-2
$\begingroup$

I'm studying robotics and one practice that I have to do is to implement my own SLAM system on ROS.

I need some advice about where to start.

Where do I start?

$\endgroup$
2
$\begingroup$
  1. Take some of the robotics courses of burgard wolfram. You can easily find his video lectures on his homepage. You don't need to watch all of the videos. Just watch and try to understand optimization and graph-based slam part.

  2. I recommend you to implement graph-based slam in matlab first. Don't go for Extended Kalman Filter slam. They are outdated. The robotics courses of the above have some toy graph slam examples. Run that first, analyze the code, and make your own.

  3. Once you are confident with the theory start to implement it in ROS. Don't try to implement something you are not 100% sure in c++ because the debugging is hell with c. Or the best way is to start with open source slams such as LOAM.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for your answer, but I don't have a lot of time to do that, only two weeks. And I think that is a lot work to do only in two weeks. Am I wrong? $\endgroup$ – VansFannel May 23 '19 at 5:32
  • $\begingroup$ You are right. With two weeks I would suggest ICP based one. You can implement frame to frame point cloud registration with PCL which is basically a simplified version of SLAM. Find LiDAR or RGBD dataset recorded in ROS bag and put them to you ICP slam ros package as an input. Two weeks could be extremely short if you are not familiar with ROS. $\endgroup$ – C.O Park May 23 '19 at 5:49
  • $\begingroup$ Thanks. I'm familiar with ROS. The point is to implement SLAM by myself, and use it in a simulation with ROS and Gazebo. I know how to move a robot in Gazebo using ROS' topics. This is only an university subject: I only need to implement SLAM algorithm by myself and, if I have time, use it in a simulation. $\endgroup$ – VansFannel May 23 '19 at 7:29
0
$\begingroup$

Where to start will very depending on your current skill-set, as well as the specifics of the course you're studying. My recommendation is start with the Extended Kalmann Filter as well as Bayesian Filtering. These will form a foundation for your SLAM (Depending on how to want to approach it). From there, I guess all you can do is have a look at what resources there are online for implementing it. I would recommend either Here or, if you can, Here. I can personally vouch for the Probabilistic robotics textbook being a fantastic resource.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Welcome to Robotoics J U R A P H. On stack exchange we try to avoid distracting readers from our answers with lines like as 'G'day,' and "Good luck!". While it may seem counter intuitive, chattiness is often considered impolite, as giving people extra text to read, even if they ignore it, is disrespectful of their time. $\endgroup$ – Mark Booth Jun 21 '19 at 14:28

Not the answer you're looking for? Browse other questions tagged or ask your own question.