MuJoCo动力学数值导数

由https://github.com/wecacuee/mujoco_py_deriv分叉

• 运行需要MuJoCo许可证。
• 包装derivative.cpp以从Python调用。

安装

1. 安装mujoco_py
2. pip install mujoco-py-derivatives

使用

准备MuJoCo模型。

import mujoco_py as mj
from mujoco_py_derivatives import MjDerivative, checkderiv

# Prepare mujoco model and data
model = mj.load_model_from_path("flat_pusher_sample.xml")
sim = mj.MjSim(model, nsubsteps=nstep)
dmain = sim.data

计算数值导数

# To compute δf/δx
f = ["qacc"]
x = ["qfrc_applied", "qvel", "qpos"]
deriv_obj = MjDerivative(model, dmain, f, x)
deriv = deriv_obj.compute()