Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

MyoChallenge Manipulation Track Metric #213

Merged
merged 5 commits into from
Aug 27, 2024
Merged

MyoChallenge Manipulation Track Metric #213

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
b1ec80a
adding evaluation metric for manipulation
cherylwang20 Aug 26, 2024
877f306
Merge branch 'MyoHub:main' into myochal-mani-p2
cherylwang20 Aug 26, 2024
ad5aebc
hotfix for linux compilation error
cherylwang20 Aug 26, 2024
710b0fe
Merge branch 'myochal-mani-p2' of https://github.com/Balint-H/myosuit…
cherylwang20 Aug 26, 2024
3281da5
Update bimanual_v0.py
cherylwang20 Aug 26, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 11 additions & 11 deletions myosuite/envs/myo/myochallenge/bimanual_v0.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -111,7 +111,7 @@ def _setup(self,

# check whether the object experience force over max force
self.over_max = False
self.max_force = max_force
self.max_force = 0

self.touch_history = []

Expand Down Expand Up @@ -177,7 +177,8 @@ def get_obs_dict(self, sim):

current_force = sim.data.sensordata[0]
if current_force > self.max_force:
self.over_max = True
self.max_force = current_force
obs_dict['max_force'] = np.array([self.max_force])

obs_vec = self._obj_label_to_obs(touching_objects)
obs_dict["touching_body"] = obs_vec
Expand Down Expand Up @@ -236,7 +237,8 @@ def get_reward_dict(self, obs_dict):
("pass_err", pass_dist + np.log(pass_dist + 1e-3)),
# Must keys
("sparse", 0),
("solved", self.check_solve(goal_dis)),
("goal_dist", goal_dis),
("solved", goal_dis < self.proximity_th),
("done", False),
)
)
Expand All @@ -258,12 +260,6 @@ def step(self, a, **kwargs):
- self.sim.model.actuator_ctrlrange[robotic_act_ind, 0]) / 2.0)
return super().step(processed_controls, **kwargs)

def check_solve(self, goal_dis):
if goal_dis > 0.01:
return False
if self.over_max == True:
return False
return True

def get_metrics(self, paths, successful_steps=5):
"""
Expand All @@ -280,15 +276,19 @@ def get_metrics(self, paths, successful_steps=5):
num_success += 1
score = num_success / num_paths

times = np.mean([np.round(p['env_infos']['obs_dict']['time'][-1], 2) for p in paths])
times = np.mean([np.round(p['env_infos']['obs_dict']['time'][-1], 5) for p in paths])
max_force = np.mean([np.round(p['env_infos']['obs_dict']['max_force'][-1], 5) for p in paths])
goal_dist = np.mean([np.mean(p['env_infos']['rwd_dict']['goal_dist']) for p in paths])

# average activations over entire trajectory (can be shorter than horizon, if done) realized
effort = -1.0 * np.mean([np.mean(p['env_infos']['rwd_dict']['act_reg']) for p in paths])
effort = -1.0 * np.mean([np.mean(p['env_infos']['rwd_dict']['act']) for p in paths])

metrics = {
'score': score,
'time': times,
'effort': effort,
'peak force': max_force,
'goal dist': goal_dist,
}
return metrics

Expand Down
Loading