Address feedback from prior students of Arend's course

Main complaint points:

The workload has consistently been way too much for 4 ECs. Students report having to spend double, triple, quadruple the time typical for this # of ECs. This is by far the most vocal and largest complaint.
- Many homeworks required the students to redo prior work multiple time (e.g. "compare you answers to last weeks methods")
- Peer review takes an extra 2-3 hours to do properly on top of the super length homeworks. You have to do the review and review the review.
- Too many assignments each week: three deadlines per week
- Since course is obligatory, they feel compelled to work more on it
The exams required the exact numerical answers to get any credit at all for a problem. This is very difficult for the complexity of the problems and that they require programming to solve makes it very hard in the allotted time.
Peer review
- Students are hesitant to be critical of others because they don't want their grade lowered in retaliation.

What students like:

They like the book as a resource and the examples in the book.
Peer review let you see how other students tried solving the problem and gave good ideas for your next attempts.
Symbolic computations were new and they liked learning it.
Like the lecture style and teaching style in lectures.
Students like what they learn and come away with from the class.
Nice that you don't have to pass everything to pass the whole course
Like the option to drop some assignment grades

Recommendations:

Let students work, at least, in pairs on the homework assignments and submit a single result for the pair of students. Allow for individual submissions also for those that prefer it.
An example problem solution library would be helpful.
Some programming introductory material would help (week 0 assignment)

Things that have been changed:

Workload should match ECs:

Students can work in pairs on the homeworks and submit one assignment as a pair.
Homeworks should be designed such that they don't require redoing prior problems.
Some of the topics will be reduced in scope: impact, integration accuracy, newton-euler equations.
No peer review, homeworks will be autograded.
Two homework grades will be dropped.
Programming will be explicitly taught along the way with an opening session on Jupyter, Python, SymPy, and later sessions on NumPy, SciPy, matplotlib, pythreejs.
The exam will not rely solely on exact numerical answers to problems. We will use otehr problem types: true-false, multiple choice, and close enough numerics. Problems will be less elaborate.
New open access book will be created along with the first offering of the class.

moorepants / me41055