Can I render/reset an environment based on a specific observation?

[cchristofi]

I am conducting an experiment where I test different starting points for the completion of the task and I want to visualize these starting points afterwards. Is there any way to visualize these points by using the observation?

[qgallouedec]

Hi, sorry for the delay. The easiest way would be to render directly when the starting point is generated:

import gymnasium as gym
from PIL import Image

import panda_gym

env = gym.make("PandaPickAndPlace-v3", render_mode="rgb_array", renderer="OpenGL")

# Render the starting point
env.reset()
Image.fromarray(env.render()).save("starting_point.png")

# Interract
for _ in range(50):
    env.step(env.action_space.sample())

env.close()

If, for some reasons, you really want to do it afterward, you can use the save/restore state feature:

import gymnasium as gym
from PIL import Image

import panda_gym

env = gym.make("PandaPickAndPlace-v3", render_mode="rgb_array", renderer="OpenGL")

# Reset and save the initial state
env.reset()
state_id = env.save_state()

# Interract
for _ in range(50):
    env.step(env.action_space.sample())

# Restore initial state and render it
env.restore_state(state_id)
Image.fromarray(env.render()).save("starting_point.png")

env.close()