Skip to content

A 2D simulator for RoboMaster AI Challenge, the environment for the training of reinforcement learning. The simulation is able to achieve efficient collision detection and significantly accelerated reinforcement learning

Notifications You must be signed in to change notification settings

Yazhou-Z/ICRA-simulator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

61 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Simulator😄

This is a 2D simulator for RoboMaster AI Challenge, the environment for the training of reinforcement learning. The simulation is able to achieve efficient collision detection and significantly accelerated reinforcement learning.

action

observation

cars

acts

punish&bonus

map_parameter

Real

053900aa88262dec18e18d70dea4fb4

action

Called orders in kernel.

        self.action_space = spaces.Box(high = 1, low = -1, shape = (8,),dtype = int)

In function orders_to_acts, np.clip is used.

for i in range(4):
	self.orders[i] = np.clip(self.orders[i], 0, 1)
名称 范围 解释 手控按键
0 x -1~1 -1:后退,0:不动,1:前进[3] s/w
1 y -1~1 -1:左移,0:不动,1:右移 q/e
2 rotate -1~1 底盘,-1:左转,0:不动,1:右转 a/d
3 yaw -1~1 云台,-1:左转,0:不动,1:右转 b/m
4 shoot 0~1 是否射击,0:否,1:是 space
5 supply 0~1 时候触发补给,0:否,1:是 f
6 shoot_mode 0~1 射击模式,0:单发,1:连发 r
7 auto_aim 0~1 是否启用自瞄,0:否,1:是 n

observation

(有待完善

        self.observation_space = spaces.Box(low = -180.0, high = 2000.0, shape = (17, ), dtype = np.float32)
num name min max
[0, 14] car_info -180.0 800.0
15 time 0.0 180.0
[16, 19] observ 0.0 1.0

cars

floatshape(car_mun,15),car_num为机器人的数量:

引索 名称 类型 范围 解释
0 owner int 0~1 队伍,0:红方,1:蓝方
1 x float 0~800 x坐标[0]
2 y float 0~500 y坐标
3 angle float -180~180 底盘绝对角度[1]
4 yaw float -90~90 云台相对底盘角度
5 heat int 0~ 枪口热度
6 hp int 0~2000 血量
7 freeze_time int 【已删除】
8 is_supply bool 【已删除】
9 can_shoot bool 0~1 决策频率高于出弹最高频率(10Hz)
10 bullet int 0~ 剩余子弹量
11 stay_time int 【已删除】
12 wheel_hit int 0~ 轮子撞墙的次数
13 armor_hit int 0~ 装甲板撞墙的次数
14 car_hit int 0~ 轮子或装甲板撞车的次数

detect&vision&observ

shape: (car_num, car_num)

detect

get_lidar_vision

vision

get_camera_vision

observ

observ: detect || vision

#	   0  1  2  3
detect = [[0, 1, 0, 0], # 0
          [0, 0, 1, 1], # 1
          [0, 0, 0, 0], # 2
          [1, 0, 0, 0]] # 3

表示:

0号车能检测到1号车

1号车能检测到2号车和3号车

2号车检测不到任何车

3号车能检测到0号车

acts

acts是一个较底层的action,类型floatshape为:(car_num,8)

引索1 名称 解释
0 rotate_speed 底盘旋转速度
1 yaw_speed 云台旋转速度
2 x_speed 前进后退速度
3 y_speed 左右平移速度
4 shoot 是否发射
5 shoot_mutiple 是否连发
6 supply 是否触发补给
7 auto_aim 是否自动瞄准

punish&bonus

class Move_Shoot:
    def __init__(self, area, time, activation):
       self.area = area
       self.time = time
       self.activation = activation
class RefereeSystem:
    move = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
    shoot = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
    red_hp = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
    blue_hp = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
    red_bullet = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
    blue_bullet = Move_Shoot(np.zeros(4, dtype='float32'), 0, None)
	def __init__(self, special_area, time, cars):
	def checkZone(self, car, bouns):
	def getMobility(self,car):
	def getShootabiliy(self, car):
	def _reset_bufzone(self):
	def update(self):

map_parameter

name 解释
length int 808 地图长度
width int 448 地图宽度
special_area float (8, 4) supply & punish area
areas float (4, 4) start areas
barriers float (9, 4) 障碍物的位置信息

以地图左上角为原点

名称 范围 解释
0 border_x0 0~808 左边界
1 border_x1 0~808 右边界
2 border_y0 0~448 上边界
3 border_y1 0~448 下边界

special_area

shape (6, 4)

area1 ,area2 centrosymmetric

reset randomly

area1 area2
red_hp blue_hp
red_bullet blue_bullet
move shoot

areas

start_area, shape (4, 4)

num team
0, 1 red
2, 3 blue

barriers

shape(9,4)

name
0 barrier_horizontal_short
1 barrier_horizontal_short
2 barrier_horizontal_tall
3 barrier_horizontal_tall
4 barrier_horizontal_tall
5 barrier_horizontal_tall
6 barrier_vertical
7 barrier_vertical
8 barrier_small

About

A 2D simulator for RoboMaster AI Challenge, the environment for the training of reinforcement learning. The simulation is able to achieve efficient collision detection and significantly accelerated reinforcement learning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages