From fa18f607fa252cfe53f5f0ffd702f467429c82a8 Mon Sep 17 00:00:00 2001
From: tactino <18781106300@163.com>
Date: Tue, 29 Sep 2026 14:06:33 -0400
Subject: [PATCH] robomimic-v1: the state-only metadata (square, lift, can,
 transport) builds - camera_names is optional, and image observations are on
 only with cameras

---
 src/plugrl_env_client/envs/robomimic/robomimic_env.py | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/plugrl_env_client/envs/robomimic/robomimic_env.py b/src/plugrl_env_client/envs/robomimic/robomimic_env.py
index 9380a76..6720ce7 100644
--- a/src/plugrl_env_client/envs/robomimic/robomimic_env.py
+++ b/src/plugrl_env_client/envs/robomimic/robomimic_env.py
@@ -135,18 +135,25 @@ def __init__(
             )
         )
 
+        # The `-img` metadata names the cameras whose images go into the
+        # observation; the state-only metadata (`square`, `lift`, ...) names
+        # none. Reading the key unconditionally raised KeyError for every one
+        # of those, and turning image observations on without cameras would
+        # have robosuite add its default `agentview` image to the states. The
+        # agentview frame below is rendered either way.
+        camera_names = list(env_meta["env_kwargs"].get("camera_names", []))
         env = _env_utils.create_env_from_metadata(
             env_meta=env_meta,
             render=False,
             render_offscreen=True,
-            use_image_obs=True,
+            use_image_obs=bool(camera_names),
         )
         assert isinstance(env, robomimic.envs.env_robosuite.EnvRobosuite)
         self.env = env
 
         self.env.env.hard_reset = False
 
-        self.image_keys = env_meta["env_kwargs"]["camera_names"]
+        self.image_keys = camera_names
         self.task = env_meta["env_name"]
         self.agentview_image_size = config.agentview_image_size