There are currently three transmission strategies being tested:
Full map
DAG
Equality & Error Correction <- aka: "lite"
Introduced in this paper, occupancy grids can be shared by quickly pinpointing all differences in the local and remote maps. I highly recommend giving the methods section a read.
Outline
The basics of the "lite" transfer strategy are as follows:
Determin if local maps are the same
IF false -> re-check equality in sub-sections of the map
By recursively checking equality, we can pinpoint all positions which are different from the local map.
From here, the paper doesn't specify how to re-build the remote map. But I'm assuming we can just copy the local map, and ask the other robot to "get values for all diff-positions" and simply fill them in.
And we are done!
We then call the evaluation function to grade a map merge, and find how many bytes were transmitted.
Why is this novel?
The interesting part of this "lite" transfer method is that it's lossy. We can't guarantee our equality function will be correct 100% of the time. This is why we are testing this transfer method, to see if "good enough" will work for map merges.
Part of #2
Action Item(s)
Background
There are currently three transmission strategies being tested:
Introduced in this paper, occupancy grids can be shared by quickly pinpointing all differences in the local and remote maps. I highly recommend giving the methods section a read.
Outline
The basics of the "lite" transfer strategy are as follows:
By recursively checking equality, we can pinpoint all positions which are different from the local map.
From here, the paper doesn't specify how to re-build the remote map. But I'm assuming we can just copy the local map, and ask the other robot to "get values for all diff-positions" and simply fill them in.
And we are done! We then call the evaluation function to grade a map merge, and find how many bytes were transmitted.
Why is this novel?
The interesting part of this "lite" transfer method is that it's lossy. We can't guarantee our equality function will be correct 100% of the time. This is why we are testing this transfer method, to see if "good enough" will work for map merges.