Skillgen demo by LJones-alt · Pull Request #4 · aarch-uol/IsaacLab-Laura

LJones-alt · 2026-02-25T16:34:49Z

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context.
List any dependencies that are required for this change.

Fixes # (issue)

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (existing functionality will not work without user modification)
Documentation update

Screenshots

Please attach before and after screenshots of the change if applicable.

Checklist

I have read and understood the contribution guidelines
I have run the pre-commit checks with ./isaaclab.sh --format
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
I have updated the changelog and the corresponding version in the extension's config/extension.toml file
I have added my name to the CONTRIBUTORS.md or my name already exists there

Copilot

Pull request overview

This pull request appears to be a demonstration or experimental implementation for "Skillgen" that adds support for a vial insertion task with various backup controllers and recovery mechanisms. The PR introduces new environment configurations, modifies termination and observation logic, adds ensemble training capabilities, and implements state machine-based backup controllers for safe robot manipulation.

Changes:

Adds new environment registration "Dev-IK-Rel-Insert-v1" for top-down vial insertion task
Implements backup controller state machines for insert operations with recovery logic
Updates training configurations and uncertainty-based switching parameters
Adds LaTeX documentation for the switching logic algorithm

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 23 comments.

Show a summary per file

File	Description
terminations.py	Modified termination conditions for object insertion with updated thresholds and added debug prints
observations.py (tasks)	Added debug print statements to observation functions
dev_ik_rel_vial_insert_top_down.py	New configuration file for top-down vial insertion task
dev_ik_rel_vial_insert.py	Updated termination parameters for success criteria
bc_rnn_low_dim.json	Modified WandB project name and increased training epochs
init.py	Registered new "Dev-IK-Rel-Insert-v1" environment
generation.py	Changed log file mode from append to write
observations.py (isaaclab)	Added debug print for joint positions
train_ensemble.py	Updated hard-coded data paths and bootstrap sampling percentage
switching_logic_algorithm.tex	New LaTeX documentation for uncertainty-based switching logic
play_with_backup.py	Changed backup controller task type to "insert"
play_ensemble_v05.py	Enabled recovery mode and updated uncertainty parameters for vial insertion
play_controller.py	Switched to BackupControllerInsertSM
backup_controller_insert_top.py	New state machine controller implementation for top-down insertion
backup_controller_insert.py	Extensive updates to state machine logic with rotation checking and new UNGRASP state
backup_controller_handler.py	Added "insert_top" task type support with duplicate import
run_experiments.sh	Updated experiment configuration for insert task
config.json	New training configuration file
find_best_checkpoints.py	Changed to absolute Docker path

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-25T16:40:33Z

    height_dist = torch.linalg.vector_norm(pos_diff[:, 2:], dim=1)
    xy_dist = torch.linalg.vector_norm(pos_diff[:, :2], dim=1)
-    #print(f"For DEBUG : xy_dist : {xy_dist}, height_dist : {height_dist}")
+   # print(f"For DEBUG : xy_dist : {xy_dist}, height_dist : {height_dist}")


Inconsistent indentation. The comment on line 164 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines.

Copilot · 2026-02-25T16:40:34Z

    #print(f"For DEBUG : STACKED STATUS : {stacked}")
    stacked_upright = stacked & upright_good
+    #print(f"For DEBUG : STACKED STATUS : {stacked_upright}, {stacked}, {upright_good} ")


Inconsistent indentation. The comment on line 191 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines.

Suggested change

#print(f"For DEBUG : STACKED STATUS : {stacked}")

stacked_upright = stacked & upright_good

#print(f"For DEBUG : STACKED STATUS : {stacked_upright}, {stacked}, {upright_good} ")

# print(f"For DEBUG : STACKED STATUS : {stacked}")

stacked_upright = stacked & upright_good

# print(f"For DEBUG : STACKED STATUS : {stacked_upright}, {stacked}, {upright_good} ")

Copilot · 2026-02-25T16:40:35Z

        robot.data.root_state_w[:, :3], robot.data.root_state_w[:, 3:7], object_pos_w
    )
-
+    #print(f"object position : {object_pos_b}")


Inconsistent indentation. The comment on line 37 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines.

Suggested change

#print(f"object position : {object_pos_b}")

# print(f"object position : {object_pos_b}")

Copilot · 2026-02-25T16:40:35Z

+    )
+    dot = wp.min(dot, float(1.0))
+    angle = float(2.0) * wp.acos(dot)
+   # print(f"angle : {angle}, threshold : {threshold_rad}")


Inconsistent indentation. The comment on line 60 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines.

Suggested change

# print(f"angle : {angle}, threshold : {threshold_rad}")

# print(f"angle : {angle}, threshold : {threshold_rad}")

Copilot · 2026-02-25T16:40:35Z

    robot: Articulation = env.scene[robot_cfg.name]
    upper: RigidObject = env.scene[upper_object_cfg.name]
    tilt_deg = upright_tilt_deg(upper.data.root_quat_w, upper.data.default_root_state[:, 3:7])
+    print(f"Tilt deg: {tilt_deg}")


This print statement should be removed before merging to production. Debug print statements left in production code can clutter logs and impact performance.

Suggested change

print(f"Tilt deg: {tilt_deg}")

Copilot · 2026-02-25T16:40:39Z

+        self.observations.subtask_terms.appr_goal=ObsTerm(func=mdp.is_object_lifted, params={"threshold":0.15}
+
+    )


Missing closing parenthesis. The ObsTerm function call on line 112 is not properly closed, which will cause a syntax error.

Suggested change

self.observations.subtask_terms.appr_goal=ObsTerm(func=mdp.is_object_lifted, params={"threshold":0.15}

)

self.observations.subtask_terms.appr_goal = ObsTerm(

func=mdp.is_object_lifted,

params={"threshold": 0.15},

)

Copilot · 2026-02-25T16:40:39Z

+        # self.scene.robot = FRANKA_PANDA_HIGH_PD_CFG.replace(
+        #     prim_path="{ENV_REGEX_NS}/Robot",
+        #     init_state=ArticulationCfg.InitialStateCfg(
+        #         joint_pos={
+
+        #             "panda_joint1":  0.3281,
+        #             "panda_joint2": -0.3684,   
+        #             "panda_joint3":  -0.2787,
+        #             "panda_joint4": -2.6138,  
+        #             "panda_joint5":  -2.7527,
+        #             "panda_joint6":  2.4991,  #  +90° → keeps hand level
+        #             "panda_joint7":  0.3331,
+        #             "panda_finger_joint1": 0.04,   # open gripper
+        #             "panda_finger_joint2": 0.04,
+        #         }
+        #     ),
+        # )


Large blocks of commented-out code (lines 81-97) reduce code readability and maintainability. If this code is no longer needed, it should be removed. If it might be useful in the future, consider documenting why it's kept or moving it to version control history.

Suggested change

# self.scene.robot = FRANKA_PANDA_HIGH_PD_CFG.replace(

# prim_path="{ENV_REGEX_NS}/Robot",

# init_state=ArticulationCfg.InitialStateCfg(

# joint_pos={

# "panda_joint1": 0.3281,

# "panda_joint2": -0.3684,

# "panda_joint3": -0.2787,

# "panda_joint4": -2.6138,

# "panda_joint5": -2.7527,

# "panda_joint6": 2.4991, # +90° → keeps hand level

# "panda_joint7": 0.3331,

# "panda_finger_joint1": 0.04, # open gripper

# "panda_finger_joint2": 0.04,

# }

# ),

# )

Copilot · 2026-02-25T16:40:40Z

+    )
+    dot = wp.min(dot, float(1.0))
+    angle = float(2.0) * wp.acos(dot)
+   # print(f"angle : {angle}, threshold : {threshold_rad}")


Inconsistent indentation. The comment on line 60 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines.

Suggested change

# print(f"angle : {angle}, threshold : {threshold_rad}")

# print(f"angle : {angle}, threshold : {threshold_rad}")

Copilot · 2026-02-25T16:40:40Z

+        #print(f"rpy version :{torch.cat([rest_pos,roll.unsqueeze(0), pitch.unsqueeze(0), yaw.unsqueeze(0)], dim =-1)}")
+        if self.tasktype == "insert_top":
+            rest_pos = torch.tensor([[ 0.6226, -0.0621,  0.3555]], device=self.device)
+            rest_rot = tensor([[ 0, 1, 0, 0]], device=self.device)


Missing import statement for 'tensor'. The code uses 'tensor' on line 105 but it's not imported. This should be 'torch.tensor' to avoid a NameError.

Suggested change

rest_rot = tensor([[ 0, 1, 0, 0]], device=self.device)

rest_rot = torch.tensor([[ 0, 1, 0, 0]], device=self.device)

Copilot · 2026-02-25T16:40:40Z


 def upright_tilt_deg(q_cur, q_init) -> torch.Tensor:
    u_cur, u_init = matrix_from_quat(q_cur)[:, :, 2], matrix_from_quat(q_init)[:, :, 2]
+   # print(f"upright good {torch.rad2deg(torch.acos((u_cur * u_init).sum(-1).clamp(-1.0, 1.0)))}")


Inconsistent indentation. The comment on line 147 has incorrect indentation (3 spaces instead of 4), which violates Python's PEP 8 style guidelines. This should be aligned with the rest of the function body.

Suggested change

# print(f"upright good {torch.rad2deg(torch.acos((u_cur * u_init).sum(-1).clamp(-1.0, 1.0)))}")

# print(f"upright good {torch.rad2deg(torch.acos((u_cur * u_init).sum(-1).clamp(-1.0, 1.0)))}")

LJones-alt added 2 commits February 25, 2026 13:55

minor changs

429b7a2

added in insert top down backup controller

eee4486

Copilot AI review requested due to automatic review settings February 25, 2026 16:34

github-actions Bot added documentation Improvements or additions to documentation isaac-mimic labels Feb 25, 2026

LJones-alt merged commit 90c657d into aarch-uol:skillgen_demo Feb 25, 2026
4 of 7 checks passed

Copilot started reviewing on behalf of LJones-alt February 25, 2026 16:35 View session

Copilot AI reviewed Feb 25, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Skillgen demo#4

Skillgen demo#4
LJones-alt merged 2 commits into
aarch-uol:skillgen_demofrom
LJones-alt:skillgen_demo

LJones-alt commented Feb 25, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	#print(f"object position : {object_pos_b}")
	# print(f"object position : {object_pos_b}")

	# print(f"angle : {angle}, threshold : {threshold_rad}")
	# print(f"angle : {angle}, threshold : {threshold_rad}")

		self.observations.subtask_terms.appr_goal=ObsTerm(func=mdp.is_object_lifted, params={"threshold":0.15}

		)

	rest_rot = tensor([[ 0, 1, 0, 0]], device=self.device)
	rest_rot = torch.tensor([[ 0, 1, 0, 0]], device=self.device)

	# print(f"upright good {torch.rad2deg(torch.acos((u_cur * u_init).sum(-1).clamp(-1.0, 1.0)))}")
	# print(f"upright good {torch.rad2deg(torch.acos((u_cur * u_init).sum(-1).clamp(-1.0, 1.0)))}")

Uh oh!

Conversation

LJones-alt commented Feb 25, 2026

Description

Type of change

Screenshots

Checklist

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants