Cyber Physical Design and Analysis

The course was not difficult and my experience pretty much matched other students' experience in older semesters. The lectures are not needed to do the assignments and the homeworks (except for one) and projects are straightforward. The TAs were pretty active and helpful -- no complaints in their effort.

My only complaint is that there was one homework where I very much disagreed with the framing of the question -- it was a simple optimization problem but the constraints were not fully explicit (granted the original paper the homework was based on did not either). We had to infer how to further add constraints to obtain a sensible result. However, I was hesitant to do so in fear of assuming too much and ended up taking a grade hit. Some students complained about the grading, like I did and I kinda felt like it was an "office hours" curve (the recorded office hours and public Piazza posts seemed as if TAs were hesitant to provide the answer though). The TAs were nonetheless very helpful, despite seeming as if they were restricted in being able to clarify the constraints as I wanted. All the other assignments were straightforward though.

Like previous reviews mentioned, Homework 5 and Project 3 are a bit more difficult. Homework 5 involved (if you used Windows like I did) using WSL to run a program correctly and answering question related to that and to hardware datasheets. Project 3 involved using AADL but the project was well written and links to many useful textbooks/ resources. In addition to that, office hours were available with a guest speaker very proficient in AADL.

This semester did mark a slight change in the course where instead of just a take home assignment, the final was split into a two portions: a shorter take home assignment and a proctored exam. The take home assignment was pretty much similar to a homework in terms of difficulty and scope. The proctored (closed notes) exam involved reading provided articles and using concepts learned on CPS to discuss it (sort of like the first homework). The exam was not difficult if you understood the basic concepts of CPS, it took only 1.5 - 2hours to write a short essay and since it was the first time we were doing it in the history of the course, the grading was very lenient. The only reason the essay portion of the final was made proctored was to dissuade use of ChatGPT to write the essay.

this course is the biggest waste of time. It has nothing to do with cybersecurity. you will learning nothing in this course. The first 2 homeworks are straightforward and not difficult. The last 2 homeworks are more challenging due to vague write ups. The first 2 projects deals with PID feedback controllers. Project 3 is using AADL, which is useless to learn, to edit files to complete a design model. This course is not difficult material however the write up for the projects are terrible and leave tons of info out. I lost several points due to this and the TAs would not give points back. The TAs run the course and are terrible. They answer questions very vaguely, office hours is a huge waste of time. The grading is poor, they take off a lot of points for minor errors, several students in spring 2023 had to drop this course because of that. The TAs also don't tell you why you miss points. you just receive a grade and is it. I asked for clarification and was told 'This is a graduate course, we cannot give you the answer, you have to watch office hours for clarification.' Overall, don't take the course if you don't have to. For some reason this is required for OMSCY and it should not be.

I expected it to be OK, but I loved it: I had some past experience in IoT and CPSs, and I find the course interesting, very relevant, and useful. Also, the Professors and TAs did a great job as well.

The videos provide a general guide, and, although you could do some projects without them (depending on your previous knowledge), I find them also very good and very useful, I 100% recommend watching them. I played most of them at an x1.25-play speed except with Professor Jerome Hugues.

The math on the videos seems hard sometimes, but do not feel afraid, it is doable, and, when used, they provide you with the right guidance, and the projects and homework are very cool.

The homework and projects are very well-designed and interesting. Take into account that Project 1 might work in the Matlab simulator, but not in Robotarium: my recommendation: finish and test it ASAP, and send P1 to Robotarium much in advance.

Final exam: this term, these were like the given homework or projects, a set of exercises you need to solve. Interesting and doable as well.

CPDA is a very mediocre course. The main shortcoming in my opinion is the uselessness of the lecture videos. I watched ~2 hours total of lecture content the entire semester (1 hour when I first started the course and another 1 hour during the take-home final exam) and I passed with a final grade of 95. The homework are very disconnected from these lectures. Honestly, most of the homeworks by themselves are actually interesting. The Robotarium project and the controls homeworks are stimulating enough, but as stated before there is zero connection between these homeworks and the lecture material. One final gripe I have is for Project 3 with AADL. This project is complicated and totally useless as far as I can tell. The instructions are vague to the point where there's about 100 questions from students on Piazza asking the TAs to interpret each question for this project. AADL itself is an outdated language. The same concepts could be much better taught in a different system environment or just with conceptual questions. Overall, I would recommend this course to any students with the Computational Perception and Robotics specialization because it means you don't have to take Computer Vision (I took CPDA, Comp Photo, and AI for Robots). Especially for students who are worried about a high workload, this is the easier alternative. However, in terms of actual value and learning, this class is very below average.

This course is required for the cyber phsical track in OMSCY. I have no idea what it has to do with cybersecurity. I expect to get an A, however I also feel that I learned nothing. Like many classes, the professor is absent beyond the video recordings. In fall 2022, the head TA apparently quit and left two other TAs holding the bag. One TA was knowledgeable of the topics, the other was not. Neither had good course management skills. Assignments are poorly written, changes are made mid-stream in assignments with no announcements (just buried commments in a pages long piazza post), and sometimes the tools (P3 VM) provided are actually broken and will not support your work. I'm sorry now for anyone who has to take this as a requirement.

If you are considering it as an elective option - just don't.

This was my 8-th course in OMSCS, with some easy and hard courses behind, and I didn't dislike the course at all. Like many others, I considered it a more reasonable option for CP&R specialization than CV in its current state (especially considering I completed CP), and I reckon it was a right choice.

I'll first try to cover the negatives generally expressed in the reviews - I understand most of the background for those, but will present my perspective. For the records, I don't have any real exposure to robotics apart from some previous courses here (like AI4R), but I do have an experience in the simulation of embedded systems, which can also be considered cyber-physical ones (still, mostly high-level exposure on a system level).

• "The lecture material doesn't relate to assignments".

  • Mostly (not fully) true. The lecture material is generally very broad, or covers specific examples not so related to the examples covered in the assignments. The way the lectures are presented can indeed be pretty hard to absorb, and I have to admit I didn't finish watching a couple of those at the end. Some concepts, like math formulas (at least for me) were presented in a way it didn't sink, apart from a very shallow idea. Some moments are recorded using not the most professional presentation skills (e.g., at one point you see a moment of "yeah, we got this episode" fist gesture at the end of the video). However, technically, if you would like to get a broad overview of the system design and ways/directions to think about (which is the main purpose of the course), the lectures hit the mark. The assignments may cover and exercise different examples on most part, but it's worth remembering that learning is the whole point here. Additionally, the final exam had a section of a specific case study, which should have been analyzed WRT the material studied in the course, so I surely don't agree with the claim that the lectures played no part here.

• "HW5 and project 3 are substantial nightmares, disconnected from the course".

  • Like most of the assignments, HW5 is another case/example evaluation and it did carry a degree of ambiguity. Some of the complaints in the reviews here here are surely justified. There were people who got the full score there - I was not among those, and some of the points I lost I can attribute mostly to ambiguity, but partly also to probably not going deep enough into thinking about aspects WRT what was intended. It's true that most of the homework was not really exercising exact material previously studied, rather being very open ended, with point deduction reasons sometimes being hidden in the maze of what lectures were intending to summarize. I think HW5 indeed was the most glaring example in the course where better background articulation/preparation could have been in order. However, I can't say it was fully useless and disconnected.

  • Project 3 and AADL. Indeed, AADL may indeed not be widely used. However, this should not matter much here, as the way I took it, it was an example on how to evaluate, adjust, simulate and analyze system design, and AADL is just a tool to see it through. It was not hard to learn it in a way of what to copy-paste. True, some fighting was there and could probably be avoided with better assignment structure, but there were hints and answers generally provided. The assignment setup including the provided VM was well prepared. Getting it done did require some fiddling with syntax, that's true, and also some basic code structure/inheritance vision being exercised. Yes, someone with no coding experience could struggle to converge fast, but by no means some substantial knowledge or experience was required. I'm not saying it was ideally smooth, but the hooks were there. I had zero experience in AADL before, and the points I lost were not related to that at all.

  • Others have mentioned that the grading was not really "completion based", and I agree.

• "Generally missing teaching objective / background for the material in the course".

  • I did not see it this way. I believe it was possible to see the point of what the instructors were trying to convey through the lectures and exercises. It was the system level design of CPS, what to think of, what to pay attention to, the big picture. Think about the course this way, and it will connect.

So now to other positives. First, the ability to exercise MATLAB (had zero exposure before the course) on some of the assignments and especially projects 1 and 2 - really nice projects, well set up, nicely visualized. Project 1 is the best one where you can exercise a rather easy example of robot planning in real life Robotarium, and project 2 included a competition on solutions (make sure you properly validate the distance etc. constraints there, as otherwise you could loose a lot of unnecessary points, as described in the assignment). Overall, those projects were fun and generally easy.

The rest of the assignments were also fine - not always straightforward, and in some cases, ideas from AI4R and sometimes other courses did help.

The exam was pretty good - take home, two elaborate problems to solve or to analyze. More or less a take home assignment (in the spirit of AI exam format, but shorter and obviously different assignment nature). It also had a few bonus points.

We also had 5% "free" points for Piazza participation.

By all means I didn't get a full score on all of the assignments, but still, I was never far away from a very high A, and again, my relevant background level was medium or lower. The load was rather low - there were a couple of weeks (e.g. during project 3 part 1 submission period of 3 weeks) where I literally did not invest time into the course at all. Try doing that in a course like CP!

The TA support on Piazza was solid - I didn't have any issue, the questions were answered timely and to the point. The atmosphere was good. The TA-s held office hours. The second professor, Dr. Hugues, who was "responsible" for the most "problematic" content of the course, was very much available and actually helpful, e.g. for the project 3.

So all in all, this course might indeed have a different setup and mindset than others (although, that can be said about many courses), but it has definitely left a positive feeling and it was not on the hard side with an open-ended approach.

I didn't like or dislike this class - it was entirely neutral for me. I took this as part of the CPR specialization because I did not want to take CV based on the really negative reviews of that course.

The subject matter of this class was not that great. If you have any engineering degree (or a BS CS), a lot of this will be review from undergrad especially around engineering design processes. I did not watch a single lecture in this class because they were extremely long and did not tie to the homeworks or projects. Each homework/project has its own resources which are usually readings. Project 1 and 2 were pretty interesting and relatively enjoyable. The grading for both of these was fair and mostly automated so you knew what you were getting when you submitted.

Homework 5 was just long - it wasn't hard per se but did take a while to get through. Project 3 was really irritating because you have to use a language called AADL which is entirely useless in the real world. Most of my time spent on Project 3 was debugging AADL setup issues and trying to figure out how to use the software. FYI the software for project 3 does not work on Mac M1 at all. M1 does not have VM support - and the software also does not work well on Intel Macs forcing you to use a VM. I really dislike using VMs because of how slow they are, so I had to fire up an old Windows laptop from college to complete this project which worked well.

There are a couple of things I want to point out though about grading. I felt that the grading in this class especially for Homework 5 and Project 3 was extremely nitpicky and totally based on accuracy and NOT completion. This conflicts with some previous reviews of this class that mentioned that grading was lenient. My Homework 5 report was extremely detailed and I still got a bunch of deductions on it because I was missing minute details the TA was looking for. The grading is totally based on accuracy which is very frustrating because the assignments have very little resources and you are kind of on your own to figure out what the TA's want. They won't really clarify what they are looking for and mention that "as long as you explain your reasoning, you will be fine with grading". This is not necessarily true though because they deduct if the answer does not contain all of the points they are looking for. I am not dinging the TAs because they did try their best to help out throughout the semester. My main point here is to go in optimizing for complete accuracy and not completion when it comes to submitting these reports.

There is a take home, open note final at the end which is fair and pretty interesting to work through. I paired this course with an easy elective and had a very chill semester overall. I probably could have paired this with an equal or harder class and still managed well.

The first half of this course was actually enjoyable and I thought the grading was decent. Then when HW5 hit and project 3, they clearly didn't take any advice from previous semesters and kept the same useless material. The grading was not generous at all on HW5 or project 3 and I tried to add every detail I could find within both those assignments. If you misunderstand their interpretation on a question, -50% for that question. Never before have I put so many hours into a project without any way to fully verify what I was doing was correct since they don't provide any material on the last two assignments (or any of the assignments really but, the first half of the course was more logic based.) This course is a required course for most so, do your best on the first half and brace for the second half.

This course is an introduction to CPS. It can be divided into two parts. The first part covers the control system design. The second part focuses on safety and security analysis. In this course, the most difficult project is using AADL to analyze an UAV system. But don't feel so frustrating with the abstract concepts. It is an open topic.

Pros

• Active Profs and TAs. For Project 3 AADL, Dr. Hugues is active on Piazza every time.
• Generous grading.
• The course videos provide a specific overview of CPS knowledge.

Cons

• Ambiguous lanuage usage in the assignment instructions. You should deduce the intentions
• Easy assignments and projects are not closely related to lectures. They make us feel confused.

Conclusion

It is an interesting but strange course. Nevertheless, the course is qualified for CPS topics. As many reviews mentioned, you may take it as one of the robotics specialization or to balance your work and life. For improvements, I hope assignments can cover formal methods and vertification in the future.