Difference between revisions of "Wall Smoothing/Melee Wall Smoothing"
m |
m |
||
| Line 5: | Line 5: | ||
For instance, [[Neuromancer]], the current melee king, marks destinations outside the “safe area” as dangerous, meanwhile he selects destination points on a circle around the robot, which is similar to wall stick in orbital wall smoothing. Therefore its movement looks just like orbital wall smoothing near wall — but its implementation is completely different. | For instance, [[Neuromancer]], the current melee king, marks destinations outside the “safe area” as dangerous, meanwhile he selects destination points on a circle around the robot, which is similar to wall stick in orbital wall smoothing. Therefore its movement looks just like orbital wall smoothing near wall — but its implementation is completely different. | ||
| − | But Wall Smoothing in Go-to style movement is much difficult, e.g. [[GoTo Surfing]] and [[Minimal Risk]]. The real issue is that you can’t adjust your heading gradually, as destinations are selected in adance; you can neither hack your go-to algorithm and add some traditional wall smoothing, as that would make you go to somewhere else, not the evaluated one. Selecting points in “safe area” works, but it still hit the wall (ocationally). But [[Wall Smoothing/Exact]] is capable of this situation, as it simulates exact robocode physics and gives you the max capabilities near wall, therefore can be used with go-to algorithm to create a wall-smoothed go-to algorithm. (I will soon post the code | + | But Wall Smoothing in Go-to style movement is much difficult, e.g. [[GoTo Surfing]] and [[Minimal Risk]]. The real issue is that you can’t adjust your heading gradually, as destinations are selected in adance; you can neither hack your go-to algorithm and add some traditional wall smoothing, as that would make you go to somewhere else, not the evaluated one. Selecting points in “safe area” works, but it still hit the wall (ocationally). But [[Wall Smoothing/Exact]] is capable of this situation, as it simulates exact robocode physics and gives you the max capabilities near wall, therefore can be used with go-to algorithm to create a wall-smoothed go-to algorithm. (I will soon post the code ;) ) |
Revision as of 01:20, 25 September 2017
| This article is a stub. You can help RoboWiki by expanding it. |
In melee, you don’t have something called “lateral direction”, therefore when facing the wall, you have to choose it on the fly. This invloves the integration of wall smoothing and movement selection system —hitting the wall is also considered as some “danger”.
For instance, Neuromancer, the current melee king, marks destinations outside the “safe area” as dangerous, meanwhile he selects destination points on a circle around the robot, which is similar to wall stick in orbital wall smoothing. Therefore its movement looks just like orbital wall smoothing near wall — but its implementation is completely different.
But Wall Smoothing in Go-to style movement is much difficult, e.g. GoTo Surfing and Minimal Risk. The real issue is that you can’t adjust your heading gradually, as destinations are selected in adance; you can neither hack your go-to algorithm and add some traditional wall smoothing, as that would make you go to somewhere else, not the evaluated one. Selecting points in “safe area” works, but it still hit the wall (ocationally). But Wall Smoothing/Exact is capable of this situation, as it simulates exact robocode physics and gives you the max capabilities near wall, therefore can be used with go-to algorithm to create a wall-smoothed go-to algorithm. (I will soon post the code ;) )
