dodox/fancy_gym
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

after deadline

Browse Source
This commit is contained in:
Onur
2022-06-21 17:15:01 +02:00
parent c47845c0dd
commit 7bd9848c31
8 changed files with 227 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
alr_envs/alr/mujoco/beerpong/beerpong.py
+25 -1
View File
@@ -185,6 +185,10 @@ class ALRBeerBongEnv(MujocoEnv, utils.EzPickle):
def dt(self):
return super(ALRBeerBongEnv, self).dt*self.repeat_action
class ALRBeerBongEnvFixedReleaseStep(ALRBeerBongEnv):
def __init__(self, frame_skip=1, apply_gravity_comp=True, noisy=False, rndm_goal=False, cup_goal_pos=None):
super().__init__(frame_skip, apply_gravity_comp, noisy, rndm_goal, cup_goal_pos)
self.release_step = 62 # empirically evaluated for frame_skip=2!
class ALRBeerBongEnvStepBasedEpisodicReward(ALRBeerBongEnv):
def __init__(self, frame_skip=1, apply_gravity_comp=True, noisy=False, rndm_goal=False, cup_goal_pos=None):
@@ -206,6 +210,25 @@ class ALRBeerBongEnvStepBasedEpisodicReward(ALRBeerBongEnv):
# internal steps and thus, the observation also needs to be set correctly
return ob, reward, done, infos
# class ALRBeerBongEnvStepBasedEpisodicReward(ALRBeerBongEnv):
# def __init__(self, frame_skip=1, apply_gravity_comp=True, noisy=False, rndm_goal=False, cup_goal_pos=None):
# super().__init__(frame_skip, apply_gravity_comp, noisy, rndm_goal, cup_goal_pos)
# self.release_step = 62 # empirically evaluated for frame_skip=2!
#
# def step(self, a):
# if self._steps < self.release_step:
# return super(ALRBeerBongEnvStepBasedEpisodicReward, self).step(a)
# else:
# sub_ob, sub_reward, done, sub_infos = super(ALRBeerBongEnvStepBasedEpisodicReward, self).step(np.zeros(a.shape))
# reward = sub_reward
# infos = sub_infos
# ob = sub_ob
# ob[-1] = self.release_step + 1 # Since we simulate until the end of the episode, PPO does not see the
# # internal steps and thus, the observation also needs to be set correctly
# return ob, reward, done, infos
class ALRBeerBongEnvStepBased(ALRBeerBongEnv):
def __init__(self, frame_skip=1, apply_gravity_comp=True, noisy=False, rndm_goal=False, cup_goal_pos=None):
super().__init__(frame_skip, apply_gravity_comp, noisy, rndm_goal, cup_goal_pos)
@@ -259,7 +282,8 @@ class ALRBeerBongEnvStepBased(ALRBeerBongEnv):
if __name__ == "__main__":
# env = ALRBeerBongEnv(rndm_goal=True)
# env = ALRBeerBongEnvStepBased(frame_skip=2, rndm_goal=True)
env = ALRBeerBongEnvStepBasedEpisodicReward(frame_skip=2, rndm_goal=True)
# env = ALRBeerBongEnvStepBasedEpisodicReward(frame_skip=2, rndm_goal=True)
env = ALRBeerBongEnvFixedReleaseStep(frame_skip=2, rndm_goal=True)
import time
env.reset()
env.render("human")