Last active
December 6, 2021 05:55
-
-
Save ugo-nama-kun/a13ec948ad494aa443f3458bda909a1c to your computer and use it in GitHub Desktop.
How to use the marker util of mujoco-py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| import numpy as np | |
| import gym | |
| from gym.envs.mujoco import mujoco_env | |
| from mujoco_py.generated import const | |
| from scipy.spatial.transform import Rotation | |
| """ Marker types in const | |
| GEOM_PLANE = 0 | |
| GEOM_HFIELD = 1 | |
| GEOM_SPHERE = 2 | |
| GEOM_CAPSULE = 3 | |
| GEOM_ELLIPSOID = 4 | |
| GEOM_CYLINDER = 5 | |
| GEOM_BOX = 6 | |
| GEOM_MESH = 7 | |
| NGEOMTYPES = 8 | |
| GEOM_ARROW = 100 | |
| GEOM_ARROW1 = 101 | |
| GEOM_ARROW2 = 102 | |
| GEOM_LINE = 103 | |
| GEOM_SKIN = 104 | |
| GEOM_LABEL = 105 | |
| GEOM_NONE = 1001 | |
| """ | |
| def euler2mat(euler, degrees=True): | |
| r = Rotation.from_euler('xyz', euler, degrees=degrees) | |
| return r.as_matrix() | |
| class SwimmerEnvRV(gym.envs.mujoco.SwimmerEnv): | |
| def render(self, **kwargs): | |
| if self.viewer: | |
| x, y = self.sim.data.qpos[:2] | |
| # Draw a sphere marker | |
| self.viewer.add_marker(pos=np.array([x, y, 0]), # Position | |
| label=" ", # Text beside the marker | |
| type=const.GEOM_SPHERE, # Geomety type | |
| size=(0.1, 0.1, 0.1), # Size of the marker | |
| rgba=(1, 0, 0, 1)) # RGBA of the marker | |
| # Draw an arrow | |
| # from: https://github.com/openai/mujoco-py/issues/31 | |
| self.viewer.add_marker(pos=np.array([0, 0, 1]), # Root Position | |
| label="Arrow", # Text beside the marker | |
| mat=euler2mat([90, 0, 90]), # Direction | |
| type=const.GEOM_ARROW, # Geomety type | |
| size=(0.1, 0.1, 2), # Size of the marker | |
| rgba=(1, 0, 1, 0.3)) # RGBA of the marker | |
| # Default swimmer renderer method | |
| super(SwimmerEnvRV, self).render() | |
| # Run | |
| env = SwimmerEnvRV() | |
| while True: | |
| env.step(env.action_space.sample()) | |
| env.render() |
All marker configuration is done by using attributes of mjvGeom.
https://mujoco.readthedocs.io/en/latest/APIreference.html#mjvGeom
How to add a mesh
(Assuming you have .stl file, Banana.stl by converting from .fbx using Meshlab)
(to do exactly same thinng below, you will need additional rescaling the banana by x0.001)
- Load the mesh file (.stl) in the xml (banana.xml) in the asset region.
- In python flile, set type option of add_maker to "GEOM_MESH". Set dataid to the right value (dataid=0 in this case).
- And run!
Example banana.xml
<?xml version="1.0" ?>
<mujoco>
<asset>
<mesh file="Banana.stl" name="banana"></mesh>
</asset>
<worldbody>
<body name="floor" pos="0 0 0.025">
<geom size="1.0 1.0 0.02" rgba="0 1 0 1" type="box"/>
</body>
</worldbody>
</mujoco>Example banana_mujoco.py
import mujoco_py
from mujoco_py.generated import const
import os
import numpy as np
import time
xml_path = 'banana.xml'
model = mujoco_py.load_model_from_path(xml_path)
sim = mujoco_py.MjSim(model)
viewer = mujoco_py.MjViewer(sim)
while True:
sim.step()
t = time.time()
for i in range(5):
for j in range(5):
viewer.add_marker(pos=np.array([0.5 * np.sin(t) + 0.2 * i, 0.5 * np.cos(t) + 0.2 * j, 0.5]),
type=const.GEOM_MESH,
# size=(0.1, 0.1, 0.1), <-- Caution. this doesn't work...
rgba=(1, 1, 0, 1),
dataid=0)
viewer.render()You will see this.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
According to MuJoCo forum, update_scene resets all virtual geoms, i.e. all markers. Unfortunately, mujoco-py does not clear the _markers list in the rgb_array (offscreen=True) mode.
Calling
del viewer._markers[:]afterrender()fixes the issue.Copied from:
openai/mujoco-py#423