The course was challenging, but fulfilling. Thank you Coursera and University of Pennsylvania for giving this wonderful experience and opportunity that I might not experience in our local community!
The topic was very interesting, and the assignments weren't overly complicated. Overall, the lesson was fun and informative , despite the bugs in the learning tool(especially, the last assignment.)
par Bakhtiar M•
The grader had quite some issues.
par JJ W•
So I'll start with the positives. The material was appropriate and interesting and well presented. CJ Taylor is an enthusiastic lecturer and the material was presented in an enjoyable easy to understand way and having finished the course, I definitely want to learn more about computational motion planning.
The problems I have with this course though are numerous. This is the second part of the Robotics Specialization and compared to the first part, this course was very weak. There was very little lecture material and the course felt thin - as if it were 2 weeks of material stretched over 4. There were many instances where the lectures could have gone in to much more detail and just didn't, I appreciate that you can't cover everything in lectures, but would it have killed you to provide or at least point to some good additional reading resources?
The assessments were the worst part. The quizzes barely tested what I had learnt and could mostly be solved by common sense. What I find shocking is that there were so few questions with few multiple choice answer that you could easily brute-force these quizzes if you really wanted to. Compare this to the Aerial Robotics course where the quizzes took time and forced me to think and understand what was discussed in lecture.
The programming assignments were shockingly bad. They were hard for completely the wrong reasons. I spent most of my time on them not coding the solution, but trying to figure out what was actually wanted and fixing bugs that were in the provided code that we WEREN'T EVEN SUPPOSED TO EDIT. The autograder would never tell you why you were wrong, just "I'm sorry, your solution didn't pass all of our test cases." This meant that finding the solution was based on guesswork rather than considered thought. This was made even worse by the fact that some simulations took a long time to run which made iterating guesses very slow - and doing this on a time limit is just pointlessly stressful.
One of the assignments had the solution already in the source code as the instructors had forgotten to take it out.
The final assignment wouldn't even run out of the box without fixing bugs in the provided code. This would have taken seconds to check had the person who wrote it bothered to check their work beforehand.
Thing is, the tasks provided in principle weren't that hard, they were actually kind of too easy. Dijkstra's algorithm isn't that difficult to implement from scratch, and yet all that was asked of us was to implement a small 10 line for-loop. That said, I appreciate that as a software engineer, I might find this sort of thing much easier than most, but even so I don't feel as though the programming assignments helped me learn anything.
Overall, regardless of how interesting the material was, this course was very shoddily put together. I appreciate this is the first time the course has been run, but this really felt phoned in and unacceptable. I paid money for this course and the quality of it is notably worse than most free MOOCs I have taken. I feel ripped off and I sincerely hope that the next section is better otherwise I doubt I will bother to continue until the end. I thought Penn University was better than this.
par Ravi T S•
Prof. Taylor's lectures were clear and easy to understand. The course taught the basics of planning pretty well.
The teaching staff, unfortunately, was completely unavailable. The solution checkers for the Matlab assignments were very brittle and wasted a lot of students' time. There were several queries about the assignments on the Discussion forums most of which did not elicit any response from the teaching staff. The students ultimately figured it out by spending several laborious hours figuring out the peculiarities of the solution checkers, and by helping each other. This certainly soured my experience with the course.
par Alejandro G•
The contents and quality of the video lectures are great.
The quality of the supplementary materials (like the quizzes and the Matlab code) provided by TAs for the assignments are very poor. The assignment guides are written with very poor grammar. The Matlab code is written using the worst coding and documenting practices. The automatic evaluation system (the system for grade the submissions), doest't provide useful feedback. As people can read in the course discussion forums, students have been asking for months for improvements to the grading systems.
par Toni T•
The assignments were hard, not because of the contents of the courser, but for understanding the context of the method being implemented (what each variable meant and how it was expressed).
Also, there should be a range of submissions that could be right but having higher or lower scores depending of how long it took to find the path, or how long the path was to reach the goal.
par Juan Á F M•
The topic is very important and interesting, but I think that lectures are way very short, and the student is left with some pieces of code that many times don't run in MATLAB 2014 or older.Finally, the graders must be tweaked after reading the forums. I don't know if that is truly the point, but a student may get too much confused with the grader response.
par thiago c•
matlab assignments are horrible, very poorly explained and documented and the submission process does not give the correct resulhalf the time. I had to change submissions that looked perfectly correct (and gave correct results) to something that did not look right just to pass the grader.
The course is good, but the matlab part really needs improvement
par Lieke V•
The assignments in this course take much more time than indicated. Feedback from staff is minimal. If you are not a good programmer, think twice. Nevertheless, the lectures are interesting and well taught. So if you do not want a certicficate, I recommend this.
par Jakob H•
Sorry but the course videos are really too short and the explanations could be better. I would appreciate if in some cases you would go more into mathematical background/detail. The programming assignments and some grading were in my opinion too strict.
par David L•
The assignments would be easy except that the provided code has bugs, and some of the assignment details are wrong, making it very difficult to pass the test cases.
There is also very little information on how the work will be graded.
par Paras S•
>>Videos are not detailed and comprehensive
>>Video lectures are very short
>>Programming excercises are difficult to do complete
>>The professor's teaching style is not engaging
>>Python should be preferred for such courses
Interesting material, and good assignments. However this is no assistance when by TAs if you have questions, which is important since there is no feedback on assignments and hard to know where you went wrong.
The course could be 1-week long instead of 4. The material offered is introductory level. Quizes are too easy, programming assignments still have bugs, and grading procedures cause a lot of confusion.
par Jie Z•
The assignments is not focusing on the contents in the lecture but some minor places. It wastes a lot of time to do meaningless work to pass the test. The autograder is really bad.
par Omkar P•
The Assignments are more of Trial and Error and not very helpful. The Course Material and Content is good. Teaching Staff is nowhere to be found.
par Barak R•
assignments are very poorly explained.
the course is very basic and I am unclear on how to use it for practical applications
par Alessandro P d M•
Many problems and mistakes found in the assignment programs and I felt a lack of staff assistance on helping solving them.
par Nilesh S•
The course is good , but the grader sucks a lot. They should program the grader in a more general way.
par Fengwen S•
Terrible auto-grader without any flexibility. Not covering how to generate configuration space at all.
par Nick L•
This is just a poor sample of a course. Very few material, low quality assignments and grading.
par Qi L•
No much staff in it
par Jason D•
The course is very bad and feels thrown together at the last minute. Learning A* and Dijkstra's algorithm is great however the assignments require you to learn not just the little details but to "discovery" techniques not even mentioned in the course material. In addition, you must have strong matlab programming skills and be familiar with much matlab functionality in order to debug some of the assignments. You must have more knowledge concerning matlab than any of the course material or pointers provides. Meaning that beginners will NOT pass this course. The automatic grader provides no feedback at all except pass or fail. This is unfortunate as it can look like your code is working correctly but, the grader is using some edge cases to grade the code but will not include any information indicating what to look for. This is really atrocious. Although the TA's do occasionally provide answers to questions. The total amount of time TA's spend answering questions is just really poor. Don't expect even well asked questions to be answered at all. In addition, the coded template quality upon which your own code depends is horrible and thrown together. You will spend way too much time analyzing it for clues as to what went wrong. Sadly, enough all of these issues have caught up with me and I was unable to pass assignment 2 part 2. Even, though everything looks like it works and achieves the desired goal and even works with all of my own test cases. The grader is merciless. Perhaps, in the feature more time can be devoted to make this course better and I can spend more time learning how the algorithms and maths work rather than matlab and the automatic grader. At this time I don't feel like my money was well spent on these courses. I don't think I would like to risk another 50 dollars learning matlab and debugging the automatic grader on any of the other courses in this specialization. That is very disappointing as I really am passionate about learning robotics and looked forward to the other courses as well.
par James L•
This is the second of the series, and I had high hopes for this one after the first one was a disappointment.
While the material was on the sparse side, the level of the course was more as advertised. The Matlab assignments were not too difficult, but were relevant enough to the material to be acceptable. Some of the assignment simulations/animations were not working, but I was able to submit the results and get full credit. I was ready to give 2 or 3 stars in this review.
Unfortunately, on the last assignment, the sample code and scripts was buggy (even towards the end of the course after it had been pointed out to the staff), and inconsistently written.
The final project submit script would not work and it was a trial and error troubleshooting to understand why. Some on the discussion boards were able to submit, but other couldn't. Although my assignment would run as I understand the project should, I could not even submit it. Judging from the discussion boards, I'm not the only one with this problem.
Again, the University of Pennsylvania has missed the mark with this class and the robotics series. It is regretful that I must again leave a one star review.
I realize this is the first time this course was offered, so there are some issues that need to be worked out. I suggest for the remainder of the classes in the series, the creators should do a pilot run first before releasing to Coursera.
par Iwan P•
The topic is interesting and the lectures itself were good. The explanations of the algorithms and concepts was clear and easy to understand. However, the amount of material covered is very little. I think there should either be more concepts thought or they should be covered in more depth.
Quizzes: The quizzes were very weak. They didn't really tell you weather you understood the concept or not. If one of the answers was wrong there was no information which one it was. So you have no idea which lecture you should watch again.
Assignments: The assignments were very bad prepared. In some assignments there were bugs in the provided code, we weren't supposed to edit. The assignments itself are rather easy and can be completed quite quickly, if there wasn't the grader. From the grading you got no information what was wrong or which tests passed or failed. Most of the time it wasn't the case that the algorithm failed, instead the output was not as expected by the grader, although it was as described in the assignment. But you never knew. This part has to be improved a lot.
Overall I have to say that I expected much more from University of Pennsylvania, especially after taking the first part of the Specialization which was of good quality.
par Nico W•
This is part 2 in the robotics specialization. The course content is way too short and basic for the price. There are about 77 minutes of lecture in total, i.e. less than 20 minutes per week. The first week covers BFS, Dijkstra, and A*, something part of most undergrad programs. The rest of the material is ok but very basic. The programming assignments are "implement Dijkstra" in week 1, "implement Dijkstra on a Torus" (and "do triangles intersect?") in week 2, a very simple for loop in week 3, and "implement gradient descent" in week 4. I could complete the course by looking at it on-and-off over one weekend.
The material that is presented is presented well, but there's not enough of it.
For reference, part 1 of the robotics specialization had over twice as much lecture content (by time) and then lots of supplementary material in addition to that, quizzes that required some thinking, and more relevant and at least slightly harder programming assignments.