Bugfix: Reset recurrent state after episode termination during evaluation in RSL-RL framework

Author: bikcrum

scripts/reinforcement_learning/rsl_rl/play.py

@@ -185,7 +185,9 @@ def main(env_cfg: ManagerBasedRLEnvCfg | DirectRLEnvCfg | DirectMARLEnvCfg, agen
             # agent stepping
             actions = policy(obs)
             # env stepping
-            obs, _, _, _ = env.step(actions)
+            obs, _, dones, _ = env.step(actions)
+            # reset recurrent states for episodes that have terminated
+            policy_nn.reset(dones)
         if args_cli.video:
             timestep += 1
             # Exit the play loop after recording one video

source/isaaclab/config/extension.toml

@@ -1,7 +1,7 @@
 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.47.2"
+version = "0.47.3"
 
 # Description
 title = "Isaac Lab framework for Robot Learning"

source/isaaclab/docs/CHANGELOG.rst

@@ -1,6 +1,15 @@
 Changelog
 ---------
 
+0.47.3 (2025-10-26)
+~~~~~~~~~~~~~~~~~~~~
+
+Changed
+^^^^^^^
+
+* Fixed an issue in recurrent policy evaluation in RSL-RL framework where the recurrent state was not reset after an episode termination.
+  
+
 0.47.2 (2025-10-22)
 ~~~~~~~~~~~~~~~~~~~