The course is very good.\n\nThe classes are well taught and show general concepts. It is necessary to do research on the internet, to solve the assignments. This is not a bad thing in my point of view
I think this is very good course of aerial robotics research. Being a student of robotics, I feel that some of stuffs in this course needs a good background in control and mechanical engineering.
автор: Barak R•
second assignment is not well explained.
автор: Anushka A•
Very thorough, neat and well explained.
автор: Vishnu T•
Very Challenging but taught me a lot
автор: Vishal G•
This Is the best course on coursera.
автор: Eric S•
Final Assignment was very unclear
автор: ATISH K•
автор: Carlos M•
Recién empecé. Pinta bien!
автор: balagopal p k•
Mind Blowing course...!!!
A bit hard for beginners
автор: Rajat S•
it was really worth it.
автор: Muhammed A E f a•
it is so awesome course
автор: Ketha J S•
автор: Fabio B•
Very good course!
автор: Yug A•
Awesome course !
автор: Joel N•
автор: Ahmed m a a f•
автор: Mauro S Q C•
автор: Naveen K•
автор: Francesco V•
the course and the lessons are really interesting and the professor is explaining really well. Especially nice are all the videos demonstrating what have been done so far.
Problem are the assignments:
-all the assignments are about controlling a quadrotor, but a matlab code 95% done is already provided and most of the times the only thing requested is to spend days in find (randomly) the correct values for the controller gains;
-the last assignment is again partly about (randomly) looking for gain values and partly about generating a trajectory as a piecewise 7th order polynomial, where the method for creating such thing is only mentioned in the assignment pdf and with wrong conditions. Even with the correct conditions, with a previous knowledge of numerical methods you can go through it, but otherwise it takes quite some time;
-the grade is not set properly.. you either take 0points or the maximum, meaning that for just passing the assignment you have the same difficulty as getting full marks.. if you arrive (after days of trial and error) to the point of passing the assignment, then you can easily get full grade with small changes.. which makes no sense;
-the explanatory pdf for the assignments or the provided codes some times contain errors, and I don't understand why they are not fixed considering that this course is already the second time that was provided;
-the assignments are fully based on matlab and to actually understand something it is required quite a deep knowledge of how to use it.. I work daily with matlab and had some difficulties in following what it is done. If somebody is new to matlab, it will be quite difficult to do even the easier tasks;
-the evaluation function for the last assignment is extremely slow.. it took me 40minutes to evaluate the results.. and considering the many iterations usually required for (randomly) finding the gains this really doesn't help.
Summarising I am quite happy about this course and glad to have done it, but please fix the many problems in the assignments. I have taken many courses online in different platforms and this was so far the most time consuming one..
Overall good course that would definitely make you spend more time reading and learning on the side. Would recommend it those who have good background in college level math like Linear Algebra and a little bit of Robotics Background from a Math perspective - like working with translations, rotations, transformation matrices of that sort to make the transition easier. (There are basic robotic courses online to help with that, like Peter Corke's course). But overall this course requires some significant effort to explore the material from an external perspective.
Some issues however -
1. Lack of added resources like reading material to support the course and help advanced students go beyond the course themselves.
2. Severe lack of activity from TA(s) on the forums. It's good that students get to interact among themselves and learn on their own, but every single post should be either answered by a TA to clarify doubts or they should acknowledge that another student's explanation was good enough to answer a particular question.
3. The in-video quizzes weren't up-to-the-mark as per me. One question asked "why" and the answer was literally "because that's how that algorithm is".
4. This is a trend in MOOCs and I don't think it can be helped, but perhaps more assignments that help understand the concepts better with examples would help students go beyond. If this course had such optional assignments that would be great for understanding the concepts with a more hands-on approach. But this is probably not the best platform to do so.
Overall would recommend future iterations of this course, especially if the first two points above are improved upon.
автор: Manoj R•
Good intro to the subject with clear instruction and responsive support by teaching staff.
Could use many more examples and/or solved problem to help independent learning.
Some notation is confusing, esp. in the later part of the course. Also, the subtle distinctions between a reference trajectory and a "desired" trajectory (in the context of the inner loop for attitude control) are not very clear. There is some art to this which is key to solving the programming assignments that the course could dwell on a lot more.
Pinned conversations in the forums are a big help in tackling the programming assignments.
Overall, the matlab sandboxes, and simulations in the programming assignments, do go a long way in helping us learn about the practical aspects of implementing controllers.
The course could benefit from a top-down (as opposed to bottom-up) overview. For example, it could be stated, early in the controls part *and in context*, that trajectory planning (which will be dealt with later), would yield a reference trajectory for which a controller is being implemented. Instead, it is simply supposed that there is a reference trajectory already available that the controller can try to track. This is just an example.
Some matlab tips on how to add more graphs would help. For example, simulations plot (x,y,z) and velocities but hints on how to visualize rotation angles would also help. The data structures used convert between 2 different "state" structures, one of which does not store angles (presumably because they are not externally specified) but this is what is available at the entry/exit into the code.
автор: Avish M•
The content of the course was good. The lectures were thorough and covered the material well. However, the programming assignments were not (in my opinion) done very well. The instructions were quite vague and didn't give a very good idea of what we were supposed to do. Secondly, the second and third programming assignments had us "tune" certain parameters so that the robot followed the specified path. However, the problem with this was that they gave us no indication of what numbers the parameters should be and given that there were 6 (for the second assignment) and 12 (for the third assignment) parameters to tune to precision, it was an enormous pain and extremely difficult. If they had given us even a range of maybe like 30 values that each parameter should be in, that would have made it so much easier, but there was no indication of any sorts. The only way I got through the assignments was due to guide put together by students and posted on the discussion boards that helped walk through the assignments and told us where to start in terms of the parameters. The content of the course was good, but there is a lot of improvement that could be done with the programming assignments.
автор: Alexey M•
The course is definitely has some value. At least it helps to feel the taste of quadrotor control. But there are some cons. 4 weeks is too short, and this entails fragmentation of course materials. A lot of reasonable questions are not answered and even not mentioned: we use Euler's angles, which lead to singularities and ambiguities, there are quaternions, but we'll not use it. Why? - no answer. PD gains tuning is manual. Is there any robust methods to build controls automatically? - no answer. And so on and so forth. And the last... lections are really boring, lector monotonically reads text, nevermind what - matrices, formulas, etc. He reads all of it in completely, with the same manner, without any interactivities with the slides. Actually it is realy demotivates. I guess one of students that made supplementary materials with marker will do much much better.
Anyway, thanks for the course. I believe that highly motivated learner can get something new from it.
автор: Akshit J•
This course is more of an introduction/overview of the parts of theory for navigation and mapping of drones. Controller part is touched more in depth with a great video on differential flatness. You will easily understand Vijay kumar's paper on "Minimum Snap Trajectory Generation and Control for Quadrotors" after watching this. Topics such as SLAM for mapping and localization, State estimation(Kalman filter) actuators are touched upon very lightly. Nowhere near with intent for implementation.
The programming exercises in this course are terrible. In a nutshell in the first 2 weeks you are told to tune some PD gains by just varying values in an input field in a GUI. In next 2 weeks everything is again about manually tuning PD control gains as dynamics equations are straight forward in writeup. All the complexity/dynamics of equations is replaced by PD gains and linear assumptions.
автор: Deleted A•
I truly love the course program and what it covers. And aiding student who can't afford the price by grants is relly nice. Then were students alloweed to download MATLAB freely! But teachers should be clear from the begining, what topics one must grasp before getting one's head into it. I don't think calculus and Linear algebra is enough, given some topics that are spoken about in the course. The teacher can explain mean things and in a second, speaks about totally incomprehensibe subject. Last but not least, I personally think the course lacks on additional materials. It would be really nice if there was something like handouts, pdf readings for deeper understanding of the concepts on offline times.