Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Implementation:Google deepmind Mujoco Engine Derivative

From Leeroopedia
Revision as of 12:45, 16 February 2026 by Admin (talk | contribs) (Auto-imported from implementations/Google_deepmind_Mujoco_Engine_Derivative.md)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Knowledge Sources
Domains Physics Simulation, Analytical Derivatives, Optimal Control
Last Updated 2026-02-15 04:00 GMT

Overview

Implements analytical derivatives of MuJoCo's smooth dynamics with respect to velocity, including derivatives of the recursive Newton-Euler algorithm, passive forces, actuator forces, and fluid drag.

Description

This module computes analytical (closed-form) derivatives of the smooth dynamics equations. It provides derivatives of spatial algebra operations (cross products, motion/force transforms), the recursive Newton-Euler algorithm with respect to velocity (mjd_rne_vel, mjd_rne_vel_dense), center-of-mass velocity derivatives (mjd_comVel_vel), quaternion integration derivatives (mjd_quatIntegrate), quaternion subtraction derivatives (mjd_subQuat), passive force derivatives including fluid dynamics models (mjd_passive_vel, mjd_ellipsoidFluid, mjd_inertiaBoxFluid), actuator velocity derivatives (mjd_actuator_vel), and flex interpolation derivatives (mjd_flexInterp_mulKD, mjd_flexInterp_addH). The top-level mjd_smooth_vel assembles the full velocity derivative of smooth dynamics.

Usage

Called during derivative computation to analytically compute the Jacobian of the forward dynamics with respect to generalized velocities. Used by optimal control, system identification, and model-based reinforcement learning algorithms that require efficient gradient computation.

Code Reference

Source Location

Key Functions

// Quaternion derivatives
void mjd_subQuat(const mjtNum qa[4], const mjtNum qb[4], mjtNum Da[9], mjtNum Db[9]);
void mjd_quatIntegrate(const mjtNum vel[3], mjtNum scale, mjtNum Dvel[9], mjtNum Dscale[3]);

// Velocity derivatives of smooth dynamics (dense and sparse)
void mjd_smooth_vel(const mjModel* m, mjData* d, int flg_bias);

// Recursive Newton-Euler velocity derivatives
void mjd_rne_vel_dense(const mjModel* m, mjData* d);

// Passive force velocity derivatives
void mjd_passive_vel(const mjModel* m, mjData* d);

// Fluid dynamics derivatives
void mjd_ellipsoidFluid(const mjModel* m, mjData* d, int bodyid);
void mjd_inertiaBoxFluid(const mjModel* m, mjData* d, int i);

// Actuator velocity derivatives
void mjd_actuator_vel(const mjModel* m, mjData* d);

// Flex interpolation derivatives
void mjd_flexInterp_mulKD(const mjModel* m, mjData* d, mjtNum* res,
                          const mjtNum* vec, mjtNum h);
void mjd_flexInterp_addH(const mjModel* m, mjData* d, mjtNum* H,
                         const int* dof_indices, int ndof, mjtNum h);

Import

#include "engine/engine_derivative.h"

I/O Contract

Inputs

Name Type Required Description
m const mjModel* Yes Physics model with body/joint/actuator definitions
d mjData* Yes Simulation state with computed kinematics and velocities
flg_bias int Varies Flag to include bias force derivatives in smooth_vel

Outputs

Name Type Description
d->qDeriv mjtNum* Derivative of smooth dynamics w.r.t. generalized velocity (sparse)
Da, Db mjtNum* Quaternion derivative matrices (3x3)
Dvel, Dscale mjtNum* Quaternion integration derivative outputs

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment