Home Brew Robotics Talk slides |

Housekeeping

• Using my course materials as a place to write my presentation!
• Yes, I will remove this before the class starts
• Forgive typos!
• Ok with me to record this
• Happy to be interrupted with questions at any time

About Me

• Computer Science B.S. from Brandeis University
• Actually was pre-med but switched to computers early
• Back in those days (80’s) getting a programming job wsa pretty easy

Worked

• DEC (Digital)
• long-forgotten-database-company “Software House” (before relational was a thing!)
• Lotus Development (Jazz, Improv, Notes, ccMail)
• Founded eRoom Technology
• Consuted in tech strategy
• Brandeis (around 10 years ago)

What I know and don’t know

• “Software Engineer” not an academic
• Although I’ve dabbled in research and have a handful of published papers
• I’ve been a programmer for xx years! Many.
• Age does change what you are good at and less good at

How I got into teaching

• Around 2010 I was done with consulting
• Always liked explaining
• Had built up a massive knowledge base in my head
• How to share it? Sure, become a college professor!
• How hard can it be?
• No PhD. In my 50’s.
• Hard to break into academia

Brandeis

• Multiple universities approached
• Developed a few different ideas for courses
• Initially co-taught a summer course. $0
• Eventually full time “Professor of the Practice”
• Yes, my alma mater but that was not the deciding factor
• Brandeis, traditionally is a liberal arts college

Computer Science at Brandeis

• My one data point so YMMV
• I was actually in the very first Comp Sci cohort at Brandeis (1976)
• Relatively small department, ~15 faculty
• About 300-400 Computer Science majors at any moment in time
• Traditionally it was focused more on theory and research
• Lately bringing in more “applied” people but still a minority

Computer Science Curriculum

• Courses offered this semester
• Bretty broad coverage of the “basics” of computer science (a few gaps closed recently)
• Very strong representation of AI, Machine Learning, Neural Nets, Computational Linguistics
• Almost all the courses have a significant “programmning” component
• Java, Python and then various other languages
• Until the Robotics course there was no course that involved actual hardware

My Courses

• It does vary depending on the semester
• Electives
  • Software Entrepreneurship (this semesster)[http://cosi102r.s3-website-us-east-1.amazonaws.com]
  • Autonomous Robotics (this semester)[http://cosi119r.s3-website-us-west-2.amazonaws.com]
  • Software Engineering Capstone (last year)[http://cosi166r.s3-website-us-east-1.amazonaws.com]
  • Software Engineering for Scalability (last year)[http://cosi105b.s3-website-us-west-2.amazonaws.com]
• Required Courses
  • Introduction to Java Programming
  • Operating Systems

Students

• I teach both undergraduate and graduate students
• There’s no rhyme or reason
• Most of my courses are elective so I have to market them and make them engaging
• Students are very intelligent and fearless
• But, they focus far too much on grades and GPAs
• They are willing to try whatever activity you invent
• But of course they are inexperienced
• You need to be careful about assuming what you would consider obvious or common sense
• Also you are often the first one to tell them an oldie-but-goodie joke or war story

Classroom Facility

• Classrooms come in all shapes and sizes
• You have almost no control over what room or what time your class is going to be held
• All classes have projectors and computer interfaces but also black boards (or white boards)
• Problem is, can the board be seen from the back of the class?
• Instructors can decide whether laptops are open or closed, whether attendance is taken, whether students can attend remotely.

Student expectations

• You need to keep them “entertained”
• Just like some of you are on your phones right now
• They almost always are there because they want to learn
• They have a lot of demands on their time
• So they are always triaging their time
• They want to know “why” they are being given an assignment or other

The genesis of the Robotics Course

• I wanted to create something that would give students a taste of a real world large software project
• Idea: Campus Rover
• I thought it was a “solved” problem
• As usual I didn’t know what I didn’t know (still don’t)
• That might be a strength :)

The Robotics course in context

• What is it like to be the only Robotics course
• It too much to fit into one course
• None of my students have any electronics engineering knowledge
• Initially had no special budget for this
• Used my tiny personal faculty budget
• Initially a “special” independent study with 6 students

General: Designing a course

• Brandeis gives us a lot of freedom
• 13 Week semesters, either 3 50 minute or two 80 minute classes per week
• Number of students expected will affect the design (scaling)
• Assessment and grading can be fraught
• Many alternative grading schemes are discussed (e.g. specifications grading)
• how to design a course, learning objectives, homework assignments, assessment and grading

Pick your poison

• Again, I didn’t realize how complicated ROS is
• “It’s just another API, how hard can it be”?
• Actually it’s taken years (as you know)
• Started with RObotis TB3. An excellent choice
• Still not 100% sure that ROS was the right choice
• WHy ROS1 and not ROS2?

About the course

• It had to be hands on
• It had to use python
• It has to work with relatively cheap robots
• Initially built it around the chapters in the PRR book
• After looking at every single book about ROS and Robots I could find
• Brandeis has a subscription to all OReilly Publishing books

Building blocks

• Learning objectices (“skills”)
• Homework assignments
• Grading rubrics
• Schedule of lectures

ROS in practice

• Running ROS is a nightmare
• Most students are not running Linux
• And their computers are in widely varying conditions
• Usually the first weeks of the course were spent getting students running
• Then we built our cluster

About the Cluster

• Kubernetes - K3S
• Docker Images
• Two pods per student, one for Linux Desktop and one for VS Code
• Docker Images
• Setup created by an excellent TA a few years ago
• I had to learn it!

About the lab

• Safety
• LIPO batteries
• Capacity of the lab
• Space is at a premium so I’ve been in many different spaces

About the “Branbot”

• Large wheels: Wanted it to be able to go outside (remember original vision)
• Cheaper: Taking advantage of maker lab and free labor

Tradeoff: Build your own robot?

• Learning: Turtlebot3 is manufactured professionally
• I have to keep a “fleet” of about 10 robots running for students
• We have gotten better at careful robust construction but it’s still not as good

Team projects

• Benefits
• Risks
• Choosing team members
• Team dynamics
• Projects need to be interesting and engaging
• But plausible and doable.

Evolution

• Over the first few years the evolution matched my learning curve
• We now have home grown robots too (why)
• Class sizes have steadily grown (reaching or passed a reasonable limit)

Continuity

• Tried really hard to have each cohort build on the last one
• Worked well when we thought we were jointly building “campus rover”
• Now students contribute to Lab Notebook

Lab Notebook

• Lab Notebook
• Useful for building “institutional memory”
• Useful for assessment
• Problems:
  • Students may put in invalid info
  • Or information becomes obsolete
  • Or info is valid but there is a much better way to do it

What about AI?

• Has upset many academic applecarts
• Everyone is deciding what their AI policy will be

Self Assessment: What can be improved

• Student wasted time trouble shooting “dumb stuff
• Too much packed into one course
• Because of that some of the learning is superficial

Self Assessment: What I am happy with

• I am proud of the course
• Students work very hard and learn a lot of the hard knocks of engineering
• They certainly learn the basics of how autonomous robots work
• Only a few of them go out for Robotics jobs (which is totally fine)