MOOC courses in robotics
In the coming years, the entire labor market will change unrecognizably and irrevocably.
Robots will replace people in agriculture, trade, at work; they will take on such seemingly “human” work as cleaning the streets, managing warehouse loaders, combine harvesters and buses, placing cans of yogurt on the supermarket shelves, putting out fires, eliminating the consequences of natural disasters and caring for the elderly and sick.
For example, in agriculture, mobile robots will be able to dose or fertilize a particular plant, visually analyzing its condition, remove every weed on the field and work 24 hours 365 days a year with interruptions for refueling and maintenance.
')
The introduction of robotics will increase labor productivity in these industries at times, multiply reducing the cost of doing business and cause significant reductions in low-skilled staff.
Changes in the labor market will be much more ambitious than those that occurred due to the introduction of steam engines, mechanized looms or computers into production - a significant number of people of working age will find themselves below the employment level without the opportunity to get any work due to their low qualifications .
The only industry that will be “in the black” by the results is the development, production, programming and maintenance of robots.
The open competitions DARPA Robotic Challenge , like the DARPA Grand Challenge and the DARPA Urban Challenge earlier, clearly demonstrate where and in what pace modern robotics are moving.
In this note, I would like to briefly describe the courses on robotics currently available on MOOC platforms.
So, in order of increasing complexity.
The Control of Mobile Robots course is available on the Coursera platform and is provided by the Georgia Institute of Technology (GT).
Prerequisites - linear algebra and MATLAB (if you want to solve practical problems).
Lecturer Dr. Magnus Egerstedt (Magnus Egerstedt).
Assistant Dr. Egerstedt Jean-Pierre de la Croix ( NASA Jet Propulsion Laboratory RoboSimian) took part in the DARPA Robotic Challenge (they took 5th place using their own design robot).
The course consists of lectures and seminars, tests and practical tasks performed in MATLAB . Due to the characteristics of the toolbox, Octave will not work, it is MATLAB that is needed.
An additional bonus of the course is the opportunity to assemble your own robot on the SparkFun Magician chassis.
The course provides a list of necessary parts and coordinates of suppliers, as well as video instructions for assembly. The BeagleBone Black is used as the computing platform for the self-assembling mobile robot.
The second run of the course was completed in 2014, now its status is “Archive”, but you can subscribe to it, see all the lectures and seminars, answer control questions. Verification works, but estimates cannot be obtained. The validation of the solution of practical problems in MATLAB also works.
The course took place over two seasons (2013 and 2014). According to Egerstedt statistics, out of 40,000 enrolled in the course in 2013, more than 20,000 were eliminated after the first week, and more than 5,000 people completed the course successfully. The result is extremely good not only for a technical course, it significantly exceeded the “average for MOOC”.
In my opinion, the course is quite simplified. The reasons for which this was done, most likely to try to artificially raise the number of "finishers".
Although the course does not have an official book, Egerstedt recommended two on the forum: Feedback Systems: Aström, Murray by Princeton University Press (downloaded from the author’s website) and Feedback Control of Dynamic Systems by Prentice Hall by Franklin , Powell, Emami-Naeini.
“Logo” and “branding” graphics of all Georgia Institute of Technology courses on the Coursera have been updated, so, perhaps, the course will be restarted.
The CS373 Artificial Intelligence for Robotics course works on the Udacity platform and is branded as provided by the Georgia Institute of Technology (GT) as part of their Online Master of Science (OMS CS) program.
On-demand course.
Classic by Sebastian Thran (Sebastian Thrun).
Tran was the leader of the Stanford University team — winners of the Grand Challenge and second place winners in the Urban Challenge (lost to the team of long-time “friend” Tran from Carnegie Mellon University (CMU)).
The wiki article gives an idea of ​​who this person is and how his work has advanced the progress of all mankind.
Prerequisites - no. Linear algebra, Python , probability theory in the volumes necessary for the course are explained in the course of the course.
Some sections of the course were built based on the 2006 book by Thrun, Burgard, Fox Probabilistic Robotics of MIT Press (you can find a draft of the first edition in Google).
The course format is “corporate” from Udacity - something like a “dialogue” of a lecturer with a student, with a check of the understanding of the material right in the course of the lectures; control questions and practical tasks on programming the course material in Python .
At the end of the course, an exam and practical work, the results of which cannot be "peeped".
In my opinion, a significant drawback of the course is that because of the format of the presentation of the material you often do not understand what you are doing - “you do not see the forest for the trees”. The presence of the book by Tran helps a lot to clarify what he is trying to explain, what it is for and most importantly, how and where all this can then be practically used.
Since 2014, Udacity has stopped giving a certificate for completing this course.
The course has undergone a re-branding and is now on the list offered by the Georgia Institute of Technology (GT) through the Udacity platform as part of their Online Master of Science in Computer Science (OMS CS) program, although the course was originally developed by Tran.
The AUTONAVx Autonomous Navigation for Flying Robots course is provided by the edX platform and developed by the Technical University of Munich (MTU) (Technische Universität München - TUM).
Lecturers Jürgen Sturm (Jürgen Sturm) from (then) MTU, he now works at Metaio GmbH, Apple’s “daughter” and Professor Daniel Kremens (Daniel Cremers) from MTU.
Prerequisites - linear algebra, Python .
According to edX statistics, out of an estimated 20,000 enrolled in the summer of 2014, 1,400 students completed it successfully.
It is assumed that the owner of the quadrocopter Parrot AR.Drone , who has successfully completed this course, will then be able to program it himself.
Since JĂĽrgen Sturm no longer works at MTU, so most likely the course will be reformatted in the future (if there will be any).
The course format is reading PowerPoint presentations, test questions and practical work in Python .
You can download a quadcopter simulator on a python, with which you can practice programming a virtual quad.
The course is based on Sturm’s lectures at MTU, the aforementioned book by Tran, and Richard Zeliski’s book (Richard Szeliski) by Springer Computer Vision: Algorithms and Applications , the draft of which can be downloaded from the author’s website.
I was particularly interested in the extreme lecture in which Sturm presented the scientific results of a group of scientists from the Computer Vision Group of the MTU - 3D reconstruction using a video camera or a Kinect sensor on a drone, visual odometry, SLAM using PTAM , etc.
The course is completed, but all the tasks and lectures are available.
The ETHx Autonomous Mobile Robots course is available on the edX platform and is developed by the Swiss Technical School of Zurich (Eidgenössische Technische Hochschule Zürich - ETH Zurich).
On-demand course - will go until December 2015.
Prerequisites - linear algebra.
There are clear introductory lectures on MATLAB , covering 100% of all that is required for the course.
Based on the 2nd edition of the book by MIT Press, the authors of the course Siegwart, Nourbakhsh and Scaramuzza Introduction to Autonomous Mobile Robots . The entire book is available as a set of pictures as an additional material to the course.
The book is sensible, however, part of the material is missing in it and for some sections of the course it is better to use lectures - in particular, there is nothing in the book about the features of the mathematical description of walking robots and industrial manipulators.
Part of the Kalman filter course is based on the relevant chapter from Tran's book.
The book is especially good with a detailed bibliography containing the newest sources of information on robotic topics for 2011 - books and articles.
Both editions of the book are in Google in the form of pdf.
Course format: lectures in the form of reading presentations, test questions and practical work on MATLAB with the Symbolic toolbox.
Strictly speaking, MATLAB for the course is not needed, all solutions can be checked directly in the browser. The number of answers to test questions and tasks is unlimited.
The lecture material is slightly simplified, the tasks in the examinations of some sections are simplified significantly.
There are additional tasks for MATLAB .
According to the results of the course they promised to give a certificate of participation to all who completed 60% of the tasks. There is the option of obtaining a valid certificate from edX .
In my opinion, the strengths of the course are its visibility, the book on which it is based and the overall "polished" course.
One of the creators of the course, Paul Furgale, initially went to work at Apple for a team working on their version of an autonomous car.
Queensland University of Technology (Queensland University of Technology, QUT) on its own edX-like platform offers two courses by Professor Peter Cork (Peter Corke) Introduction to Robotics (went first week) and Robotic Vision (will begin on October 19).
Both courses are based on his book, published in 2011 by Springer Robotics, Vision and Control: Fundamental Algorithms in MATLAB (now the professor is finishing a new edition). Parts of the book are kindly provided by the professor as an outline of the courses.
The book can be found in "Google". In my opinion, the book is heavily biased to the practical plane - how to properly use its toolboxes - with a minimal explanation of the theory.
The bibliography of the book contains relevant sources on the subject of robots - books and articles.
Course format - video lectures, quizzes and assignments for MATLAB .
MathWorks licenses MATLAB for the duration of the course.
They use their own toolbox Cork, developed by him for MATLAB - Machine Vision Toolbox and Robotics Toolbox .
The main emphasis in the presentation is made on industrial robots.
An additional option is to assemble your own robotic arm from the Lego Mindstorms constructor and perform a practical task on it.
Everyone who will show the result is not worse than 50%, they promise a certificate.
A two-part course from Seoul National University (Seoul National University, SNU) on the edX platform.
The first part ended on May 9, 2014, the second - on August 3, 2014.
Lecturer Professor Frank Chongwoo Park.
The status of courses on the edX "Archive", but you can subscribe to them and all materials (including verification of the test) are available.
I am delighted with this course. Fair. You do not often see such a clear presentation of such a complex material, which is promoted by the magnificent English professor and his subtle sense of humor.
Course format - high quality lectures and answers to test questions.
The focus of the course is on the description of industrial robot arms.
The course is based on, apparently, parts of the future book of the lecturer. They can be downloaded by clicking on the “Lecture Notes” link, right here and here .
The following books are recommended: Craig John J. Introduction to robotics: Mechanics and control , 3rd edition of 2004 published by Prentice Hall , R. Murray, Z. Li, S. Sastry A 1994 CRC Press published Matrix Introduction to Robotic Manipulation author, M. Spong, S. Hutchinson, M. Vidyasagar Robot Modeling and Control 2006, published by Wiley and B. Siciliano, L. Sciavicco, L. Villani, G. Oriolo Robotics: Modeling, Planning and Control 2008 by Springer .
The Underactuated Robotics course is provided on the edX platform.
Lecturer Professor at the Massachusetts Institute of Technology (Massachusetts Institute of Technology - MIT) Russ Tedrake led the Team MIT team at the DARPA Robotic Challenge (their Google Atlas took 6th place).
The course ended on December 19, 2014, the status on edX “Archive”, but it is possible to subscribe to it and all materials (including verification of the verification works) are available.
The point of the course is how to use the dynamics of the robot as efficiently as possible, for example, to increase its mobility and reduce the energy costs of moving.
The course program and bibliography here .
Most mathematically complex of all. No simplifications, the course is based on lectures that Tedraik read for several years at MIT and, apparently, on a future book .
Prerequisites: linear algebra, mathematical analysis, dynamics, linear and nonlinear systems, TAU, MATLAB .
The professors own toolbox is used - Drake .
Robots will replace people in agriculture, trade, at work; they will take on such seemingly “human” work as cleaning the streets, managing warehouse loaders, combine harvesters and buses, placing cans of yogurt on the supermarket shelves, putting out fires, eliminating the consequences of natural disasters and caring for the elderly and sick.
For example, in agriculture, mobile robots will be able to dose or fertilize a particular plant, visually analyzing its condition, remove every weed on the field and work 24 hours 365 days a year with interruptions for refueling and maintenance.
')
The introduction of robotics will increase labor productivity in these industries at times, multiply reducing the cost of doing business and cause significant reductions in low-skilled staff.
Changes in the labor market will be much more ambitious than those that occurred due to the introduction of steam engines, mechanized looms or computers into production - a significant number of people of working age will find themselves below the employment level without the opportunity to get any work due to their low qualifications .
The only industry that will be “in the black” by the results is the development, production, programming and maintenance of robots.
The open competitions DARPA Robotic Challenge , like the DARPA Grand Challenge and the DARPA Urban Challenge earlier, clearly demonstrate where and in what pace modern robotics are moving.
In this note, I would like to briefly describe the courses on robotics currently available on MOOC platforms.
So, in order of increasing complexity.
Control of Mobile Robots
The Control of Mobile Robots course is available on the Coursera platform and is provided by the Georgia Institute of Technology (GT).
Prerequisites - linear algebra and MATLAB (if you want to solve practical problems).
Lecturer Dr. Magnus Egerstedt (Magnus Egerstedt).
Assistant Dr. Egerstedt Jean-Pierre de la Croix ( NASA Jet Propulsion Laboratory RoboSimian) took part in the DARPA Robotic Challenge (they took 5th place using their own design robot).
The course consists of lectures and seminars, tests and practical tasks performed in MATLAB . Due to the characteristics of the toolbox, Octave will not work, it is MATLAB that is needed.
An additional bonus of the course is the opportunity to assemble your own robot on the SparkFun Magician chassis.
The course provides a list of necessary parts and coordinates of suppliers, as well as video instructions for assembly. The BeagleBone Black is used as the computing platform for the self-assembling mobile robot.
Course sections:
- Introduction to Automatic Control Theory
- Mobile robots
- Linear systems
- Control systems design
- Hybrid control systems
- Navigation
The second run of the course was completed in 2014, now its status is “Archive”, but you can subscribe to it, see all the lectures and seminars, answer control questions. Verification works, but estimates cannot be obtained. The validation of the solution of practical problems in MATLAB also works.
The course took place over two seasons (2013 and 2014). According to Egerstedt statistics, out of 40,000 enrolled in the course in 2013, more than 20,000 were eliminated after the first week, and more than 5,000 people completed the course successfully. The result is extremely good not only for a technical course, it significantly exceeded the “average for MOOC”.
In my opinion, the course is quite simplified. The reasons for which this was done, most likely to try to artificially raise the number of "finishers".
Although the course does not have an official book, Egerstedt recommended two on the forum: Feedback Systems: Aström, Murray by Princeton University Press (downloaded from the author’s website) and Feedback Control of Dynamic Systems by Prentice Hall by Franklin , Powell, Emami-Naeini.
“Logo” and “branding” graphics of all Georgia Institute of Technology courses on the Coursera have been updated, so, perhaps, the course will be restarted.
Artificial Intelligence for Robotics
The CS373 Artificial Intelligence for Robotics course works on the Udacity platform and is branded as provided by the Georgia Institute of Technology (GT) as part of their Online Master of Science (OMS CS) program.
On-demand course.
Classic by Sebastian Thran (Sebastian Thrun).
Tran was the leader of the Stanford University team — winners of the Grand Challenge and second place winners in the Urban Challenge (lost to the team of long-time “friend” Tran from Carnegie Mellon University (CMU)).
The wiki article gives an idea of ​​who this person is and how his work has advanced the progress of all mankind.
Prerequisites - no. Linear algebra, Python , probability theory in the volumes necessary for the course are explained in the course of the course.
What is being studied:
- Statistics
- Probability theory
- Kalman filter
- Particle Filter
- Planning
- SLAM
Some sections of the course were built based on the 2006 book by Thrun, Burgard, Fox Probabilistic Robotics of MIT Press (you can find a draft of the first edition in Google).
The course format is “corporate” from Udacity - something like a “dialogue” of a lecturer with a student, with a check of the understanding of the material right in the course of the lectures; control questions and practical tasks on programming the course material in Python .
At the end of the course, an exam and practical work, the results of which cannot be "peeped".
In my opinion, a significant drawback of the course is that because of the format of the presentation of the material you often do not understand what you are doing - “you do not see the forest for the trees”. The presence of the book by Tran helps a lot to clarify what he is trying to explain, what it is for and most importantly, how and where all this can then be practically used.
Since 2014, Udacity has stopped giving a certificate for completing this course.
The course has undergone a re-branding and is now on the list offered by the Georgia Institute of Technology (GT) through the Udacity platform as part of their Online Master of Science in Computer Science (OMS CS) program, although the course was originally developed by Tran.
AUTONAVx Autonomous Navigation for Flying Robots
The AUTONAVx Autonomous Navigation for Flying Robots course is provided by the edX platform and developed by the Technical University of Munich (MTU) (Technische Universität München - TUM).
Lecturers Jürgen Sturm (Jürgen Sturm) from (then) MTU, he now works at Metaio GmbH, Apple’s “daughter” and Professor Daniel Kremens (Daniel Cremers) from MTU.
Prerequisites - linear algebra, Python .
According to edX statistics, out of an estimated 20,000 enrolled in the summer of 2014, 1,400 students completed it successfully.
It is assumed that the owner of the quadrocopter Parrot AR.Drone , who has successfully completed this course, will then be able to program it himself.
What is being taught:
- Different types of representation of coordinates and orientations of a solid in 2D and 3D, transitions between them
- Homogeneous transformations
- Actuators and control systems
- PID controller
- Kalman filter
- Localization
- Visual odometry
- SLAM and 3D reconstruction
Since JĂĽrgen Sturm no longer works at MTU, so most likely the course will be reformatted in the future (if there will be any).
The course format is reading PowerPoint presentations, test questions and practical work in Python .
You can download a quadcopter simulator on a python, with which you can practice programming a virtual quad.
The course is based on Sturm’s lectures at MTU, the aforementioned book by Tran, and Richard Zeliski’s book (Richard Szeliski) by Springer Computer Vision: Algorithms and Applications , the draft of which can be downloaded from the author’s website.
I was particularly interested in the extreme lecture in which Sturm presented the scientific results of a group of scientists from the Computer Vision Group of the MTU - 3D reconstruction using a video camera or a Kinect sensor on a drone, visual odometry, SLAM using PTAM , etc.
The course is completed, but all the tasks and lectures are available.
ETHx Autonomous Mobile Robots
The ETHx Autonomous Mobile Robots course is available on the edX platform and is developed by the Swiss Technical School of Zurich (Eidgenössische Technische Hochschule Zürich - ETH Zurich).
On-demand course - will go until December 2015.
Prerequisites - linear algebra.
There are clear introductory lectures on MATLAB , covering 100% of all that is required for the course.
Based on the 2nd edition of the book by MIT Press, the authors of the course Siegwart, Nourbakhsh and Scaramuzza Introduction to Autonomous Mobile Robots . The entire book is available as a set of pictures as an additional material to the course.
The book is sensible, however, part of the material is missing in it and for some sections of the course it is better to use lectures - in particular, there is nothing in the book about the features of the mathematical description of walking robots and industrial manipulators.
Part of the Kalman filter course is based on the relevant chapter from Tran's book.
What is being taught:
- Displacement and kinematics
- Direct and inverse kinematic problems
- Sensors
- SLAM
- Planning
The book is especially good with a detailed bibliography containing the newest sources of information on robotic topics for 2011 - books and articles.
Both editions of the book are in Google in the form of pdf.
Course format: lectures in the form of reading presentations, test questions and practical work on MATLAB with the Symbolic toolbox.
Strictly speaking, MATLAB for the course is not needed, all solutions can be checked directly in the browser. The number of answers to test questions and tasks is unlimited.
The lecture material is slightly simplified, the tasks in the examinations of some sections are simplified significantly.
There are additional tasks for MATLAB .
According to the results of the course they promised to give a certificate of participation to all who completed 60% of the tasks. There is the option of obtaining a valid certificate from edX .
In my opinion, the strengths of the course are its visibility, the book on which it is based and the overall "polished" course.
One of the creators of the course, Paul Furgale, initially went to work at Apple for a team working on their version of an autonomous car.
Introduction to Robotics
Queensland University of Technology (Queensland University of Technology, QUT) on its own edX-like platform offers two courses by Professor Peter Cork (Peter Corke) Introduction to Robotics (went first week) and Robotic Vision (will begin on October 19).
Both courses are based on his book, published in 2011 by Springer Robotics, Vision and Control: Fundamental Algorithms in MATLAB (now the professor is finishing a new edition). Parts of the book are kindly provided by the professor as an outline of the courses.
The book can be found in "Google". In my opinion, the book is heavily biased to the practical plane - how to properly use its toolboxes - with a minimal explanation of the theory.
The bibliography of the book contains relevant sources on the subject of robots - books and articles.
Course format - video lectures, quizzes and assignments for MATLAB .
MathWorks licenses MATLAB for the duration of the course.
They use their own toolbox Cork, developed by him for MATLAB - Machine Vision Toolbox and Robotics Toolbox .
The main emphasis in the presentation is made on industrial robots.
Sections:
- Introduction to Robotics
- Representation of orientation and movement in 2D and 3D
- Time-varying coordinate systems
- Direct and inverse kinematic problems
- Speed ​​in 2D and 3D
- Control of robot links
- Dynamics of solids
An additional option is to assemble your own robotic arm from the Lego Mindstorms constructor and perform a practical task on it.
Everyone who will show the result is not worse than 50%, they promise a certificate.
Robot Mechanics and Control
A two-part course from Seoul National University (Seoul National University, SNU) on the edX platform.
The first part ended on May 9, 2014, the second - on August 3, 2014.
Lecturer Professor Frank Chongwoo Park.
The status of courses on the edX "Archive", but you can subscribe to them and all materials (including verification of the test) are available.
I am delighted with this course. Fair. You do not often see such a clear presentation of such a complex material, which is promoted by the magnificent English professor and his subtle sense of humor.
Course format - high quality lectures and answers to test questions.
The focus of the course is on the description of industrial robot arms.
Sections:
- Representation of coordinates and orientation of a solid in 2D and 3D
- Direct and inverse kinematic problems
- Representation of orientation and movement of links through the parameters of Denavit-Hartenberg
- Exponential representation of the orientation and movement of the links of the manipulator
- Singularities
- Speed ​​and acceleration links
- Manipulator link control
The course is based on, apparently, parts of the future book of the lecturer. They can be downloaded by clicking on the “Lecture Notes” link, right here and here .
The following books are recommended: Craig John J. Introduction to robotics: Mechanics and control , 3rd edition of 2004 published by Prentice Hall , R. Murray, Z. Li, S. Sastry A 1994 CRC Press published Matrix Introduction to Robotic Manipulation author, M. Spong, S. Hutchinson, M. Vidyasagar Robot Modeling and Control 2006, published by Wiley and B. Siciliano, L. Sciavicco, L. Villani, G. Oriolo Robotics: Modeling, Planning and Control 2008 by Springer .
Underactuated Robotics
The Underactuated Robotics course is provided on the edX platform.
Lecturer Professor at the Massachusetts Institute of Technology (Massachusetts Institute of Technology - MIT) Russ Tedrake led the Team MIT team at the DARPA Robotic Challenge (their Google Atlas took 6th place).
The course ended on December 19, 2014, the status on edX “Archive”, but it is possible to subscribe to it and all materials (including verification of the verification works) are available.
The point of the course is how to use the dynamics of the robot as efficiently as possible, for example, to increase its mobility and reduce the energy costs of moving.
The course program and bibliography here .
Most mathematically complex of all. No simplifications, the course is based on lectures that Tedraik read for several years at MIT and, apparently, on a future book .
Prerequisites: linear algebra, mathematical analysis, dynamics, linear and nonlinear systems, TAU, MATLAB .
The professors own toolbox is used - Drake .
Source: https://habr.com/ru/post/265093/
All Articles