Robo Home | Rednaxela | Changes | Preferences | AllPages

I'm not sure if this idea is new or not, however I was just thinking that it might be interesting to try a "surfing rambot" of sorts, or at least a way for a close range bot to get close while not being a sitting duck to 'modern' targeting. So, what I'm thinking, is adapting TrueSurfing to a mode that approaches the enemy. Instead of evaluating the danger of stopping, going one way, or going the other way, it would evaluate the danger of moving heading towards the enemy, towards one perpendicular, and towards the other perpendicular. When it is close enough to perpendicular (when the following becomes significantly more efficent), the option of turning towards the other perpendicular becomes reversing into the other perpendicular. I'm thinking this would be an interesting way to get a bot up close to the enemy with relative safety. It might not be a bad idea to use plain TrueSurfing when the bot is at it's preferred range and use this variation to get to the range it prefers. Another thought which might be good at least for a rambot, is instead of directly chasing the opponent, moving towards where it would expect to intercept the opponent given movement predictions about it, that might help counter dive-in protection as well as making the interception faster. I'm not sure this would the best tactic for most battles, but I think a relatively advanced surfing rambot like this might create a ProblemBot for quite a few of the bots that prefer longer range battle. Another crackpot idea of mine :) -- Rednaxela

The way Dookious adjusts its distance is similar to what you describe: if it is far from its desired distance, it will dive (or move away ... climb?) at a sharper angle, but if it's at its desired distance (or close to it), it will move perpendicular (or close to it). It still evaluates clockwise, stop, and counter clockwise, as normal, but it applies an attack angle (offset from perpendicular) formula to those options based on current vs desired distance. If the PointBlankCompetition gets going again, that would be an excellent place to test this idea. =) -- Voidious

Actually the recent activity on the PointBlankCompetition page is what made me think about this. I was also thinking about one thing that made my old bot Mervin not-so-horrible in actual battle despite it's simplicity. One on Mervin's movement patterns was an "Intercept" movement, which would try to intercept where the enemy was moving, using the same accurate circular predictor as the gun. Against bots with strong perpendicular movement (i.e. Reiko or WaveSurfing style), the result of moving to intecept, would have meant that the enemy's guessfactors would have been "copied" in a sense (keep in mind the actual angle would be "compressed" by how it's a diving movement instead of perpendicular), thus leading a diving pattern that weaved around bullets very effectively so long as I was firing at a similar time as the opponent. I'm pretty sure this explains why many high level bots seemed to have some issues hitting Gwynfor when it was diving. I think this "Dive Surfing" thought would essentially produce the strong movement I was seeing in Mervin without relying on firing at the same time or strong enemy movement. -- Rednaxela

A simple drone I put together (see [1]) implements a very straightforward "dive bombing" movement, ending up circling its target. For a simple drone with no real targeting or bullet power logic, and developed just to be a "drone" for target practice, I am pleasantly surprised at both how well it does (against the sample bots and other example bots, like the AntiGravityBot?) and how easily the really good bots (like Yngwie etc) really whack it. -- Excession

Robo Home | Rednaxela | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited June 5, 2008 1:34 EST by Excession (diff)