reference-motion
Load, classify, interpolate, validate, and safely inspect MyoSuite fixed, randomized, and tracked reference motions.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- reference-motion
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Load, classify, interpolate, validate, and safely inspect MyoSuite fixed, randomized, and tracked reference motions.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/api-reference.md2.2 KB
- 参考文档references/troubleshooting.md3.1 KB
- 参考文档references/workflows.md2.2 KB
03 SKILL.md 正文
MyoSuite reference motion
Use this route when a task involves reference trajectories, motion playback,
fixed/random/track targets, interpolation, initialization states, reference
logs, or examine_reference diagnostics.
Route the request
- Core task IDs, reset/step, and action/observation contracts:
environments. - Window/offscreen rendering and MuJoCo camera behavior:
simulation-rendering. - JAX parity or accelerated MJX reference handling:
mjx-acceleration. - Long-running policy training or learner configuration:
training-integration.
Read the API reference for the data contract,
workflows for bounded recipes, and
troubleshooting before diagnosing malformed
files or out-of-range times. The bundled
scripts/reference_motion_smoke.py is a safe in-memory check and does not need
original repository data.
Fast path
- Choose NumPy
ReferenceMotionfor a normal CPU environment. - Supply either a dict or a
.npz/.pkl/.picklepath containingtimeand the relevantrobot,robot_vel, andobjectarrays. - Let the class infer
FIXED(one frame),RANDOM(two low/high frames), orTRACK(more than two frames). - Call
get_init(), thenget_reference(time)at bounded times; callreset()before replaying a trajectory from its beginning. - Use
--render nonefor CLI playback in headless verification. Rendering is a separate route, not a prerequisite for checking interpolation.
import numpy as np
from myosuite.logger.reference_motion import ReferenceMotion
reference = {
"time": np.array([0.0, 1.0, 2.0]),
"robot": np.zeros((3, 2)),
"robot_vel": np.zeros((3, 2)),
"object": np.zeros((3, 7)),
}
motion = ReferenceMotion(reference)
robot_init, object_init = motion.get_init()
frame = motion.get_reference(0.5) # linearly interpolated TRACK frame
Validate dimensions and time bounds before coupling a motion to a task. Keep reference arrays and returned structures copied when storing them beyond one simulation step.
Verification boundary
Core NumPy reference handling and bounded synthetic fixed/random/track behavior
are suitable for CPU verification. JAX parity, large .npz corpora, and full
playback remain optional or data-dependent; never claim they pass from a CPU
import alone. The source-native candidate is the reference-motion test family,
run only when its data and optional dependencies are available.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.