Modern Robotics

Recent discussion: https://news.ycombinator.com/item?id=19123506 and top comment explains why this book is useful: https://news.ycombinator.com/item?id=19124052

something related to robotics also discussed recently: https://news.ycombinator.com/item?id=19125858

I've read the book / finished the course. What this course doesn't teach you: getting sensor data, figuring out the environment ( this part is called mapping ) from your sensor input ( for ex. locating feature points / getting 3d position of these points from a stereo camera ), locating your own position from the sensor data. In short it doesn't teaches you SLAM ( simultaneous localization and mapping ), which is a good thing, because that is a very big subject.

What it teaches you:

- basics of robotics ( movement freedom, a little bit of topology )

- math necessary for robotics ( describing rotation and displacement in a single matrix, representation of rotations around a point, representation of forces on a body, solving differential equations given these things )

- robot kinematics: what will be the end position if my joint angles are x ( forward kinematics ), what should the joint angles be if my end position is x ( inverse kinematics )

- robot dynamics: same thing like kinematics, but positions => forces/torques. I think this is the hardest part of the whole subject. Although these problems can be solved in a close form, it's not practical, and you'll need some physics knowledge to be able to comprehend everything. Nothing really fancy, it's just needs more time if you physics is not up to this level.

- control: controlling those robots with motors. Describes feedback / feedforward control. In short: if you have a good enough model, then it's not that hard to control a robot, since you only have to control the 'difference' between your model and reality.

- Path planning with obstacles

- Generating control for paths ( your robot has physical constraints, for ex. it takes time to accelerate. You have to take those things into considerations )

- How to make your robot follow a path / how to make paths with physical constraints ( for ex. a car can not go sideways )

- Grasping

I'd say the book is pretty awesome, it's between the line of academic-too high level, and in a good way. You won't be a pro after reading it, since almost every problem gets harder in reality, but you'll get a good enough understanding which will make diving into a subject easy. Lot of excersizes.

Also, it is very readable, I haven't lost myself in any of the parts. Another good thing I've found that it's easy to just skim trough the book, and have general understanding. The youtube videos help a lot with that, I'll suggest you watch them before/after reading a chapter.

Or probably the intent is to make robots to do web development ;-)

I have zero experience with web development except for some basic html stuff and hacking around with java-script of third party websites. I am an electrical engineering undergraduate. Mastering java-script is indeed in my wish list, but it lags far behind robotics and control systems.

Robots are cool. Robotics is just slightly out of the experience of the average dev to feel interesting.

YES!

recently? Robotics has done well on HN for a long time.

Robotics is an enormous field. This book is heavily steered towards kinematics (ie how stuff moves). There's another book on their recommended list which details how to drive motors and use sensors. That doesn't begin to cover machine vision for perception and navigation, general control theory, or the sorts of algorithms in probabilistic robotics. Let alone machine learning!

If you want a fairly all-inclusive reference, look at the Springer Handbook of Robotics. You'll need a pickup truck to carry it home though. It's more an encyclopaedia than a textbook, but it is surprisingly comprehensive.

Also, most robots don't use reinforcement learning (and most people don't have a basement full of robot arms learning to grasp things). That's a nice research area, but in practice it's not needed for a lot of tasks.

I like the suggestion but I think I understand why it's not there. Given that RL applied to robotics is a research area with quite a lot of knowledge prerequisites that are not covered in classical control, I think that a single chapter would not be enough to be of practical use. There should be a mention of RL for sure (I'm not sure if there is one), but it seems reasonable to consider that proper coverage would bring too much extra scope.

It seems to be more of a robotics engineering textbook, giving known solutions to common problems, than a robotics science textbook. Reinforcement learning in robotics is an interesting field of research but you're not going to use it to build cars.

Printable version with 10pt font and margins [1] Tablet version [2]. Both with hyperlinks. [3], [4] for without hyperlinks.

FYI, these links are in the Book section [5]

[1] http://hades.mech.northwestern.edu/images/7/7f/MR.pdf [2] http://hades.mech.northwestern.edu/images/4/4e/MR-tablet.pdf [3] http://hades.mech.northwestern.edu/images/8/81/MR-largefont.... [4] http://hades.mech.northwestern.edu/images/b/b2/MR-2up.pdf [5] http://hades.mech.northwestern.edu/index.php/Modern_Robotics...