KR: Is there something about this project, or any other that makes it inherently NP?
The cross map was interesting. Where do you place the lights? What if you only had 1 light?
Dfed: Use 1 light to draw mosquitoes into 1 region, then put the collector in that region
Dan: If you put a collector at the very end of a line, then you may see it more likely that a mosquito will hit it directly.
Elba: It's very important to make sure that we are using as much of the light as possible (minimal shadow, maximal shine in bounds)
KR: Take the group2board2 map, and think about it with 3 and 10 lights. Manually, come up with the best 3 light and 10 light arrangements.
How can you prune the search space?
Eva: Find only the interesting parts.
EPK: How do you determine how much of the board is empty space? Could you just use a random walk?
KR: What kind of representation do you use?
Blake: Create "areas" as units of reasoning
Jesse: Close off the small gaps, then use that to define the areas
Zach: You couldn't just go end-to-end, you have to do line-to-line
Dfed: We store for each meter of granularity, what lines are closest and how far
David: You can observe the intended direction of flow through a channel, by looking at which directions the obstacles are
KR: In a map with no walls, what's the most efficient way to collect the mosquitoes
Deliverables:
No more maps. A few manual deliverables:
A 3 and 10 light version of an empty map, and a 3 and 10 light version of the group2board2.
Next week, come up with improved players that do something interesting and is informed by your thinking from the manual investigations.
No comments:
Post a Comment