# Panda Gym Push closed-loop starting policy Panda Gym Push closed-loop starting policy. Native IK actions, privileged simulator observations and a frozen three-seed evaluation; no state teleportation after reset. Record version: `1.0.0` · Hub commit: `d800c48cfd84f618a5c6ab84f8f7feec74ab0b56` ## Executable interface ```json { "version": "1", "kind": "task", "abstraction_level": 2, "execution_domain": "simulation", "language": "python", "input_transport": "json_file_or_inline", "output_transport": "json_stdout", "output_selector": "outputs", "conventions": { "linear_unit": "m", "angular_unit": "rad", "time_unit": "s", "coordinate_frame": "right_handed_world_xyz; +z_up", "joint_order": "not_used", "quaternion_order": "not_used" }, "input_schema": { "type": "object", "required": [ "seed" ], "properties": { "seed": { "type": "integer", "minimum": 0, "maximum": 10000 } }, "additionalProperties": false }, "output_schema": { "type": "object", "required": [ "achieved_goal", "desired_goal", "error_m", "steps", "environment_success" ], "properties": { "achieved_goal": { "type": "array", "minItems": 3, "maxItems": 3, "items": { "type": "number", "minimum": -30, "maximum": 30 }, "x-unit": "m", "x-frame": "right_handed_world_xyz; +z_up" }, "desired_goal": { "type": "array", "minItems": 3, "maxItems": 3, "items": { "type": "number", "minimum": -30, "maximum": 30 }, "x-unit": "m", "x-frame": "right_handed_world_xyz; +z_up" }, "error_m": { "type": "number", "minimum": 0, "maximum": 50 }, "steps": { "type": "integer", "minimum": 1, "maximum": 1500 }, "environment_success": { "type": "boolean" } }, "additionalProperties": false }, "preconditions": [ "Use the exact recorded source, Linux Python 3.10 environment and pinned direct packages.", "No physical robot, network actuation or external robot credentials." ], "postconditions": [ "All reported checks must be true and all output numbers finite.", "Success covers the declared goal and frozen cases, not arbitrary tasks or physical hardware." ], "failure_codes": [ "invalid_input_or_environment", "execution_failed", "task_failed" ], "composition": { "depends_on": [], "automatic_frame_conversion": false, "automatic_hardware_execution": false } } ``` ## Exact environment | Runtime | Robot | Controller | Observation | OS | Python | Evidence | | --- | --- | --- | --- | --- | --- | --- | | panda-gym 3.0.7 | franka-panda | native_pybullet_ik_cartesian_feedback | goal_dict | linux x86_64 | >=3.10,<3.11 | import_verified | Privileged simulator state and declared goals; no perception or real sensor validation. Compatibility status: **source_reviewed**. An environment match does not establish task completion. Runtime verification: **import_verified**. ## Source and license - Repository: https://github.com/qgallouedec/panda-gym - Source commit: `14d944120037a5bb637e629965f88bb9e13dba3b` (no tag) - Original authors: Original panda-gym contributors (preserve pinned LICENSE) - License: [MIT](https://github.com/qgallouedec/panda-gym/blob/14d944120037a5bb637e629965f88bb9e13dba3b/LICENSE) - Source archive SHA256: `fd2fde8e90118b215f27c70654bccb0a0bd0042eb10cc986b3ae5168d5635fb1` - Upstream files: `README.md`, `setup.py`, `LICENSE`, `panda_gym/pybullet.py`, `panda_gym/utils.py`, `panda_gym/version.txt`, `panda_gym/__init__.py`, `panda_gym/envs/panda_tasks.py`, `panda_gym/envs/core.py`, `panda_gym/envs/__init__.py`, `panda_gym/envs/tasks/stack.py`, `panda_gym/envs/tasks/flip.py`, `panda_gym/envs/tasks/slide.py`, `panda_gym/envs/tasks/push.py`, `panda_gym/envs/tasks/pick_and_place.py`, `panda_gym/envs/tasks/reach.py`, `panda_gym/envs/tasks/__init__.py`, `panda_gym/envs/robots/panda.py`, `panda_gym/envs/robots/__init__.py`, `panda_gym/assets/colored_cube.png`, `panda_gym/assets/__init__.py` - Upstream contribution rules: https://github.com/qgallouedec/panda-gym ## Find, inspect, prepare ```sh python /PATH_TO_SKILL/scripts/rsh.py find "task" --runtime panda-gym --runtime-version 3.0.7 --robot franka-panda --json python /PATH_TO_SKILL/scripts/rsh.py inspect pandagym-panda-push --json python /PATH_TO_SKILL/scripts/rsh.py prepare pandagym-panda-push --project ./experiment --expected-source-commit 14d944120037a5bb637e629965f88bb9e13dba3b --expected-record-version 1.0.0 --plan --json ``` Remove `--plan` to download the checksum-verified pinned source and create an isolated project environment. Preparation returns the remaining configuration. Dependency installation is an explicit `--install-deps` step; it can execute package build hooks. Robot simulation is a separate command returned by prepare, never run by find/inspect. Direct package pins: `panda-gym==3.0.7`, `pybullet==3.2.7`, `gymnasium==1.2.0`, `numpy==1.26.4`. Direct runtime versions are pinned and measured. The receipt retains the complete environment freeze; prepare records the installed environment. System drivers are separate host requirements. Hardware: Linux x86_64; CPU simulation; see retained receipt for the exact host environment. Rendering: Headless. EGL libraries are required for MuJoCo imports; PushT uses SDL dummy video. No interactive display.. External API keys: none. ## Invocation contract Entrypoint: `python run.py`. ```json { "argv": [ "{python}", "{recipe_dir}/run.py", "--source-dir", "{source_dir}", "--example" ], "inputs": { "seed": 0 }, "outputs": { "format": "single JSON object", "payload_selector": "outputs", "schema": { "type": "object", "required": [ "achieved_goal", "desired_goal", "error_m", "steps", "environment_success" ], "properties": { "achieved_goal": { "type": "array", "minItems": 3, "maxItems": 3, "items": { "type": "number", "minimum": -30, "maximum": 30 }, "x-unit": "m", "x-frame": "right_handed_world_xyz; +z_up" }, "desired_goal": { "type": "array", "minItems": 3, "maxItems": 3, "items": { "type": "number", "minimum": -30, "maximum": 30 }, "x-unit": "m", "x-frame": "right_handed_world_xyz; +z_up" }, "error_m": { "type": "number", "minimum": 0, "maximum": 50 }, "steps": { "type": "integer", "minimum": 1, "maximum": 1500 }, "environment_success": { "type": "boolean" } }, "additionalProperties": false } } } ``` Success: All declared goal and finite-state checks pass; exit 0 with task_complete. Native task success plus final Euclidean goal error <.04 m. Airborne pick/place goals retain the grasp; ground goals are released. Timeout: 180 seconds. Terminate the adapter at the recorded timeout. No physical hardware is opened. Failures: Invalid input or source checksum; Different pinned package version; Unavailable runtime/driver; Goal not reached within bounded steps. ## Actual evidence 1/3 frozen cases completed; retained failures are not promoted. Reviewer: Hub finite-case Lambda execution harness · 2026-10-20. Commands/checks actually performed: - python tools/verify_runtime_expansion.py --backend pandagym --source-cache SOURCE_CACHE --output OUTPUT Submitted evidence and hub verification remain separate. A schema check, PR creation or merge does not establish task success or publication. ## Reusable fixes Change the runtime expansion manifest/template, regenerate, and rerun all frozen cases. Preserve source/model/checkpoint licenses and distinguish simulation from hardware. # Panda Gym Push closed-loop starting policy Panda Gym Push closed-loop starting policy. Native IK actions, privileged simulator observations and a frozen three-seed evaluation; no state teleportation after reset. Native task success plus final Euclidean goal error <.04 m. Airborne pick/place goals retain the grasp; ground goals are released. Source and model terms: MIT. The Hub wrapper is MIT. See the record for input JSON, exact environment, frozen execution receipts and preparation requirements.