Introductory SLAM Course

[Tim Craig]

I just finished watching the videos to an online SLAM course.  This is a "short course" format a la the Udacity courses.  It covers developing a robot motion model, Bayes filtering with Histogram Filter, Kalman Filter, Extended Kalman Filter, and Particle Filter.  The material is centered around data gathered from a Lego based tracked robot with a laser scanner in an arena with cylindrical markers.   Programming assignments are in Python with students working within a supplied framework and adding specific functions.  The results of the instructor's correct implementation are shown and discussed.  Sebastian Thrun's "Probabilistic Robotics" is used as the reference material.

Claus Brenner of Leibniz University is the instructor.  While the instructor is German, the course is in English and his English is excellent and only mildly accented.

Students are assumed to be familiar with calculus, matrices, vectors, and basic probability with conditional probability and Bayes Theorem.  If you don't have the math, your return from this course will be limited.  You MAY be able to get an informal introduction to SLAM from watching the demos of the solved problems and listening to the explanation.  But there's no buildup to the mathematics so that part moves pretty quickly.

Here's the instructor's website for the course:  http://www.clausbrenner.de/slam.html
It contains the "partial" Python programs and data for the assignments.  It also has a link to the videos on YouTube.  There are 72 videos making up 7 topics.  None of the videos are more than about 20 minutes long with most considerably shorter.  I did the course in 7 sessions while eating lunch, minus the programming which I didn't do.

Tim

[Joe Landau]

This appears to be covering much the same territory as the Udacity course "AI for Robotics"

https://www.udacity.com/course/cs373

The lecturer is Sebastian Thrun, who is a pleasure to watch.  I did the assignments, though not while at lunch.  I thought they were not too hard, and would be very valuable if you were going to do any implementation.  As Tim says, there is math, but I don't think you need to understand all the math to do the assignments.  You are just using it.

Joe

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To post to this group, send email to hbrob...@googlegroups.com.
Visit this group at http://groups.google.com/group/hbrobotics.
For more options, visit https://groups.google.com/groups/opt_out.

[Tim Craig]

I did the Udacity CS373 course a while back. It covered SLAM is one
component of the self driving car. Since this presentation covers only
SLAM, I think there's a bit more detail. I got quite a bit out of
watching the parts where the robot was run on the data and the
commentary on why certain things were happening. Such as why the error
ellipses were changing size and orientation. Both Thrun and Brenner
have similar lecture styles although the one criticism I have for
Brenner is that in a few spots he seems to rush through the math. I
haven't checked to see how well it fits with the presentation in Thrun's
book although it would have been nice if Brenner had provided a set of
notes the equations and his manipulations could be studied at leisure.
It's really wonderful that we now have such quality material available
to us and it's free.

Tim

[tdeyle]

Hello;

Did you end up completing the code for Unit B? If so, I am running into issues with trying to replicate the results that are shown in the video. I was wondering if you would be able to assist me.

Any help would be appreciated.

Theo

[Andy Jang]

Hello:

I watched the first few videos in A and got hooked. So I signed up for udacity.
Do you know where to get the hardware. I really want to try out the Lidar sensor.

[Austin Hendrix]

You can purchase some of the smaller Lidar units from Acroname, but they tend to be pricey: http://www.acroname.com/products/index_Hokuyo.html

A cheaper alternative is to buy a Neato and disassemble it for the low-cost lidar. The downside here is that you’ll have to build the appropriate support circuitry to attach it to your project, and they tend to be a little particular to work with. You can also access the Neato’s lidar data through the debug port, without disassembling it.

-Austin

P.S. The Neato makes a nice vacuum, too!

[Patrick Goebel]

Some generous person donated a URG-04LX-UG01 to Pi Robot a few years ago and I'm amazed to see the price is still the same today.� What will it take for these prices to come down?

--patrick

http://www.pirobot.org

On 01/20/2014 06:14 PM, Austin Hendrix wrote:

You can purchase some of the smaller Lidar units from Acroname, but they tend to be pricey:�http://www.acroname.com/products/index_Hokuyo.html

A cheaper alternative is to buy a Neato and disassemble it for the low-cost lidar. The downside here is that you�ll have to build the appropriate support circuitry to attach it to your project, and they tend to be a little particular to work with. You can also access the Neato�s lidar data through the debug port, without disassembling it.

[Chris Albertson]

For experimenting you can use a Sharp sensor.  These have a beam with of about 2 or 3 degrees and you can spin them on a stepper motor.  I have not done this yet myself but it is on my short list of projects. 

 What got me started was this image. 

pointscan.jpg

 It is a 3D distance map where distance is converted to grey scale made by mechanically scanning a $10 sharp sensor using hobby servo motors.  

 

Sure it's slow but 100x less expensive.

On Mon, Jan 20, 2014 at 6:26 PM, Patrick Goebel <pat...@pirobot.org> wrote:

Some generous person donated a URG-04LX-UG01 to Pi Robot a few years ago and I'm amazed to see the price is still the same today.  What will it take for these prices to come down?

--patrick

http://www.pirobot.org

On 01/20/2014 06:14 PM, Austin Hendrix wrote:

You can purchase some of the smaller Lidar units from Acroname, but they tend to be pricey: http://www.acroname.com/products/index_Hokuyo.html

A cheaper alternative is to buy a Neato and disassemble it for the low-cost lidar. The downside here is that you’ll have to build the appropriate support circuitry to attach it to your project, and they tend to be a little particular to work with. You can also access the Neato’s lidar data through the debug port, without disassembling it.

-Austin

P.S. The Neato makes a nice vacuum, too!

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To post to this group, send email to hbrob...@googlegroups.com.
Visit this group at http://groups.google.com/group/hbrobotics.
For more options, visit https://groups.google.com/groups/opt_out.

--

Chris Albertson
Redondo Beach, California