from for collaboration and interaction among the students
 15

Fig. 14 Post-activity survey: Histograms of the results of motivation and collaboration questions.

that emerges in a natural way. Examples of this were 5. W. Burgard, A. B. Cremers, D. Fox, D. Hahnel, G. Lake-
the landmark assignation in the See-You-There activity, meyer, D. Schulz, W. Steiner, and S. Thrun. Experiences
with an interactive museum tour-guide robot. Artificial In-
or the tools and roles distribution in the Graph-Plotter telligence, 114(1-2):3–55, 1999.
activity. 6. J. R. Carbonell. AI in CAI: An artificial-intelligence ap-
It is important to note that in relation with tech- proach to computer-assisted instruction. IEEE Trans. Man-
nology usability, students of the three schools needed Machine Systems, 11(4):190–202, 1970.
7. D.H. Clements. Teaching length measurement: Research
no more than half a session to become proficient users challenges. School Science and Mathematics, 99(1):5–11,
of the technology. This is highly remarkable given that 1999.
most of the students had never used a handheld device 8. P. Dillenbourg. Collaborative learning: cognitive and com-
or interacted with a robot before. Also, the pieces of putational approaches. Pergamon-Elsevier Science Ltd., Ox-
ford, England, 1999.
hardware did not interfere with the face-to-face commu- 9. T. Fong, I. Nourbakhsh, and K. Dautenhahn. A survey of so-
nication among the students, a key factor for achieving cially interactive robots. Robotics and Autonomous Systems,
good results in collaborative work groups. 42(3):143–166, 2003.
10. R. Gockley, A. Bruce, J. Forlizzi, M. P. Michalowski,
Finally, we believe that part of the success of the
A. Mundell, S. Rosenthal, B. P. Sellner, R. Simmons,
proposed educational framework resides in the fact that, K. Snipes, A. Schultz, and J. Wang. Designing robots for
even though it has to operate in real time in the real- long-term social interaction. In Proc. of IEEE/RSJ Int.
world, the activities implemented possess a high-level of Conf. on Intelligent Robots and Systems (IROS), pages
2199–2204, 2005.
structure that helps the robot to correctly guide the ac-
11. J. M. Harakiewicz, K. E. Barron, J. M. Tauer, S. M. Carter,
tivities and the students through them. As the Robotics and A. J. Elliot. Short-term and long-term consequences of
field advances and more powerful and more adaptive achievement goals in college: Predicting continued interest
robotic technologies emerge, we will be able to increase and performance over time. Journal of Educational Psychol-
ogy, 92:316–330, 2000.
the degree of autonomy of the AEMRMs. In particular,
12. J. Hartley and D. H. Sleeman. Towards more intelligent
we plan to include more sophisticated planning strate- teaching systems. Int. Journal of Man-Machine Studies,
gies on the AEMRMs, that include mechanisms to help 5:215–236, 1973.
the robots to decide the best way to proceed with the 13. S. Hidi. An interest researcher’s perspective on the effects
of extrinsic and intrinsic factors on motivation. Intrinsic
educational activities, based on an estimation of the
Motivation: Controversies and New Directions, pages 309–
learning achievements and motivation of the students. 339, 2000.
14. W. D. Ihlenfeldt. Virtual reality in chemistry. Journal of
Molecular Modeling, 3(9):386–402, 1997.
15. M. Jansen, M. Oelinger, K. Hoeksema, and U. Hoppe. An in-
References teractive maze scenario with physical robots and other smart
devices. In Proc. of 2nd IEEE Int. Workshop on Wireless
1. D. Ahlgren and I. Verner. An international view of robotics and Mobile Technologies in Education (WMTE’04), 2004.
as an educational medium. In Int. Conf. on Engineering 16. P. Jermann, A. Soller, and M. Muehlenbrock. From mirroring
Education, 2002. to guiding: A review of the state of the art technology for
2. A.B. Arons. A Guide to Introductory Physics Teaching. John supporting collaborative learning. In European Perspectives
Wiley, 1990. on Computer-Supported Collaborative Learning, EuroCSCL,
3. R. Avanzato. Mobile robot navigation contest for undergrad- pages 324–331, 2001.
uate design and k-12 outreach. In Proc. of Conf. of American 17. D.W. Johnson and R. Johnson. Learning Together And
Society for Engineering Education (ASEE), 2002. Alone: Cooperative, Competitive, and Individualistic Learn-
4. R. J. Beichner. Testing student interpretation of kinematics ing (5th Ed.). Englewood Cliffs, Prentice-Hall, New Jersey,
graphs. American Journal of Physics, 62(8):750–762, 1994. USA, 1999.
 View publication stats
16

18. F. Klassner and S. Andreson. Lego mindstorms: Not just for 39. D. V. Schroeder and T. A. Moore. A computer-simulated
k-12 anymore. IEEE Robotics and Automation Magazine, stern-gerlach laboratory. American Journal of Physics,
10(2):12–18, 2003. 61:798–805, 1993.
19. J. Lalonde, C. Bartley, and I. Nourbakhsh. Mobile robot 40. A. Soto, P. Espinace, and R. Mitnik. A mobile robotics course
programming in education. In IEEE Int. Conf. on Robotics for undergraduate students in computer science. In Proc. of
and Automation (ICRA), 2006. IEEE Latin American Robotics Symposium (LARS), pages
20. G. Leinhardt, O. Zaslavsky, and M.K. Stein. Functions, 187–192, 2006.
graphs, and graphing: Task, learning, and teaching. Review 41. C. Stein. Botball - Autonomous students engineering au-
of Educational Research, 60(1):1–64, 1990. tonomous robots. In Proc. of Conf. of American Society for
21. M.C. Linn, J.W. Layman, and R. Nachmias. Cognitive Con- Engineering Education (ASEE), 2002.
sequences of Microcomputer-Based Laboratories: Graphing 42. P. K. Tao. Confronting students alternative conceptions in
Skills Development. Contemporary Educational Psychology, mechanics with the force and motion microworld. Computers
12(3):244–253, 1987. in Physics, 11(2):199–207, 1997.
22. L.C. McDermott. Research on conceptual understanding in 43. S. Thrun, M. Bennewitz, W. Burgard, A.B. Cremers, F. Del-
mechanics. Physics Today, 37(7):24–32, 1984. laert, D. Fox, D. Haehnel, C. Rosenberg, N. Roy, J. Schulte,
23. L.C. McDermott. Millikan Lecture 1990: What we teach and D. Schulz. Minerva: A second generation mobile tour-
and what is learned - Closing the gap. American Journal guide robot. In Proc. of the IEEE Int. Conf. on Robotics
of Physics, 59(4):301–315, 1991. and Automation (ICRA), pages 1999–2005, 1999.
24. L.C. McDermott, M.L. Rosenquist, and E.H. Van Zee. Stu- 44. D.E. Trowbridge and L.C. McDermott. Investigation of stu-
dent difficulties in connecting graphs and physics: Examples dent understanding of the concept of velocity in one dimen-
from kinematics. American Journal of Physics, 55(6):503– sion. American Journal of Physics, 48(12):1020–1028, 1980.
513, 1987. 45. E. Wang and R. Wang. Using legos and robolab (LabVIEW)
25. T. Mikropoulos, A. Katsikis, E. Nikolow, and P. Tsakalis. with elementary school children. In 31st Conf. on Frontiers
Virtual environments in biology teaching. Journal of Biolog- in Education, volume 1, pages T2E–T11, 2001.
ical Education, 37(4):176–181, 2003. 46. J. Weinberg, G. Engel, K. Gu, C. Karacal, S. Smith,
26. R. Mitnik. The robot as an autonomous mediator of the W. White, and X. Yu. A multidisciplinary model for us-
learning experience and the social interactions. PhD thesis, ing robotics in engineering education. In Proc. of Conf. of
Dept. of Computer Science, Pontificia Universidad Catolica American Society for Engineering Education (ASEE), 2001.
de Chile, 2008. 47. G. Zurita and M. Nussbaum. Computer supported collabora-
27. R. Mitnik, M. Nussbaum, and A. Soto. Mobile robotic sup- tive learning using wirelessly interconnected handheld com-
ported collaborative learning (MRSCL). In Lecture Notes in puters. Computers & Education, 42(3):289–314, 2004.
Artificial Intelligence, volume 3315, pages 912–921. Springer
Verlag, 2004.
28. R. Murphy. Competing for a robotics education. IEEE
Robotics & Automation Society Magazine, June:44–55, 2001.
29. T. Murray. Authoring intelligent tutoring systems: An anal-
ysis of the state of the art. Int. Journal of Artificial Intelli-
gence in Education, 10:98–129, 1999.
30. I. Nourbakhsh, J. Bobenage, S. Grange, R. Lutz, R. Meyer,
and A. Soto. An affective mobile educator with a full-time
job. Artificial Intelligence, 114(1-2):95–124, 1999.
31. I. Nourbakhsh, E. Hammer, K. Crowley, and K. Wilkinson.
Formal measures of learning in a secondary school mobile
robotics contest. In IEEE Int. Conf. on Robotics and Au-
tomation (ICRA), 2004.
32. I. Nourbakhsh, C. Kunz, and T. Willeke. The mobot mu-
seum robot installations: A five year experiment. In Proc.
of IEEE/RSJ Int. Conf. on Intelligent Robots and Systems
(IROS), pages 3636–3641, 2003.
33. S. Papert. Mindstorms: Children, Computers, and Powerful
Ideas. Basic Books, Inc., New York, 1980.
34. M. Petre and B. Price. Using robotics to motivate ‘back door’
learning. Education and Information Technologies, 9(2):147–
158, 2004.
35. J. Piaget, T. A. Brown, C. E. Kaegi, and M. R. Rosenzweig.
Intelligence and affectivity. Their relationship during child
development. Annual Reviews Monograph, 1981.
36. J. Pineau, M. Montemerlo, M. Pollack, N. Roy, and S. Thrun.
Towards robotic assistants in nursing homes: Challenges and
results. Robotics and Autonomous Systems, 42(3-4):271–281,
2003.
37. M. Rosenblatt and H. Choset. Designing and implementing
hands-on robotics labs. IEEE Intelligent Systems and their
Applications, 15(6):32–39, 2000.
38. W. Rourk. Virtual biochemistry - A case study. Future Gen-
eration Computer Systems, 17:7–14, 2000.