In this proposal, we design the next generation robot laboratory. We have
conquered LEGO robot instruction, and now we are ready to look beyond it. This step is
a difficult one, since higher level robots have been designed for research, not for
classroom use. In addition, we want to make this laboratory accessible to as wide a range
of instructors and students as possible. We especially want to include those that normally
are precluded from state-of-the-art research. In order to accomplish these complex goals,
we have designed an approach to combine the development of the hardware, software,
Project-based learning can be an effective learning method. In a pilot program
that compared a project based introductory programming course to a traditionally
lectured course, most of the project students “felt they were engaged in meaningful
In addition, whereas only 5 percent of the PBL class found the course
overwhelming, over 30 percent of the matched-CS1 class found themselves in this
category. The PBL class had most students indicating that they felt positive about the
course, whereas the matched-CS1 class had the majority of students indicating that they
felt either bored or anxious about the course” [Greening et al 1997, p. 204]. They also
found that the course developed several skills, including problem solving, group work,
software development, report writing, verbal skills in computer science and testing of
The expected outcomes of the proposed work are as follows:
Full development of innovative materials for teaching project-based robotics
modules in a variety of computer science courses based on a prototype developed
at Bryn Mawr College.
Multiple assessments of the effectiveness of the materials at different types of
institutions serving students with diverse backgrounds and career goals.
Longitudinal studies will be performed to assess the long-term affects of the
curriculum on the students and faculty.
We have 8 faculty members at 6 test sites who are prepared to use the materials in
Fall 2003. We will be recruiting 8 to 10 additional schools in each of Years 2 and
3 to implement the curriculum in their schools. This will result in a test set of 22
to 26 schools over three years.
Dissemination of information about the developed materials through a web site,
conference papers and presentations, and publication of a textbook.
Self-sustaining national distribution through the web site archives and discussion
groups, as well as the textbook distribution by its publisher.1
2 Results from Prior NSF Support
Deepak Kumar and Lisa Meeden were awarded a two year NSF ILI-IP grant
starting in 1996 in the amount of $58,000 for a proposal entitled "A robot laboratory for
teaching artificial intelligence."2
The objective of this project was to improve the
Publisher is still to be determined. Talks have already occurred with acquisitions editors at Prentice Hall
and MIT Press. Both were interested in the book, although no formal agreements have been made.
Award Number: NSF ILI-IP 9651472
instruction of undergraduate AI courses by providing a unifying theme based on simple,
LEGO-based robots with Handyboard controllers. Each topic in AI was presented as a
robot task. Students then built their own robots and programmed them to accomplish
As a result of this grant, curricular materials were developed and integrated into
undergraduate courses taught at Bryn Mawr College and Swarthmore College [Kumar
1998, Kumar and Meeden 1998a, Kumar and Meeden 1998b]. Furthermore, a resource
kit of laboratory materials was prepared and distributed as part of the dissemination
component to enable other schools to adapt materials that developed for the project
[Kumar and Meeden, 1998c]. Kumar and Meeden have documented at least 50 faculty
members who have used these materials to create similar laboratories at their own
3 Goals and Objectives
The LEGO-based robot has served undergraduate faculty well for the last five
Since the curriculum development by Kumar and Meeden, the Handyboard
controller has been redesigned as a mass market consumer product called LEGO
Mindstorms. These developments have enabled the introduction of LEGO-based robots
into middle schools, high schools, and even elementary schools [for example, see the
Kiss Institute's Botball competition at www.kipr.org]. Because of this, more and more
incoming undergraduate students are already quite knowledgeable about basic robotics
concepts and are ready to delve more deeply into research-level questions.
The more advanced robots we are proposing to use now are quite similar to those
used in various state-of-the-art robot deployments such as Mars rovers, museum tour
guides, and hazardous landscape exploration. However, in introducing these more
advanced robots into our own courses we have found that they can be difficult and
intimidating for undergraduates to use. This is mostly due to the fact that the robots are
primarily marketed for high-level research or industrial deployment and there isn't
sufficient usable documentation and software for novices.
We propose to change the way that topics in Artificial Intelligence are taught by
developing software and curriculum for a wide variety of project modules (described in
The project-based modules can be used as an independent course or
integrated into other courses (see Section 4.3 for a discussion of an independent course
using the modules and Appendix A for examples of how the modules could be integrated
into existing courses).
Our goal is to make research-level robotics hardware and methodologies
accessible to computer science faculty who may not have robotics experience and to their
undergraduate students. Due to past successes with project-based learning, we believe
that undergraduate students who are engaged by research-level projects of this kind will
be more likely to do well in the rest of their computer science studies and, ultimately, to
attend graduate school. We will track the grades, attitudes, and career paths of students
who take courses using our curriculum and to compare them to other students graduating
with computer science degrees from the same institutions who have not been exposed to
our curriculum to assess the success of the project in meeting our objectives.
4 Detailed Project Plan
We look to accomplish our objectives by building a series of modules and then
mentoring interested faculty at other institutions. We will provide these faculty with a
solid foundation for introducing research-level robotics by recommending a hardware
platform (and associated simulator) that is more advanced than the Handyboard but still
has good user support, by providing a software system designed for experimenting with a
wide range of more complex robot controllers, and by introducing a curriculum
integrated with the robot platform and software system that could be used in a variety of
Mentored faculty will be encouraged to submit CCLI-A&I proposals in order to
obtain the recommended robot hardware. However, many of the materials we will
provide can be used with a robot simulator alone.3 In addition, a number of faculty
interested in participating in this project have already acquired some of the recommended
hardware, but have not yet been able to introduce it effectively into their courses (for
example, see the attached letters of intent from Bowdoin and Bloomsburg University).
Our project proposes the integration of three key components to creating the next
generation robot laboratory: a hardware platform, a software platform, and a curriculum.
In the next sections, each of these components will be described in detail.
While many of the modules can be done in simulation, students will benefit more from using the robot
hardware. Interaction with the world can be simulated, but not fully realized. The simulator will be unable
to run the vision module and the gripper is not included in the simulator. Additionally, without the robot
hardware, project opportunities during and after the course will be lost. We believe that students will have
a much richer project experience through the use of the hardware.
4.1 Hardware Platform: The Pioneer 2 Robot
Kurt Konolige designed the Pioneer robot about seven years ago as a small
research platform that could be used for advanced robotics classes. The technology has
been licensed to ActivMedia, allowing other faculty to buy the robots and use them for
teaching and research.
The robot hardware is very reliable and robust, making it an
excellent choice for a classroom situation. According to ActivMedia, there are 1,000
Pioneer robots (I and II) in the field. Of these robots, only 10 are sent in for repair each
The configuration that we are proposing for this work includes an embedded
computer, a front and rear sonar array (used for obstacle detection and mapping), a
gripper mounted to the front of the robot (used to create more interesting lab scenarios,
allowing the systems to include manipulation in addition to mobility), a single camera
pan-tilt-zoom head (used for mobility, planning and teaching students vision algorithms),
wireless Ethernet (used for programming and controlling the robot from a remote
computer, either in the lab or across the internet, and for multi-robot communication),
and a rear bumper.
In addition to supporting our modules well, this robot configuration allows the
robot to be used for student research projects that go beyond the course materials.
4.2 Software Platform: Pyro
Pyro stands for Python Robotics. The purpose of the Pyro software is to provide a
stable, integrated, environment that can be used for experimenting with robot controllers
on several robot platforms as well as simulators. Currently, the robots supported include:
the Pioneer family (Pioneer and Pioneer 2 robots) and the Khepera family (Khepera and
Khepera 2 robots). There are also two simulators available, both of which simulate the
Pioneer family of robots.
Pyro has the ability to define different styles of controllers. For example, the
control system could be a neural network, a subsumption architecture, a collection of
fuzzy logic behaviors, or a symbolic planner. Any program that controls the robot
(physical or simulated) is referred to as the brain. It is written in Python and usually
involves extending existing class libraries.
The libraries help simplify robot specific features and provide insulation from the
lowest level details concerning hardware drivers. In fact, the abstraction provided
uniformly accommodates the use of actual physical robots or their simulations even
though vastly different drivers and communication protocols may be used underneath the
abstraction layer. Consequently, a robot experimenter can concentrate on the behaviorlevel details of the robot. Robots may feature a disparate set of sensors and movements,
and yet, depending on the specific robots or simulators used, Pyro provides a uniform
way of accessing those features without getting bogged down by low-level details.
Pyro also provides facilities for visualization of various aspects of a robot
experiment. Users can easily extend the visualization facilities by providing additional
Python code as needed in a particular experiment. For example, you can easily create a
graph to plot some aspect of a brain with just a few lines of code.
In picking a development language for our project pilot at Bryn Mawr, we were
faced with a series of constraints: we wanted a system that was easy for beginning
students to learn, provided a modern, object-oriented paradigm, would run on many
platforms, would allow exploration of many different control paradigms and
methodologies, would remain useful as users gained expertise, could be easily extended,
and allowed for creating modern-looking visualizations.
First, we examined existing projects to see if any fit our constraints. There are
many open sourced robotics programming environments available; however, most are
committed to a particular control strategy. Separating the control strategy code from the
rest of the system code seemed to require a major rewrite in all cases that we examined.
However, Teambots is one open source project that satisfied many of our goals
[Balch 1998]. Teambots is written in Java, and, therefore, is object-oriented. However,
because security is of such importance in Java, there are some additional burdens placed
on the programmer. For example, multiple inheritance must be implemented through
single inheritance combined with interfaces. Although Sun has been developing a
standard 3D programming interface for Java, Teambots has not yet taken advantage of
We decided to build a prototype using the extensible modeling language XML
and C++ [Blank 1999]. Although this system had some nice qualities derived from its
XML roots, it turned out to have all the complexities of XML and C++ combined, and
was therefore difficult for introductory students to learn.
Having learned from our first prototype, we decided to build another, but this time
to focus on the usability from the perspective of the new user. We found that the
language Python meets many of our goals. Python is an object-oriented, interpreted
language that recently has been used as an introductory programming language, as well
as for solving real-world complex programming problems. For example, [Prechelt 2000]
found in some specific searching and string-processing tests that Python was better than
Java in terms of run-time and memory consumption, and not much worse than C or C++.
Python has a fairly clean object and inheritance syntax that supports multiple
inheritance. It also has bindings to TCL and OpenGL, two mature graphics APIs for 2D
and 3D drawing, respectively.
Pyro was designed so that all aspects of the robot control could be studied, and
altered, by non-experts. For example, we have implemented most of the Saphira fuzzy
logic behavior engine in 200 lines of Python code.
Pyro was successfully used in the Spring 2002 semester in the Bryn Mawr
College course "Androids: Design and Practice." Two of the students from that course
have been awarded summer interships to continue the development of Pyro.
proporsed curriculum developed for this grant and described in the next section is based
on the material covered in this Androids course. Due to its past success, we plan to adapt
these materials for use in other schools.
4.3 Curriculum: Project-oriented modules
The curriculum has been divided into a collection of largely independent modules.
These modules could be used in the order given to outline a single course, or could be
used piecemeal to supplement other courses. The first two modules, which introduce the
Pyro software and basic robot control concepts, should be used prior to any other module.
Each module will have related exercises, reading materials, and software support.
Module 1: Introduction
This module provides an overview to using the Pyro system. Topics include:
starting up the software, connecting to a simulator, connecting to a robot, using
existing robot controllers, and learning the basics of the python programming
language. An on-line Pyro tutorial, developed for the prototype at Bryn Mawr,
can be found at http://emergent.brynmawr.edu/wiki/index.cgi/PyroTutorial.
Module 2: Reactive Control
This module introduces the student to the most simple robot controllers, starting
with Braitenberg vehicles [Braitenberg 1984] which connect motor responses
directly to sensor inputs. Topics include: understanding sensor responses (light,
infrared, sonar, and bump), understanding actuator behavior (differential drive
and gripper), recognizing the problem of noise in the real world, and learning to
tightly couple sensors and actuators for effective real-time control (see, for
example, [Flynn and Brooks 1989]).
Module 3: Behavior-Based Control
This module discusses the idea of behavior-based robotics control [Arkin 1998].
Two main methodologies are explored: subsumption architecture [Brooks 1986],
and a more general approach using fuzzy logic [Saffiotti, Ruspini, and Konolige
1993]. Topics include: behavior design, multi-tasking, motor and perceptual
schema, fuzzy logic, finite state autonoma, and creating behaviors for obstacle
avoidance, picking up trash, and going to specific locations.
Module 4: Vision
This module explores visual processing for mobile robots. The main focus is
integrating vision as a sensor in robotics tasks (see, for example, [Konolige 1997]
and [Briggs et al 2000]). Topics include: vision algorithms (edge detection, blob
detection, filters, convolution, optic flow, color histograms), and using vision
algorithms to locate an object by color or by shape, detect motion, track motion,
Module 5: Planning and Reasoning
This module will focus on the deliberative aspects of mobile robotics. Graph
search methodologies and logic form the foundation of this module [Stentz 1994,
Stentz 1995, Konolige 2000].
Topics include: first-order logic, state-space
diagrams, and various search methods such as A*.
Module 6: Learning
This module will explore robot adaptation. Two major paradigms are explored:
neural networks and evolutionary computation [Meeden 1996, Meeden and
Topics include: designing appropriate tasks, neural networks
architectures and learning methods, genetic algorithms, combining neural
networks and genetic algorithms, and adapting solutions to tasks that were
Module 7: Mapping and Localization
This module explores issues in creating and following topological and spatial
maps [Martin and Moravec 1996, Gutmann et al 1998, Konolige and Chou 1999,
Gutmann and Konolige 2000, Thrun 2002]. Topics include: building a map,
following a map, localization, occupancy grids, and probabilistic states.
Module 8: Multi-Agent Robotics
This module will explore coordination and communication issues in multi-agent
robotics. At least two robots will be required to implement this module outside of
the simulator. Topics include: emulating behaviors of groups of animals, building
a shared map of a space, coordinated behavior to solve problems that a single
robot could not accomplish, and communication methods. ([Balch and Parker
2002] contains a large number of relevant papers in this field and [Yanco 1994]
specifically addresses communication issues.)
4.4 Pilot Schools
In Year 1, the modules will be tested at a variety of school types:
Bloomsburg University is a four-year co-educational public university which has
graduate degree programs (but not in Computer Science). It has approximately
7,000 undergraduates and 700 graduate students. There are 7-8 FTE committed to
CS classes in the Department of Mathematics, Computer Science and Statistics.
Bowdoin is a private liberal arts institution with approximately 1,600
undergraduate students. The Computer Science Department graduates about 15
majors or minors each year. The department has 4 tenure track faculty, but are
staffed for leaves, equating to 3 FTEs.
Bryn Mawr is a private women’s college with approximately 1,300
undergraduates. (Bryn Mawr has a co-ed graduate program with approximately
400 graduate students, but there is no graduate computer science program.) The
Computer Science Department graduates an average of 3 majors and 2 minors
each year. The department has recently increased to 3 FTEs and expects to
increase its graduations to 8-12 majors and 4-6 minors each year soon.
Stanford is a private university with 6,400 undergraduates and 7,500 graduate
students. Stanford offers BS, MS, and PhD degrees in Computer Science. The
Computer Science Department graduates 100-150 undergraduates, 100-150
master’s students, and 5-10 doctoral students each year. The department has
approximately 41 faculty members.
Swarthmore is a private college of liberal arts and engineering with approximately
1,375 undergraduate students. The Computer Science Department graduates an
average of 25 majors and minors each year. The department has 4 tenure track
faculty with about 5 FTEs counting adjuncts.
UMass Lowell, one of five campuses of the University of Massachusetts, has
6,000 undergraduates, 4,000 continuing education students, and 2,500 graduate
students. UMass Lowell offers BS, MS and ScD degrees in Computer Science.
The Computer Science Department graduates about 60 undergraduates, 55
master’s students, and 5 doctoral students on average each year. With recent
hires, the department will have 20 FTEs in Fall 2003 and 21 FTEs in Spring 2003.
The schools represented in our first year pilot are small and large, co-ed and single
sex, public and private, schools that are undergraduate only and schools with graduate
programs, and liberal arts colleges and large universities.
The pilot classes will be targeted to undergraduate students. However, we will also
test the materials in a graduate robotics class at UMass Lowell, starting in Spring 2003.
In Years 2 and 3, we plan to expand our pilot program to include additional schools.
In Year 2, each of the PIs will mentor one or two new schools, adding eight to ten
colleges and universities to the pilot program. We plan to recruit these Year 2 schools in
Year 1 through a wide variety of methods including a chairs mailing, the SIGCSE e-mail
list, the AAAI e-mail list, and personal contact. These schools will be encouraged to
apply for the DUE CCLI A&I grant in November 2003, which will bring them on board
for Summer 2004 to get them ready to teach the course in Fall 2004. We will recruit
again in Year 2 for Year 3, adding another eight to ten schools to our pilot.
We will run three workshops for faculty who will be teaching courses using our
materials. The workshops will be one week long, held during the summer. In these
workshops, we will discuss the curriculum and will teach the faculty how to set up, use,
and program the robots. We want to support faculty without robotics experience in their
use of the robots to engage their students in the material; the workshops will be used to
assist them with getting up to speed.
The first workshop will be held at UMass Lowell in the summer of 2003. The
second and third workshops will be held in conjunction with AAAI’s National
Conference on Artificial Intelligence in order to expose non-robotics faculty to additional
robotics research demonstrated at the annual AAAI Robot Competition and Exhibition.
Holding the workshop at the AAAI conference will also allow us to bring together
students and faculty from the prior year’s pilot classes together with the new faculty
learning about the course for the next year. Finally, we hope to encourage faculty to
pursue projects with their students after the course ends, and we plan to suggest the
AAAI Robot Competition and Exhibition as a possible project path for students.
Workshop materials will be archived on our curriculum web site, providing access
to faculty outside the pilot program who wish to implement the materials. We will also
maintain a discussion group web site where we will encourage faculty from the pilot
program to help support new faculty who are adapting the curriculum for their schools.
Our goal is to include a variety of colleges and universities in our pilot program,
which will allow us to test the materials in many different situations. Senior faculty who
do research in robotics at a large university are likely to have different experiences than
computer science faculty at a small college which has never had a robotics class. To
support the novices, we plan to provide mentoring for the schools in our pilot program.
The five PIs will each support one or two schools during our Year 2 and Year 3 in
the fall semester rollout of the materials. This support will include being available for
phone calls and e-mail to answer questions about the robots and the materials. All PIs
will have at least one robot in the same configuration as the robots purchased by
participating schools with funding from their CCLI-A&I grants. This hardware will
allow us to offer remote support when a faculty member sends us their code or explains
what they are trying to do.
In addition to providing remote support, mentors will travel to their schools at
least once during the semester to meet with the faculty and students. We plan to offer to
act as a guest lecturer and also to give a research talk to expose the students to robotics
research. We also would like to rotate the mentors to other schools, so that each school
gets a visit from two different PIs. This will require each mentor to make at most four
trips in the fall semester (two to their schools, two to other schools).
Mentoring will begin at the workshop and continue throughout the school year.
We will encourage faculty to teach the course during the fall semester, when we have
planned the mentoring activities described above. In the spring semester, mentors will
visit two other schools, providing each school with an additional research talk in robotics.
During the spring, we will encourage the students from the fall semester to work on
projects for the AAAI Robot Competition and Exhibition. (Of course, schools may
choose to teach the course in the spring semester, but this limits project possibilities and
only exposes seniors to one semester of research talks from visiting mentors.)
At the completion of this three year project, we will have created a community
that will be available to informally mentor others who want to adapt the materials for use
in their school. A discussion board on the web site will be used for current discussions
and to archive past discussions.
4.7 Continued Project Work
We will encourage faculty to teach the course in the fall semester, leaving open
the option of independent project work for the students in the spring semester. Students
who are interested in robotics will then have the chance to do larger projects on a
platform with which they are already very familiar. However, unstructured project time
is likely to be unsuccessful for some students. To solve this problem, we are planning to
encourage the students to develop systems for the AAAI Robot Competition and
Exhibition. Each year, the competition lays out several events with specific rules to
guide the development process. More independent students could choose to create a
system for the exhibition, which encourages a wide variety of demonstrations.
To bring together the prior year’s pilot schools with the new batch of schools, we
plan to hold the workshops in Years 2 and 3 at the AAAI’s National Conference on
Artificial Intelligence held in the summer, usually at the end of July.
5 Experience and Capability of the Principal Investigators
The course will be developed by five AI/Robotics researchers at different types of
schools. The team has expertise with developing the robot hardware to be used in the
course, with developing widely distributed courses through NSF’s CCLI program, and
with organizing the AAAI Robot Competition and Exhibition. All teams members have
taught at least one robotics and artificial intelligence course during their academic
Holly Yanco is an Assistant Professor in the Computer Science Department at the
University of Massachusetts Lowell. She has many years teaching experience at MIT,
Wellesley College, Boston College, and ArsDigita University. While a graduate student
at MIT, Holly was invited to teach recitation sections for MIT’s flagship introductory
course (Structure and Interpretation of Computer Programs), which is a position normally
reserved for faculty members. In 1996, she was awarded the Hennie Teaching Award by
the EECS department at MIT. In 2002, she was awarded a Teaching Excellence Award
by UMass Lowell. She has developed course materials for a wide variety of classes,
including Artificial Intelligence, Robotics, Introduction to Computer Science for NonMajors, and C Programming. She is active in outreach activities to middle and high
school students, chairing the Massachusetts Botball committee and running other teacher
workshops. In 1997, she chaired the AAAI Robot Exhibition. In 2001 and 2002, she cochaired the AAAI Robot Competition and Exhibition. Holly’s research addresses the
problems of shared control between robots and people in applications such as assistive
robotics and urban search and rescue.
Douglas Blank is an Assistant Professor in the Math and Computer Science
Department at Bryn Mawr College in Philadelphia. He received a joint Ph.D. from
Indiana University in Cognitive Science and Computer Science in 1997. Doug has
developed and taught a wide range of courses for majors and non-majors in computer
science, including "Androids: Design and Practice" and "Robots Gone Berserk: A Look
at Robots in Film". He has been active in Boosting Engineering, Science, and
Technology (B.E.S.T.), a program designed to get high school students interested in those
topics by building remote controlled robots. He has been involved in the AAAI Robot
Competition, as a participant and organizer. His competition teams have won two
Technical Achievement Awards there. Doug's current research interests include creating
neural network models of analogy-making, and building the new field of developmental
Deepak Kumar is an Associate Professor of Computer Science at Bryn Mawr
College. He received his PhD from The State University on New York at Buffalo based
on work on the design of rational agent architectures in Artificial Intelligence. Since
1993, he has been developing a new computer science program at Bryn Mawr College.
The program has since evolved to 3 FTEs, and is continuing to grow. Several innovative
curricular improvements have been incorporated by him and have also become models
for adaptation at other institutions. In 1995, along with Lisa Meeden of Swarthmore
College, he introduced the use of small robots in the undergraduate curriculum. He has
carried out several outreach activities for Philadelphia-area public schools designed to
introduce school teachers to the latest developments in technology and worked on mentor
programs to encourage students from deprived school districts to enroll at Bryn Mawr for
higher education. His research interests in Artificial Intelligence include intelligent
robotics, cognitive robotics, robot learning, and more recently developmental robotics.
He has served on several international program committees of research and educational
Lisa Meeden is an Associate Professor in the Computer Science Department at
Swarthmore College. In graduate school at Indiana University, she received an award for
Outstanding Associate Instructor. In 2001 at Swarthmore College, she received the
Lindback Award for excellence in teaching. She has served as an instructor at the NSF
sponsored summer faculty enhancement workshops on teaching undergraduate artificial
intelligence in 1995, 1996, and 1997. She has developed course materials for a wide
range of courses, including artificial intelligence, robotics, and several seminars on
computational models as well as introductory CS courses such as data structures, objectoriented programming in Java, imperative programming in C, and functional
programming in Scheme. She has co-led winning Swarthmore student teams at the
AAAI robot competitions in 1999 and 2000.
Her research is currently focused on
creating developmental architectures for adaptive robots.
Kurt Konolige is a Senior Computer Scientist at the Artificial Intelligence Center
of SRI International, a Consulting Professor of Computer Science at Stanford University,
and a Fellow of AAAI. He received his PhD in Computer Science from Stanford
University in 1984. His recent research has concentrated on real time perception and
navigation for mobile robots.
He teaches a course in mobile robotics at Stanford
University, and co-developed the Pioneer and AmigoBot robot line and the Saphira robot
Relevant recent projects where he serves as PI include visual
mapping for the Army Combined Technology Alliance robotics effort; navigation and
perception for the NASA Personal Satellite Assistant project; mapping, environment
reconstruction, and navigation for the DARPA Tactical Mobile Robotics project; a lunar
rover navigation project for Nissan Aerospace; and design of robotics navigation and
perception software for commercial robotic vehicles. He has been an invited lecturer at
universities and institutions in many different countries, and is or has been on the
editorial board of various academic publications, including Fundamenta Informaticae,
Journal of Applied Non-Classical Logics, International Journal of Applied Intelligence,
Artificial Intelligence Journal, and the Journal of Artificial Intelligence Research. He has
authored over 100 scientific publications, including 3 books and Best Papers at the 1995
IJCAI conference and the 1998 IROS conference. He is a co-founder of ActivMedia
In addition to their individual experience, several of the team members have
collaborated with each other before. Lisa Meeden and Deepak Kumar collaborated in the
design of the robot lab course described in Section 2. Kurt Konolige and Holly Yanco
are both consulting on a project to design a robotic wheelchair with ActivMedia. All of
the PIs have worked together for the AAAI Robot Competition and Exhibition.
6 Evaluation Plan
We plan to use several evaluation methods to inform the revision of our modules
and study the impact of our curriculum, as described in [Frechtling 2002]. During the
initial years of the curriculum development, formative evaluations of the modules from
the perspective of the faculty and students at the pilot schools will be conducted. The
formative evaluations of the curriculum will be used to update the materials throughout
the three year grant period. Additionally, we will track students throughout the proposal
period to create a summative evaluation to measure the the impact of the course upon the
student’s performance in subsequent computer science classes and career choice postgraduation. A summative evaluation of the faculty will be performed to measure the
number of faculty who continue to use course modules, the software and the hardware
beyond their participation in the initial pilot semester.
There are several opportunities to track the faculty members who participate in
the pilot program each year. The pre-workshop questionnaire will ask the faculty why
they are participating in the program, what they expect to learn from the workshop, how
they expect the course might change their teaching style, and what prior teaching
experience they have had with a variety of courses from programming to robotics. The
post-workshop study will focus on how the workshop was taught, asking questions that
will guide the redesign of the workshop for later years. Before the course starts, we will
ask the faculty members which modules they plan to use during the school year and in
what course, how much preparation time they have invested in the course so far, what
additional materials they might have read (if any), planned course hours, and how many
students have preregistered for the course. During the course, we will ask the faculty
members to fill out post-module surveys immediately after they complete each module,
in addition to cataloging all e-mail contact with questions and comments about the
materials. A sample post-module faculty survey is shown in Appendix B. At the end of
the course, the faculty will be asked to reflect upon the past semester, answering
questions such as the number of students completing the course, how they taught the
course (number of exams, which modules were actually used, how many class meetings
there were each week, the length of lab periods (if any)), and their thoughts on how the
students perceived the course.
Students will be surveyed at the start of the course, immediately after each
module, and at the end of the course. So that we may track student comments across the
semester, each student will be assigned a number to put on their surveys. Pre-course
surveys will ask for the students’ prior programming experience and other CS courses
they have taken, what they expect from the course, and ask if they have any robotics
experience. A sample post-module student survey is shown in Appendix B; the survey
would be adapted for the specific contents of the module. Finally, a post-course survey
will be used to ask the students their opinions on the course and how to change it. In this
survey, we will attempt to separate the students’ feelings about the faculty member from
their feelings about the course materials.
With the faculty, we can track problems with the materials by cataloging their emails. However, we will not have this direct interaction with the students. To provide us
with a more complete tracking of the student experiences with the materials, we plan to
encourage the faculty to have their students maintain a course notebook, as suggested in
[Scherz and Polak 1999]. The student notebook will allow completed materials to be
collected, while providing a place for students to record their thoughts about the course.
We will ask for student notebooks to be photocopied and sent to us for evaluation.
Faculty could also choose to maintain a notebook, which would keep track of their
lecture and lab preparations, other readings, and intra-module thoughts about the course.
We will also create two web forums: one for the faculty teaching the course at
different schools and one for the students at the different schools to talk amongst
themselves. Both will be archived for study. (In the web forum, people will be able to
choose a nickname, bringing them anonymity.) UMass Lowell already has a server
running web forums, which we can use for this purpose.
After the course has ended, we plan to ask the faculty to help us track the students
in a longitudinal study to determine if the project course has an effect upon their grades in
later classes or their career choice. We have hypothesized that a project class which
engages students will motivate them in other computer science courses; anecdotally, we
have seen students become engaged by a robotics class, resulting in great improvement in
their overall school performance. By measuring student grades in computer science preclass and post-class, we will be able to judge if our class has engaged students
sufficiently to improve their performance in later classes. Additionally, we have seen
that students who participate in robotics projects, particularly projects outside of class
such as the AAAI Robot Competition and Exhibition, tend to attend graduate school at a
greater rate than those who do not.
For example, at Swarthmore College, of the 18
students who have participated in the AAAI Robot Competition since 1997, eight
students have gone on to do graduate work in artificial intelligence or robotics at
Carnegie Mellon University, Georgia Institute of Technology, Purdue University,
University of Edinburgh, University of Massachusetts at Amherst, and University of
Michigan. A longitudinal study of students as they graduate will allow us to test this
hypothesis; the control group will be the other graduating computer science majors in the
school who did not take the course. We recognize that a self-selected group of students
may take the robotics course, so we will also use historical data of graduate school
attendance from each school, if it exists.
Summative evaluations will be done to answer the questions of how the course
impacts the grades and career paths of the students and if the modules are reused by the
faculty in later semesters.
7 Dissemination of Results
We have several dissemination methods planned for our curriculum, including
pilot schools, a web site with materials and discussion boards, and publishing a textbook.
Throughout the project period, the materials will be tested in undergraduate
classes at the home institutions of the PIs (Bryn Mawr, Stanford, Swarthmore and UMass
Lowell). Additionally, the materials will be tested with graduate students at UMass
We will use pilot schools in all three years for testing the materials. We have
tried to select a variety of schools in the home institutions of the PIs and our two Year 1
pilot schools (Bowdoin and Bloomsburg University). In Years 2 and 3, we will recruit an
additional eight to ten schools, again attempting to reach a broad variety of schools so
that the materials may be tested in many different situations.
Materials developed will be distributed via a web site to be hosted at UMass
Lowell. We have budgeted funds to pay an undergraduate student to design and maintain
our web site during the project period, allowing us to ensure that it will be kept up-to-date
with curriculum and software updates. This web site will also host discussion groups for
the faculty and students participating in our pilot program; UMass Lowell already has
software that will enable these discussion boards. The discussions will be archived for
adaption into FAQs.
After the project period ends, the discussion boards will remain on the site to offer
informal mentoring between faculty who have already adapted the class and faculty who
wish to adapt the class for their institution. All materials will remain on the web site for
We are also investigating the possibility of publishing a textbook that use the
developed materials as its basis. There is currently no textbook in the market that teaches
project-based robotics for high level platforms. We have already had conversations with
acquisition editors at Prentice Hall and MIT Press, both of whom were interested in
discussing the matter further. We will have a draft of this text completed between Years
2 and 3.
We plan to write papers about the curriculum, its development and the evaluation
studies for conferences such as SIGSCE and will also look to publish in education related
Year 1 (1/1/2003 – 12/31/2003):
Initial development of course: lecture notes, assignments, software,
projects, robot set up guide
Initial course modules to be tested at UMass Lowell
Recruit schools for Year 2
Summer 2003: Full development of course modules
Workshop for professors from pilot schools to be held at UMass Lowell
Course to be taught at non-PI pilot schools (Bowdoin and Bloomsburg)
and co-PI schools
Year 2 pilot schools apply for CCLI-A&I grants
Throughout year: Evaluation of assessment materials used to redesign course materials
Year 2 (1/1/2004 – 12/31/2004):
Full course to be tested at UMass Lowell
Independent project work at pilot schools
Recruit schools for Year 3
Summer 2004: Evaluation of assessment materials used to redesign course materials
Lecture notes from modules refined into more of a textbook format
Workshop at AAAI-2004 for eight to ten Year 2 pilot schools
Year 1 pilot schools bring robots to AAAI Robot Competition and
Course to be taught at ten Year 2 pilot schools and schools from Year 1
Year 3 pilot schools apply for CCLI-A&I grants
Year 3 (1/1/2005 – 12/31/2005):
Full course to be tested at UMass Lowell
Independent project work at participating schools
Summer 2005: Evaluation of assessment materials used to redesign course materials
Workshop at AAAI-2005 for eight to ten Year 3 pilot schools
Year 1 and 2 pilot schools bring project work to AAAI
Competition and Exhibition
Course to be taught at ten Year 3 pilot schools and schools from Years 1
Summative longitudinal studies
Publication of textbook
We are proposing the full development of an innovative curriculum for the next
generation robotics laboratory, which has already been prototyped at Bryn Mawr. By
moving to research robots from Handyboards and Lego, we will be teaching technology
that is currently used in robotics research to undergraduates. This will enable students to
participate in research projects, giving them an out of class educational experience.
We plan to recruit our pilot schools from a diverse set of institutions, with two
pilot schools in Year 1, eight to ten more schools in Year 2, and eight to ten additional
schools in Year 3. The course will also be taught at the four institutions represented by
We plan to make multiple assessments of the effectiveness of the materials at
different types of institutions serving students with diverse backgrounds and career goals.
Longitudinal studies will be performed to assess the long-term affects of the curriculum
on the students and faculty.
We will disseminate information about the developed materials through a web
site, conference papers and presentations, and publication of a textbook. There will be
self-sustaining national distribution through the web site archives and discussion groups,
as well as the distribution of the textbook by its publisher.