0

I'm new to SLAM and I'm sure the question is dumb, but I'd really like to know how this pose and area data accumulation goes in SLAM. Once an algorithm initializes it starts to track pose (and write somewhere (where?) a sequence of coordinates) and recover a map (a point cloud stored somewhere (where?)). Am I correct? Then the algo iterates over all persisted points for loop closure purposes? Are the points stored in some DB? Or what data structure is used?

Help me please to clarify the situation! (Since it's the head of yours that may occasionally get sliced by the drone's propellers, which is driven by my future SLAM!))

Thanks in advance!

Sophour
  • 106
  • 9
  • This might get a faster answer on Signal Processing, which has plenty image processing and computer vision questions. https://dsp.stackexchange.com/ S.O. tends to be more specific technical software problems with languages, compilers, build systems, general computer science algorithms and such. – DarenW Mar 16 '18 at 18:42

2 Answers2

1

What you are asking is an implementation detail. It depends on which SLAM algorithm you are planning to use and how you want to implement it considering the performance issues for your project setting.

As far as I know, rtabmap uses a database to store the features with estimated location information to detect a loop closure in real-time.

nayab
  • 2,332
  • 1
  • 20
  • 34
-1

I have the same curiosity.I found some people use a pre-defined grid and a reference point at which the robot start from and as the robot moves discrete steps,for example 10cm for each direction ,the robot sensors triggered and the mape (grid) is updated.

The None
  • 87
  • 1
  • 1
  • 7