The Development and Validation of a Low Cost, Flexible, Open Source Robot for Use as a Teaching and Research Tool Across the Educational Spectrum

Development and Validation of a Low Cost, Flexible, Open Source Robot for Use as a Teaching and Research Tool Across the Educational SpectrumDissertation DefenseBy Abraham L. HowellThursday, March 29, 20121Committee MembersRoy T.R. McGrann, Chair - ME Dept.

Bruce T. Murray, Member ME Dept.

Richard R. Culver, Member - ME Dept.

Richard R. Eckert, Member - CS Dept.

Harold W. Lewis III, Member - SSIE Dept.

Patrick P. Madden, Outside Examiner CS Dept.

2Defense OutlineBackground and motivation

Review of engineering education literature

Statement of Research

Developed teaching and robotics research tool, Open-Robot

Initial investigative research results for Open-Robot

Open-Robots use in swarm research

Final research results for Open-Robot

Final conclusions and future work3Background and Motivation Robots as a Teaching Tool

CaptivateMotivateReinforce Concepts & TheoriesReal World ContextExtend Concepts & TheoriesSTEM Careers4Background and Motivation Robots as a Research Tool for Swarm Robotics

Validate Simulator ResultsTest New Hardware DesignsCostCentralized & DecentralizedTest New AlgorithmsReal World Interactions5Review of Engineering Education LiteratureEducational SpectrumK-12Undergraduate EngineeringRobotics, Programming, EngineeringScience, Technology, Engineering & Mathematics (STEM)Classroom, Laboratory, Problem-Based, and Project-Based Learning,Design Competitions and Early Research InvolvementSummer CampsCapstone Design ProjectsAttitudinal SurveysPre-Post TestsAssessmentDigi-Robot & concept of angle for elementary (no assessment). Using NXTs to teach MS math (pre- & post-test). MS & HS underwater robot design using NXTs (design-based, teacher surveys & pre-/post-test). Nature inspired robotics camp for MS students (pre-/post-questionnaire & (2) pre-/post-questions). HS PBL w/Trash Can robot (no assessment). PBL w/MS math & science students (teacher pre-/post-surveys). Villanova University exposed HS students to swarm robotics research. Teaching HS physics w/PBL & robots (student pre-/post-surveys.

6Statement of ResearchThe primary focus of this dissertation has been to develop and validate a low-cost, flexible, open source robot that can be used as a teaching tool across the educational spectrum and as a research tool in the area of swarm robotics.

Question 1Can a specially developed robot teaching tool positively impact student learning?

Question 2Can a specially developed robot research tool be used to validate swarm robotics algorithms and hardware designs?

Hypothesis 1A symbiotic relationship between engineering educators and swarm robotics researchers can be created when a common robot platform is utilized by both parties.7Developed Teaching and Robotics Research Tool - Open-Robot

8Initial Investigative Research Results for Open-RobotUse in undergraduate bioengineering course.Autonomous Agents course.

Use in undergraduate computer science course.Microcontrollers and Robotics course.

Use in high school programming classes.Java and C++ classes.

Use in high school physics class.Tele-research with Minnesota High School.

9Use in Undergraduate Bioengineering CourseIn the Spring of 2006, an earlier version of Open-Robot, BIObot, was used in BE380B Autonomous Agents as part of a National Science Foundation (NSF) Curriculum and Laboratory Improvement (CCLI)Grant.

All the laboratories were designed to leverage BIObot as a teaching tool.Specially designed Pocket PC software provided students with a means of interacting and controlling BIObot wirelessly.

Course focused on the following: habituation, sensitization, reflexes, classic control theory, fuzzy logic systems, neural networks, genetic algorithms and genetic programming.10Working with the BIObot robots helped me to understand the course concepts better.The software interfaces for the BIObots and the Pocket PC's were easy to work with.Using the BIObots was fun.The BIObots helped me to learn how to use the various methods and techniques learned in class.The concepts and ideas from the lectures were not clear to me until we used them with the BIObots in the lab.The BIObots made me want to learn more so that I could make the robots do more things.It was easy to figure out how to program the BIObots.The BIObots made me appreciate the difference between a classroom example and running a system in the real world.Overall we observe a high-level of agreement with the 8-statements.Statement#1 & 5 are coupled.11Use in High School Programming ClassesAn earlier version of Open-Robot, BIObot, was leveraged in (2) high school programming classes that focused on Java and C++.Robot intervention occurred near the end of the class and just prior to the final project.

A total of (3) lectures and corresponding laboratories provided an introduction to the robots, sensors, serial communication and specially designed C++ and Java class libraries.First lecture and lab focused on introducing robot and its sensors.Second lecture and lab familiarized students with corresponding class library.Third lecture discussed how to create behaviors such as obstacle avoidance and light tracking.

12Using BIObots was fun.It was easy to figure out how to program BIObot.The BIObots made me want to learn more so that I could make the robots do more things.The BIObots made me appreciate the difference between a classroom example and running a system in the real world.After working with BIObot I am more interested in science.The use of BIObots in this class is a good idea.Using BIObots enhanced my interest in this class.After working with BIObot I am more interested in engineering.Overall we observe a high-level of agreement with the 8-statements.Statement#5 & 8 attempt to quantify student perception related to science & engineering.13Use in Undergraduate Computer Science CourseIn the Spring of 2007, an earlier version of Open-Robot, BIObot, was integrated with a microcontrollers and robotics course.

Robot was used as a teaching tool in (3) out of the (12) laboratories.Low-Level Serial I/OWireless Communication with BluetoothHigh-Level Robot Behaviors

Results of this work were published and presented at the 2008 ASEE Annual Conference and Exposition, Pittsburgh, PA 14I enjoyed working with the BIObot robot in this lab.BIObot helped to clarify the concepts associated with this lab.It was easy to interface BIObot with the QwikFlash microcontroller board.BIObot helped me to better understand real-world hardware/software interaction.Working with BIObot increased my interest in this lab.BIObot helped me to better understand serial communication.I recommend using BIObot in this lab for future course offerings.I would like to work with BIObot again.A low-level of agreement on statement#1,2 & 3 could be related to technical difficulties that occurred in lab. Open-Robot baud rate set at 115,200 bps.It appears that the technical difficulties helped students to better understand real-world hardware/SW issues.Even though students had to struggle in this lab they still had a high-level of agreement with statement# 5,6,7 & 8.15I enjoyed working with the BIObot robot in this lab.BIObot helped to clarify the concepts associated with this lab.Connecting BIObot with a desktop computer was not difficult.BIObot helped me to better understand wireless robot control.Working with BIObot increased my interest in this lab.Programming BIObot was not difficult and it helped me to better understand the issues associated with wireless control.I recommend using BIObot in this lab for future course offerings.After working with BIObot, I am more interested in learning about wireless control.Overall we observe a high-level of student agreement with all 8-statements.16I enjoyed working with the BIObot robot in this lab.BIObot helped to clarify the concepts associated with this lab.It was easy to program BIObot for obstacle avoidance.Programming BIObot for two competing behaviors (obstacle avoidance and light tracking) is difficult.Working with BIObot increased my interest in this lab.BIObot helped me to better understand how robots sense objects and navigate in unknown environments.I recommend using BIObot in this lab for future course offerings. BIObot helped me to better understand how to program robots for real-world applications.Overall we observe a high-level of student agreement with all 8-statements.A low-level of agreement with statement#4 means that students did not perceive difficulty in programming for 2-competing behaviors.17Use in High School Physics Class Long Distance Educational ResearchIn 2009 a total of (10) Unassembled Open-Robot kits were leveraged by a Minnesota high school.

Robots were used as part of a project-based learning opportunity for a physics teachers class.

A total of (30) students worked in small groups to perform all the required soldering and mechanical assembly.Also had to perform initial electrical debug.

In order to satisfy the final deliverable each group had to develop a program that controlled their Open-Robot with a behavior of their own design.18S1: Ability to identify a capacitor, resistor, and voltage regulator.S2: Knowledge and understanding of common electronic components i.e. capacitor, resistor, and voltage regulator.

S3: Ability to solder through-hole components.S4: Knowledge and understanding of robot sensors.S5: Ability to assemble a programmable robot.S6: Knowledge of robot programming.

Overall we observe a significant change in student perception regarding specific skills, abilities and knowledge.19

I enjoyed assembling OPEN-ROBOTs circuit boards.I enjoyed assembling OPEN-ROBOTs mechanical components.I did not know how to solder until working with OPEN-ROBOT.It was fun to work with a real-world robot like OPEN-ROBOT.Working with OPEN-ROBOT has increased my interest in this class.I want to learn how to program OPEN-ROBOT.I recommend using OPEN-ROBOT in future classes.I would like to work with OPEN-ROBOT again.Overall we observe a high-level of agreement with all 8-statements except for statement#3.A low-level of agreement with statement#3 simply means that a fair amount of students already had some level of soldering experience.20

I enjoyed assembling OPEN-ROBOTs circuit boards.I enjoyed assembling OPEN-ROBOTs mechanical components.I did not know how to solder until working with OPEN-ROBOT.It was fun to work with a real-world robot like OPEN-ROBOT.Working with OPEN-ROBOT has increased my interest in this class.I want to learn how to program OPEN-ROBOT.I recommend using OPEN-ROBOT in future classes.I would like to work with OPEN-ROBOT again.It is interesting to see that the majority of girls agreed with statement#1 & 2.From statement#3 it is apparent that most of the girls did not have previous experience soldering.Good to observe strong agreement with statement#5 & 7.

21Use in High School Physics Class Comments from StudentsAwesome

This was probably the best part of my senior year. Thanks for the opportunity.

It was a great experience!!!

Fun. Robot = learning physics.

This was the highlight of my senior year in high school.

It was a good experience.These student statements speak for themselves!22Initial Research Results Issues with Assessment MethodologyQuestion 1: Can a specially developed robot teaching tool positively impact student learning?Based upon the initial results we have gained additional insight and say yes this appears to be possible!

However, these results only reveal that students overall perceived an increase in knowledge, skills/abilities and interest in the subject matter under investigation.

We must directly quantify changes in student knowledge and correlate this to working with the teaching tool, so that this work can be deemed a rigorous validation.How do we achieve this goal?23Open-Robots Use in Swarm ResearchUsing RFID and Open-Robot to evolve foraging behavior.

Using ZigBee to control a swarm of Open-Robots.

Cultural Transmission in a swarm of Open-Robots.24Using RFID and Open-Robot to Evolve Foraging BehaviorAs part of the Swarm Robotics Research component, Open-Robot was used in an experiment to evolve foraging behavior.

RFID tags were loaded with virtual food and then embedded in a 4x8 foot environment.

Custom C# software was developed and leveraged Genetic Programming as a means of evolving foraging behavior for Open-Robot.

Results of this work were published and presented at the 2006 Genetic and Evolutionary Computation Conference, which was held in Seattle, WA25Using ZigBee to Control a Swarm of Open-RobotsA total of (3) Open-Robots were outfitted with a custom-designed ZigBee circuit board.ZigBee is a low-cost, low-power wireless alternative to WiFi 802.11.Allows for one-to-one and one-to-many network communication.

A custom centralized software controller was developed and used to provide foraging behavior to all (3) Open-Robots.Demonstrated how ZigBee could be used to wirelessly control a swarm of robots.

Results of this work were published and presented at the 4th International Conference on Cybernetics and Information Technologies, Systems and Applications (CITSA 2007)

26Cultural Transmission in a Swarm of Open-RobotsA total of (12) Open-Robots were used as part of a graduate students masters degree thesis work in BUs Bioengineering department.

This work focused on developing a novel, decentralized control technique for swarms of robots.RFID tags were leveraged as a mechanism for robots to transfer their respective behaviors indirectly to other robots in the swarm.Robots searched the environment using different motion behaviors.

Results of this work were presented and published at the 2nd IEEE Symposium on Artificial Life (2009), which was held in Nashville, TN.27Final Research Results for Open-RobotAssessment of student performance.Specially developed pre/post/post testing methodology.

Course layout.Learning modules and objectives.

Results of Pre/Post testing.

Results of Sentence Stem surveys.

Final conclusions and future work.

28Assessment Methodology Need for a Novel Quantitative MethodThe initial investigative research phase solely leveraged attitudinal surveys in an attempt to solicit student feedback with respect to Open-Robots educational effectiveness in specific learning environments across the educational spectrum.

Attitudinal surveys or student-centric perspectives do not actually quantify the degree to which a students learning is affected, but instead they quantify each students attitude or perception regarding the usage of Open-Robot as a teaching tool.

The above-statements greatly illuminate the need for a novel quantitative methodology that will rigorously measure changes in student knowledge throughout the learning process.How do we achieve this goal?29Specially Developed Assessment Methodology Ideal ModelStudent KnowledgeProgression of Time within Learning ModulePre-Lecture TestPost-Lecture TestPost-Laboratory TestLearning Module30Final Research Results for Open-RobotAssessment of student performance.Specially developed pre/post/post testing methodology.

Course layout.Learning modules and objectives.

Results of Pre/Post testing.

Results of Sentence Stem surveys.

Final conclusions and future work.

31Learning Objectives- Learning Module#7 Serial IODemonstrate the fundamentals of synchronous and asynchronous serial communication.Be able to represent asynchronous and synchronous serial communication data frames in a graphical manner.Be able to calculate baud rates and character transmission rates.

Interface a microcontrollers UART with a desktop or laptop computers RS-232 serial port.

Configure a microcontrollers UART for asynchronous serial communication using assembly level programming.

Develop and debug assembly level code for receiving and sending ASCII readable characters.

32Learning Objectives- Learning Module#8 Embedded CUnderstand the fundamentals of A/D hardware and how it can be used to interface robotic sensors.

Interpret non-linear sensor outputs and correlate to real-world units.

Understand the fundamentals of robotic odometry and how wheel encoders can be used for control feedback.

Develop and debug C code for a mobile robot that receives serial commands, executes commands, and sends back command responses.

Debug C code using an in-circuit programmer/debugger tool.

33Learning Objectives- Learning Module#9 Wireless Robot ControlUnderstand the fundamentals of wireless control in the context of robots.WiFi, ZigBee, Bluetooth and infrared communication.Communication protocols and how to interface with hardware.

Understand the fundamentals of simple control algorithms and develop high-level programs that control a robot across a wireless connection.

Understand the fundamentals of proportional control algorithms and develop high-level programs that control a robot across a wireless connection.

Be able to tune the proportional gain parameters of a given proportional control system.

Be able to select a suitable control algorithm based upon system requirements.

34Learning Objectives- Learning Module#10 Control ArchitecturesUnderstand the fundamentals of reactive, deliberative and subsumption control architectures.Be able to select a viable control architecture for a given problem.Design suitable subsumption control architecture for a given problem.

Develop and debug high-level code that implements subsumption control for a wireless robot.Leverage multithreading or simple timers along with a task coordinator.

35Learning Objectives- Learning Module#11 Fuzzy Inference Systems for Robot ControlUnderstand the fundamentals of fuzzy inference systems (FIS).Define a suitable FIS using a block diagram.Setup and configure a FIS.Define and represent fuzzy relations for linguistic variables.Calculate triangular and trapezoidal membership functions.Evaluate If-Then Rules.Calculate fuzzified and defuzzified outputs.State the difference between Mamdani style and Fuzzy Singleton defuzzification.

Design, implement and tune a FIS that supplies different behaviors that control a wireless robot.

36Final Research Results for Open-RobotAssessment of student performance.Specially developed pre/post/post testing methodology.

Course layout.Learning modules and objectives.

Results of Pre/Post testing.

Results of Sentence Stem surveys.

Final conclusions and future work.

37Results of Pre/Post/Post TestingEach student was assigned a number, so that their corresponding pre-lecture, post-lecture and post lab results could be analyzed.This provides a more in-depth view into how a specific students knowledge evolved throughout the learning process.

Unfortunately there were only a total of 9 students that participated in this study.In order to deem the results from each learning module statistically significant they were first analyzed for normality using Anderson-Darling.A Paired T-Test was used to determine whether or not the post-lecture and post-lab score changes were significant.38Results of Pre/Post/Post Testing Testing for Normality using Anderson-Darling TestH0: Data follows a Normal DistributionH1: Data does not follow a Normal Distribution

If P-Value > .05 & AD < Critical Value (CV), then accept null hypothesis and conclude that data follows a normal distribution.

If P-Value .05 & AD > Critical Value (CV), then reject null hypothesis and conclude that data follows a non-normal distribution.

39Results of Pre/Post/Post Testing Results of Paired T-TestH0: Test A to Test B Differences are not significantH1: Test A to Test B Differences are significant

If P-Value < .05, then reject null hypothesis.If P-Value .05, then accept null hypothesis.

Paired T-Test reveals that Post-Lecture and Post-Lab changes are statistically significant.

40

Learning Module#7 Serial I/OLearning Module#8 Introduction to Open-Robot & Embedded CLearning Module#9 Wireless Robot ControlLearning Module#10 Behavior Control ArchitecturesLearning Module#11 Fuzzy Logic Inference Systems for Robot ControlOverall students displayed low levels of knowledge with regard to the pre-lecture tests for LM#7, 8, 10 & 11.Students display a significant increase in knowledge on the post-lecture tests for LM#10 & 11.41

Why did Student#3 and 9 regress?Student#3 leveraged an incorrect solution methodology on question#1.Question#1 required an iterative solution process, so that the baud rate error could be minimized, but Student#3 did not use an iterative process.Student#9 answered two parts of question#3 correctly on the pre-lecture test, but only one part on the post-lecture.Question#1 & 2 were answered incorrectly on both the pre-lecture and post-lecture tests.42

Lab Group#Student#s11, 2, 423, 6, 935, 7, 8It is possible that student#6 was responsible for the success of group#2 in the Serial IO laboratory.43

Why did Student#7 stagnate after the Post-Lecture test?Question#2 and 3 were answered in a similar manner on the Post-Lecture and Post-Lab.Had difficulty answering question#1 on all three tests.Question#1 focuses on learning objective#1 & 2, which is related to concepts associated with robotic sensors that output a non-linear, varying voltage and must be interfaced with a microcontrollers analog-to-digital hardware.44

Lab Group#Student#s11, 2, 423, 6, 935, 7, 8After examining the individual tests in detail, it is interesting to find that student#5 and 8 also had difficulty with question#1 even through their respective level of knowledge continued to grow from pre-lecture to post-lecture and post-lecture to post-lab.45

What happened to Student#3 and 7?

A closer look at the tests reveals that Student#3 made a mathematical error on question#3 for the Post-Lab test. The methodology was correct, so in essence student#3 stagnated.

Student#7 simply stagnated after the post-lecture test.46

Pre-Test shows that students knew very little about Subsumption Architecture.

A majority of the students converge upon saturation after the pre-lecture.Is it possible that the depth of penetration into behavior control architectures was too shallow and therefore students quickly learned the material after lecture?47

Pre-Test results show that students knew very little about Fuzzy Logic.

Student#3 did not answer any of the questions on the pre-lecture and only question#3 on the post-lecture and post-lab test.An informal inquiry with student#3 revealed that he/she had been extremely busy with other courses and simply had not been able to devote the appropriate amount of time in or out of class.48

Lab Group#Student#s11, 2, 423, 6, 935, 7, 8The post-lecture and post-lab tests for student#6 show that he/she partially answered question#1, 2, and 3 in a similar manner, but simply did not answer question#4.Question#4 attempted to measure student knowledge relative to the evaluation of If-Then-Rules.49Final Research Results for Open-RobotAssessment of student performance.Specially developed pre/post/post testing methodology.

Course layout.Learning modules and objectives.

Results of Pre/Post testing.

Results of Sentence Stem surveys.

Final conclusions and future work.

50Sentence Stem Survey Results

The main purpose of the sentence stem survey was to collect additional data regarding OPEN-ROBOTs effectiveness as a teaching tool, but in the context of each students own words.The Sentence Stem Survey was administered via Blackboard and at the end of the course.

Two main categories: (1) Students attitude toward OPEN-ROBOT in relation to laboratory and final project usage and (2) Students perception of OPEN-ROBOTs educational effect.51

fuzzy logic control and subsumption architecture, I was able to get hands-on experience.real world problems, mechanical factors complicated things beyond the software levelthat robotics is very complicated, it was clear that there are so many factors that influence the strategy we used to find a solution to our lab problemsfuzzy logic control and subsumption architecture, I was able to get hands-on experience.that robotics is very complicated, it was clear that there are so many factors that influence the strategy we used to find a solution to our lab problems

It is very important for students to have the opportunity to experience real-world type situations and issues because it forces them to extend what they have learned in class, but in a non-sterile sense.52

my understanding of the class, the hands on experience was easier to understand than just lecturecementing the concepts covered in lecture, it gave me a chance to apply what I'd learnedmy understanding of the class, the hands on experience was easier to understand than just lecturecementing the concepts covered in lecture, it gave me a chance to apply what I'd learned

It is good to learn that students perceived OPEN-ROBOT as a catalyst for helping them with programming, concepts and theories from class, and problem solving, but through hands-on experience and application of classroom knowledge in a laboratory setting.53

improved, the robot presented challenges that I have not faced beforeincreased, I was working with new types of problems, outside of just debugging codeimproved, Ive seen that even with a correct algorithm, a system may not work properlyimproved, I had to apply different methods of debugging

During the wireless control laboratory it was noted that most students disregarded the fact that wireless communication can be prone to packet loss and therefore only planned for the Golden Path in their software. This lack of robustness planning caused numerous debugging difficulties for most of the groups and so they were forced to learn the hard way.54

increased, I had to troubleshoot new problems after working with open robotimproved, I now know of other things to look for as the root of the problemincreased, I had to troubleshoot new problems after working with open robotimproved, I had to find bugs in an environment that I was not familiar with

It is not clear as to the exact details of how Open-Robot helped, but it could be related to the fact that Open-Robot is a real-world device that is prone to real-world issues.55

Subsumption, helped, it showed it in a hands on fashiona lot of the material, is where I learned most, just listening doesn't settle in my head but actually doing made things clear to mefuzzy logic, helped, I got to write a fuzzy logic inference system and actually see how it all worked without doing all the lower level workfuzzy logic, helped, I got to write a fuzzy logic inference system and actually see how it all worked without doing all the lower level worka lot of the material, is where I learned most, just listening doesn't settle in my head but actually doing made things clear to me56Qualitative assessment data that is solely based in student perception cant be used to develop rigorous conclusions. However, when perception data is directly paired with quantitative pre- and post-test type data its usefulness is greatly amplified because it can be used to develop a deeper understanding from the students perspective. Sentence completion surveys force students to respond to specially designed statements, but in a minimally constrained yet partially guided manner.57Final Research Results for Open-RobotAssessment of student performance.Specially developed pre/post/post testing methodology.

Course layout.Learning modules and objectives.

Results of Pre/Post testing.

Results of Sentence Stem surveys.

Final conclusions and future work.

58Final ConclusionsA low-cost, flexible, open source robot has been developed and validated as both a viable teaching and swarm robotics research tool.

A novel quantitative assessment methodology was developed, so that Open-Robots educational impact could be measured.

Sentence Stem analysis revealed that students perceived the hands-on nature of Open-Robot to be the most beneficial from a learning perspective.Open-Robots Real-world aspect was perceived to be the next most important.

Future WorkFuture work could involve pursuing the following paths.Investigate the validity of Hypothesis 1. Develop a symbiotic relationship between swarm researchers and educators.

Investigate using Open-Robot as a tool for improving student knowledge, skills and attitudes related to the ABET a-k outcomes.Focus on using Open-Robot as a tool for improving the most important outcomes.

Continue to evolve Open-Robot as a teaching tool.Recently developed a 32-bit version that is compatible with the Arduino series of controllers.WiFi & ZigBee wireless boards developed.Vision & memory boards currently under development.Question and Answer PeriodThank you to everyone involved because this day would not have been possible without you!