Difference between revisions of "Talk:GuessFactor Targeting (traditional)"
| (4 intermediate revisions by 4 users not shown) | |||
| Line 11: | Line 11: | ||
* [[Bin Smoothing]] is to affect all (or some) bins depending on how close they are to the "exact" bin. If you are to update bin #12, then bins 11 and 13 are very close and get some data added to it too. A simple function may be <code>bin[i] += 1 / (1 + Math.pow(index - i, 2));</code>. You can see that the values added are 1, 1/2, 1/5, 1/10, ... | * [[Bin Smoothing]] is to affect all (or some) bins depending on how close they are to the "exact" bin. If you are to update bin #12, then bins 11 and 13 are very close and get some data added to it too. A simple function may be <code>bin[i] += 1 / (1 + Math.pow(index - i, 2));</code>. You can see that the values added are 1, 1/2, 1/5, 1/10, ... | ||
* For weighting you can weight non-firing waves less than firing waves, where a firing wave is a wave that was shot at the same time as a real bullet was. Exact weights are unknown. | * For weighting you can weight non-firing waves less than firing waves, where a firing wave is a wave that was shot at the same time as a real bullet was. Exact weights are unknown. | ||
| − | * And you should probably add [[Rolling Average]] as well, it means dropping old data as new data comes in. It's implementation is straight forward with [[Paul Evans]]'s magic formula. | + | * And you should probably add [[Rolling Average]] as well, it means dropping old data as new data comes in. It's implementation is straight forward with [[User:Paul Evans|Paul Evans]]'s magic formula. |
Hope this helps. --[[User:Zyx|zyx]] 16:52, 15 March 2009 (UTC) | Hope this helps. --[[User:Zyx|zyx]] 16:52, 15 March 2009 (UTC) | ||
| + | |||
| + | See [[GuessFactor Targeting Tutorial]]. For [[Rolling Averages]], see its page! I have normalize it to work with that tutorial. » <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> » 01:31, 16 March 2009 (UTC) | ||
| + | |||
| + | == Considering Multiple Waves == | ||
| + | |||
| + | Have any of you tried or considered trying some version of the inverse of surfing multiple waves? It's something I am planning on trying out with my gun one of these days. I haven't worked out how it will work yet, but it seemed like a worthwhile idea to explore. -- [[User:Skotty|Skotty]] 23:52, 19 September 2011 (UTC) | ||
| + | |||
| + | I'm not sure there is something analogous for targeting really. The point of multiple waves has to do with the pathfinding nature of the movement problem, which doesn't exist in the targeting problem. If there's something I'm missing, I'd be quite interested, but I don't really see a way in which there could be something equivalent. --[[User:Rednaxela|Rednaxela]] 00:22, 20 September 2011 (UTC) | ||
| + | |||
| + | I agree, multiple waves exist to help the robot navigate the movement problem, to better determine locations and their future risk. As the targeting problem is only for the most part a 1 dimensional problem for a normal gf gun, that is what gf to fire at. I don't see there as being a way to translate one to the other in a general fashion that would be much help to targeting. I cannot really think of a way to do multiple wave targeting, let alone one that is useful. Since each wave is handled in a discrete fashion without need for considering other waves. Unless of course you want to predict your own targeting ahead to take that into consideration while targeting. But that is somewhat silly, as you would have to predict the enemy movement to do so, which is what you are already doing. — <span style="font-family: monospace">[[User:Chase-san|Chase]]-[[User_talk:Chase-san|san]]</span> 00:54, 20 September 2011 (UTC) | ||
Latest revision as of 01:54, 20 September 2011
Archived discussion pages
Advanced Stuffs
How do I implement bin smoothing, weighting and other stuff? I don't get it. I used: p.nfshost.com/GFTTutorial.shtml to teach me Guess Factor Targeting. Sorry, I cant put it in a link, the captcha image won't load.
I don't think there is a rule but I'll try to give you some hints.
- Bin Smoothing is to affect all (or some) bins depending on how close they are to the "exact" bin. If you are to update bin #12, then bins 11 and 13 are very close and get some data added to it too. A simple function may be
bin[i] += 1 / (1 + Math.pow(index - i, 2));. You can see that the values added are 1, 1/2, 1/5, 1/10, ... - For weighting you can weight non-firing waves less than firing waves, where a firing wave is a wave that was shot at the same time as a real bullet was. Exact weights are unknown.
- And you should probably add Rolling Average as well, it means dropping old data as new data comes in. It's implementation is straight forward with Paul Evans's magic formula.
Hope this helps. --zyx 16:52, 15 March 2009 (UTC)
See GuessFactor Targeting Tutorial. For Rolling Averages, see its page! I have normalize it to work with that tutorial. » Nat | Talk » 01:31, 16 March 2009 (UTC)
Considering Multiple Waves
Have any of you tried or considered trying some version of the inverse of surfing multiple waves? It's something I am planning on trying out with my gun one of these days. I haven't worked out how it will work yet, but it seemed like a worthwhile idea to explore. -- Skotty 23:52, 19 September 2011 (UTC)
I'm not sure there is something analogous for targeting really. The point of multiple waves has to do with the pathfinding nature of the movement problem, which doesn't exist in the targeting problem. If there's something I'm missing, I'd be quite interested, but I don't really see a way in which there could be something equivalent. --Rednaxela 00:22, 20 September 2011 (UTC)
I agree, multiple waves exist to help the robot navigate the movement problem, to better determine locations and their future risk. As the targeting problem is only for the most part a 1 dimensional problem for a normal gf gun, that is what gf to fire at. I don't see there as being a way to translate one to the other in a general fashion that would be much help to targeting. I cannot really think of a way to do multiple wave targeting, let alone one that is useful. Since each wave is handled in a discrete fashion without need for considering other waves. Unless of course you want to predict your own targeting ahead to take that into consideration while targeting. But that is somewhat silly, as you would have to predict the enemy movement to do so, which is what you are already doing. — Chase-san 00:54, 20 September 2011 (UTC)
