awesome

Genial

super!

great

О

This course is a great course for beginners who are stepping into the world of robotics and the perfect course to start a specialization. The course content is exciting and there are a lot of instructional concepts. However, I was not satisfied with the way the content was presented. Being used to the chalk and talk environment, and following online videos where the tutor writes the equations on the screen as they go through them, the " PPT slides and voiceover" method did not attract my attention. But the content of the course kept me going. I found some of the qualitative quiz questions quite difficult as they were not taught in the class explicitly.

Understanding quadrotors, I see that there could not be much done when it comes to programming tasks. Programming tasks Weeks 2 and 3 were extremely easy to code, but tuning the PD gain parameters was the most infuriating task. Even when I got the correct code for Week 3, I was doubting whether I made any mistakes in the code for a long time while the issue was caused by a bad PD gain ratio. The tuning part could be removed by providing a considerable range of possible values of the gains. This would save a lot of wasted time. That being said, I found the final task to be insanely interesting and the satisfaction of solving the final task was extraordinary. I would also appreciate the work done in creating the simulation model and the way the quadcopter was visualized and the trajectory was plotted.

Overall, this course pretty much teaches you all the basic theory you need to know to control a quadrotor. If you dont mind wasting a few hours tuning 6-8 values, you would walk out of this course with great satisfaction.

PS: There are a few videos of quadrotors doing insane stunts both as a swarm and individually inside the course.

I really enjoyed this course and it really hooked me onto this topic, definitely this will not be the last robotics course I am going to follow. That being said there are some points of improvement.

The main parts necessary to program the simulation assignments are covered in the materials, but there are still some of them that need to be pursued independently. Some assumptions have to be made that are not so obvious, I do not know whether it is the intention of this course to do so or not, but since they are minor footnotes that may hinder completely your results, I think they should be made explicit.

The Matlab files available for the course work as intended, what is not so obvious though is how they are intended to work, that is because there is very little documentation about them, it would be great if there was more documentation available, especially for people who have never used such tools before.

Finally, even though I personally enjoyed this course and I think it had an achievable level of challenge for me, I would definitely not recommend this course for total beginners, or people who do not have followed a math, physics or engineering degrees. The way the course is laid out, even if you do not understand the concepts behind each topic you will still manage to answer the quizzes and get the simulations to work eventually, however to truly understand what they mean and be able to replicate them in other environments requires some prior knowledge that cannot be taught in just a few hours.

I really enjoyed this course and felt the material was appropriate in scope and quality for an introduction to Aerial Robotics. I have a few issues with it though.

While this wasn't a problem for me, I feel those who haven't done maths with vectors and matrices before would struggle with the maths presented here.

There was some inconsistency in notation throughout the lectures - though I expect this will be fixed in future runs of this course.

Given how much the programming assignments involved tuning PD controllers, I think there should have been some additional material provided on how to go about doing this.

The difficulty of the last programming assignment came out of nowhere and given that you need to pass all the assignments in order to pass the course, this made it especially stressful. Having gotten almost full marks on the course until that point, the prospect of that all being for naught because I couldn't complete the final assignment in time was incredibly stressful and unfair. The simulation would fail with a Matlab ode45 error - with absolutely no information as to why that should fail. Trying to debug something by guesswork on a time limit is not a valuable educational experience.

I feel all of these issues are fixable though and on balance I think this was a very good course.

I really liked this course, as it presents a researcher's insight into simple, but effective quadcopter motion planning and control strategies. I think this course does require some background engineering knowledge. The last assignment was particularly challenging and really enjoyed. However, I have the following recommendations: 1) the time required for the assignments should be increased to realistic values. Although I found the provided PDFs (in addition to the lectures) useful, it still took time to derive the matrix equation for the snap-trajectory coefficients based on the boundary conditions. Unless the equations are provided in some form, I would suggest a minimum of 9 hours for that assignment. 2) all assignments entail some tuning of coefficients for PD controllers. It can be very confusing to obtain those values, especially if you are not sure what you have done is right and what sort of values to look at. For this reason, I recommend to add an extra step per assignment where the controller is run for a fixed point. That way it is clear from the marker if the equations are correct. Then, a document should be provided on sound strategies for obtaining the coefficients, without a need to look at the blogs all the time.

Overall a good course. But I would suggest the following updates:

-there are many formulas in the course that are not explained, just dropped in the slides

-there are many inconsistencies in notations (for instance the Kv, Kp used sometimes, other times Kd and Kp)

-programming labs should be given in an incremental approach; ex. instead of just requesting a full solution for a given problem, a stepwise approach where the functions are built up to produce intermediate results and slowly add complexity to reach that final goal is much more efficient as a learning tool

-supplementary notes are not available

-slides for the main lectures are available but they are attached to the videos; it would be better if they are added as an item in the course outline. Many other courses on Coursera do the same thing. Also it would spare the presenter to break down the slide pack into parts corresponding to the videos; just place the whole slide pack for a lesson.

This is a more difficult online course that you will come across. I give 4 stars because of the pretty nice content. Learning about writing linear controllers for the desired trajectory in 3D is the best thing we can get out of the course. I don't think this course qualifies as a beginner specialization. Basics (and perhaps little more details) of the topics from Dynamics(Euler equations, differentiation of a vector expressed in one frame wrt to another frame), controls and bit of computation(complexity) is needed. Topics like "Calculus of variation" were simply new to me and you have to work little more harder to understand the different equations which will be presented without any derivation. Biggest downside is TA's or other support staff is not pro-active. Perhaps they will improve in the future runs.

Quite a bit of the assignments was just tuning control parameters. If this is that important a part of designing controllers, you should spend more time on how to tune them. Discuss what physically reasonable or expected gains or damping to see on real world drones in lectures, and physical implications of the control parameters (i.e. is it reasonable to have pitch error gain of 50? 1000? should damping be expected to be 5 or 500? why?). It would also be helpful to discuss more of the dynamics of quad-copters, such as coupling of controls. Are there specific modes of coupled responses analogous to, say Dutch Roll? Are there any controls inputs that can be decoupled form others? Knowing this would make the tuning exercises more educational rather than just guessing.

This course was not so easy( and I mean, that's good ) Some assignments were rather difficult and they forced me to spend several nights for looking the right solution. So, sometimes I felt myself a bit tired) Fortunately I've passed that, before over of session; Good side of this course, it has a lot of material for self-study, i mean, lectures, supplement materials, related some aspects of math and matlab tutorials also good references on advanced level materials ( like Trajectory Generation and Control for Quadrobots ). Although, this stuff requires a thoughtful studying, I have a great interest to get it. I was really impressed by videos of experiments with robots. In other side, some lectures and materials have mistakes.

O curso é bastante interessante, mas você deverá estar com um nível legal de conhecimento em Cálculo e Álgebra Linear se quiser realmente aproveita-lo.

Não gostei muito de algumas das avaliações, pois existe a chance de a gente passar apenas chutando, ou seja, se errar, só tentar novamente em seguida ou em algumas horas. Como é o meu primeiro curso aqui no Coursera, não sei se todas as avaliações são assim, mas é algo que deveria mudar, de alguma forma.

Mas o que mais gostei do curso foi ele ter despertado a vontade para estudar para valer "sistemas de controle". É o que farei a seguir, antes de partir para os outros cursos dessa especialização em robótica.

Very interesting course, with some toothing problems and unrealistic timescales for many.

For me, it would have been impossible to complete the course without the support of some extremely bright and helpful classmates, active in discussion fora (almost a given in a MOOC, but nevertheless, a course should be designed such that one could complete it without the fora).

More guidance on the programming assignments would be the first thing I would focus on for the next edition, as well as setting some realistic timescales (there is probably no way for the majority of participants to complete the last assignment within the 3 hours advised for it...).

This course covered a lot of material, provided a well-rounded intro to the physics of 3D motion in the quad rotor context, included excellent supplemental videos to explain some of the tough mathematical concepts, and included well-designed practical assignments.

The bridge between lecture and assignment, particularly the trajectory planning question in the last assignment, could be refined as the course iterates, although the material I found myself struggling through that very tough problem may be that which I retain most from this class.

Over all an excellent class, and I'm looking forward to the rest of the series.

Overall course is very good to get an in-depth knowledge of control systems and motion planning for UAVs.Having worked in the field of UAVs for the past one year as a researcher the only problem that I found was that there could have been another video stating its implementation on widely available drone platforms such as on ROS and Pixhawk interface or on any other computer-flight controller interfaces so that one could try to implement this in real life on actual quadcopters. Also if any mentor or fellow learner has already done an implementation I would like to know. Overall great course and the team of mentors.

I loved the course, it was way beyond my expectations. Everything well explained. I only have on critic. I think the assignment of 5 waypoints trajectory was not in accordance with the rest of the course. Imagine the tests following a level line and abruptly changes to something too hard. I think the purpose here is to learn and not find out people with special skills. Many of our colleagues that were following the course simple gave up due to this assignment. And also took 25 hours average for many of us, that not dispose of this time, since the proposed time to complete the tasks are of three hours.

This course is quite tough, and I often thought to myself: "If this is an intermediate level, what does a professional level look like?" Especially the later assignments are quite difficult and time consuming to solve (it says, you need approximately three hours to solve the last one. Well, it took me about 30 hours to come up with a half-decent attempt).

However: if you really want to learn something new about cool flying robots (and who doesn't?) and are willing to spend quite an amount of effort and time, this is the course for you!

The course was very useful, but the study material had several printing mistakes in formulae/ derivations that caused problems for some learners. Luckily there were more experienced people around to help, but the course organisers should be more careful in their material. Also, it would help if free software like octave is used rather than matlab.....it was a nightmare to download and install on my ubuntu computer and i still get annoying telephone calls from matlab people asking me if i'm interested in buying the full version.

It is a great course for me. At the end of the course, I learned how to generate a trajectory having several set waypoints and have my drone follow the trajectory autonomously in simulation.

As opposed to the stated prerequisite, I believe the course requires engineering/math/physics background.

On the down side, there is no TA presence in the discussion forum. You will need to rely on your classmates if you are stuck. Also the second assignment requires hours of tuning, which serves almost no educational purpose.

Great amount of material for a novice. For me who has never done anything like this, but who has a ton of programming experience, the hardest part was to understand the concepts behind all the formulas. Slides and instructor's explanations were good, but required a good deal of studying to really understand it. Overall, I learned a ton and really got to know the math behind the flight control. Programming assignments got significantly more challenging as the course progressed, but also more informative.

The only thing I think that put this course on the back foot is the absence of mentors and TAs. They should be pretty active enough to clear the doubts that pops up while clearing quizzes or programming assignments. Its just the people on the forum that comes handy but even when they are not sure about the concept then the going becomes tough. Apart from that, the course is one of the best of its kind and trust me you will never regret doing this certification !!! Good job team UPenn !!!

決定提供一個中文的心得。

這堂課程為期四週，每週會有一個作業，我認為這堂課程對於並無基本力學或線性代數概念的人會比較辛苦一點，可能一不小心就要去下一班了，因為畢竟只有四週，所以很多部分會比較快速帶過（但會提供你很多參考資料來對於這方面做加強，且也提供了一些論文和講義來讓你有更深刻的理解，當然這並不是通過本課所必須的）。

前三門作業蠻符合課程的難度，只要投影片內容都理解的不錯應該就能夠穩穩寫完，但是第四週的作業難度就比較飆升了，不過說難度飆升其實也是你完全照課程內容的方式去做(minimum snap)，實際上若只是要通過的話課程提供的測試程式並沒有設定的非常嚴格，就算用比較簡易的實作方式也完全能夠應付。

給四顆星的理由在於，其實我希望這堂課程能夠開成一整個學期的長度啊，老師裡面介紹的東西感覺都非常有趣，很想要更加的深入這塊領域，老師是無人機界的大師，四週課程實在有點意猶未竟，希望能有個拉高難度至penn本身實體課程難度的版本並且拉長課堂時間，可惜這應該是不太可能的。

大致上這是個我蠻推薦的課程，但若你手邊沒有實體的無人機能玩，那就可惜了一點。

The material is very informative and generally well presented. I would have liked more video explanation about the minimum snap trajectories as this was a large part of the final assignment.

If the Matlab resource was an actual desktop application instead of online, the assignments would have been much easier and quicker to accomplish. Matlab online, in my experience, is slow and hard to work with in a single browser window.

Overall, I enjoyed this course.