Commit Graph

12 Commits

Author SHA1 Message Date
cfccd12820 V3.1.1 优化代码
优化可读性与规范化命名方式
2023-06-30 18:30:12 +09:00
64e477d6c2 V3.1 控制模式完成
完成控制模式中UI,解决所有逻辑问题。控制模式于Unity中正常运行,可正常判断结束。
2023-06-29 15:18:10 +09:00
ff094aaba5 V3.0 追加控制模式,改进代码
追加控制模式,实现鼠标在地图中的坐标映射
追加控制UI
Todo:
游戏流程修改
游戏State对应修改
2023-04-09 23:35:38 +09:00
faa4e7b065 V2.8 移动算法更新
过去X与Y可同时达到maxspeed致使速度会超过maxspeed,本次修正了该错误。
追加移动Penalty。
2023-02-15 21:07:10 +09:00
9b2ba7fb46 V2.5 Change reward function
Add chart Toggle to turn on reward chart real time in game.
Add enemy num changer, now can change enemy num in game.
Change facing to enemy reward function, while facing center line far from enemy will not get reward any more.
2022-12-10 09:11:42 +09:00
f9b806de02 V2.4 Add new reward function
add new reward function in attack mode
calculate distance between closest enemy and facing center line.
let agent could spawn in whole map area.
add penalty while mouseX is moving.
2022-12-08 06:27:49 +09:00
74c85c4296 V2.1 Easier game!
Build-ParallelEnv-Target-OffPolicy-SingleStack-SideChannel-EndReward-Easy-V2.1
move remain time position after target state
easier game reward get, while watching or shoot at target area u get reward(attack mode), watching enemy u also get reward(Freemode)
2022-12-08 06:27:49 +09:00
fa90ac3bb6 V2.1 Add spin penalty and reward adjust
Build-ParallelEnv-Target-OffPolicy-SingleStack-SideChannel-EndReward-Easy-V2.1
add spin penalty while agent keep spin will give a penalty reward.
lower Go target in area reward.
2022-12-08 06:27:49 +09:00
0d60c857b7 V2.0 Bug Fix, scene added
fix scene only spawn in a small area in center.
delete room type attack target.
make no more enemy spawn in go target.
free mode enemy number change to 3.
2022-12-08 06:27:49 +09:00
8570a5ba5e bug fix
fix target attack always win in first round.
2022-12-01 19:52:52 +09:00
9585845ba2 Side channel Added
add side channel to let python side know which target got win or lose.
fix update time bug. may cause double gameover check.(got another lose after reset the game.)
2022-11-30 06:39:56 +09:00
25eac00c53 Target & dynamic reward System V1.0
太多了...总之能用了,就这样吧,也懒得用英文写了(;´Д`)
ParameterContainer用于管理所有的Reward和Parameter.
TargetController用于生成目标,然后再让SceneBlockContainer实际生成目标块,并且兼顾reward计算功能和目标观察结果的获取.
EnemyContainer用于生成和删除敌人.
SceneBlockContainer用于生成和删除目标块.

States用于管理HP.
SceneBlock用于管理目标块的...一堆东西,比如目标大小,目标区内人数等和目标所属状态等.
2022-11-29 07:54:08 +09:00