root / branches / slam / code / projects / slam / computer / slam.txt @ 599
History | View | Annotate | Download (1.5 KB)
1 |
This is the game plan if anyone is interested. |
---|---|
2 |
Each robot will go for an unspecified amount of time, |
3 |
accruing encoder, BOM, and IR data. The IR and encoders |
4 |
will be used to generate heat maps for each robot, starting |
5 |
at the center, decreasing in certainty (i.e. the amount of |
6 |
'color' put on during a time unit will be spread over a greater |
7 |
area) over time according to the physical limitations of |
8 |
the measuring devices. |
9 |
|
10 |
When the uncertainty gets too high, a new map will be allocated for |
11 |
the robot. |
12 |
|
13 |
The BOM data will be used to orient individual maps to one another. |
14 |
Along with physical limitations on the size of the environment, and |
15 |
hopefully some brainstorming to do with triangularization and what not, |
16 |
the heat maps will become increasingly well oriented. |
17 |
|
18 |
Finally, after a currently unspecified amount of time, the maps which have |
19 |
been properly oriented (this could probably just be all of them due to the nature |
20 |
of the heat maps,) will be summed up into one meta map! Metamap will undergo some |
21 |
signal processing, probably in the form of a second order gradient, to get a map |
22 |
of magnitudes, which we will try to further reduce to lines. These lines are the |
23 |
walls of the environment. |
24 |
|
25 |
In time, it is possible that this "unspecfied amount of time" could become a continuous |
26 |
computation, but I haven't put much thought into how this will scale. It will almost |
27 |
definitely depend upon the chosen resolution of the heat map. |
28 |
|
29 |
I do want people working with me, this should be a lot of fun, |
30 |
I think this is a solid method of mapping. |