TypingCat / spatial-topology-teleoperation

Teleoperation for Mobile Robot using Semantics
GNU General Public License v3.0
4 stars 0 forks source link

How to represent spatial structure #7

Closed TypingCat closed 3 years ago

TypingCat commented 3 years ago

There are 2 suggestions what the spatial structure represents. I think the first one is better.

  1. Every nodes
    • Represent spatial structure directly
    • Heavy
  2. Intersection nodes only
    • Has angles of its legs
    • Light
TypingCat commented 3 years ago

I brought 5 frames from issue #6 for a thought experiment: topology merging. This robot is passing through an intersection.

extract_topology_frames

According to the first proposal, intuitive results can be obtained. This intersection is represented with 20 nodes.

extract_topology_thought_1

According to the second proposal, the operation is minimized. This intersection is represented with only 6 nodes. However, there is some confusion when merging. Look at the central node that occurred in the 38 second frame. Each time a new topology is merged, a new leg grows. There were 5 legs at the end, but it should be 3 legs.

extract_topology_thought_2

TypingCat commented 3 years ago

The first proposal was passed.

TypingCat commented 3 years ago

There was a debate on merging nodes of samples.

  1. The first proposal is estimating position of nodes with closest nodes. It's fast(see https://github.com/TypingCat/waffle/issues/11#issuecomment-771331661) and may can respond to dynamic environment, but it cannot guarantee the exact position of intersection.
  2. The second proposal is cluster recently extracted nodes and define nodes with probability. It provides an exact position of intersection, but requires a lot of samples to be statistically significant.

I wonder if the second proposal can be computed in real time.

TypingCat commented 3 years ago

Results of the test #13 shows the second proposal is robust and not that slow. Therefore the spatial structure will be represented by intersection nodes only.