Configure option_learning agent for domino task, tighten release_z range

Author: yichao-liang

predicators/ground_truth_models/skill_factories/place.py

@@ -42,7 +42,7 @@
     ("target_x (world x position for placement)", 0.4, 1.1),
     ("target_y (world y position for placement)", 1.1, 1.6),
     ("target_yaw (placement orientation in radians)", -np.pi, np.pi),
-    ("release_z (world z height to open gripper)", 0.4, 0.6),
+    ("release_z (world z height to open gripper)", 0.5, 0.6),
 ]
 
 

scripts/configs/predicatorv3/predicator_v3.yaml

@@ -3,14 +3,14 @@
 #    python scripts/local/launch_simp.py -c mara_bench.yaml
 ---
 APPROACHES:
-  human_interaction:
-    NAME: "human_interaction"
-    FLAGS:
-      human_interaction_approach_use_scripted_option: True
-      human_interaction_approach_use_all_options: True
-      scripted_option_dir: "scripted_option_policies"
-      script_option_file_name: "domino2.txt"
-      skill_phase_use_motion_planning: True
+  # human_interaction:
+  #   NAME: "human_interaction"
+  #   FLAGS:
+  #     human_interaction_approach_use_scripted_option: True
+  #     human_interaction_approach_use_all_options: True
+  #     scripted_option_dir: "scripted_option_policies"
+  #     script_option_file_name: "domino2.txt"
+  #     skill_phase_use_motion_planning: True
   # oracle:
   #   NAME: "oracle_process_planning"
   #   FLAGS:
@@ -25,17 +25,17 @@ APPROACHES:
   #     terminate_on_goal_reached_and_option_terminated: True
   #     agent_planner_isolate_test_session: True
   #     agent_sdk_use_docker_sandbox: True
-  # mf_option_learning_agent:
-  #   NAME: "agent_option_learning"
-  #   FLAGS:
-  #     explorer: "agent"
-  #     option_learner: "agent"
-  #     demonstrator: "oracle_process_planning"
-  #     terminate_on_goal_reached_and_option_terminated: True
-  #     agent_planner_isolate_test_session: True
-  #     # agent_sdk_use_docker_sandbox: True
-  #     agent_sdk_use_local_sandbox: True
-  #     agent_sdk_max_agent_turns_per_iteration: 40
+  mf_option_learning_agent:
+    NAME: "agent_option_learning"
+    FLAGS:
+      explorer: "agent"
+      option_learner: "agent"
+      demonstrator: "oracle_process_planning"
+      terminate_on_goal_reached_and_option_terminated: True
+      agent_planner_isolate_test_session: True
+      # agent_sdk_use_docker_sandbox: True
+      agent_sdk_use_local_sandbox: True
+      agent_sdk_max_agent_turns_per_iteration: 40
   # mb_agent:
   #   NAME: "agent_abstraction_learning"
   #   FLAGS: