Implementation:Google deepmind Mujoco Engine Vis Interact

Knowledge Sources	Google_deepmind_Mujoco
Domains	Physics Simulation, Visualization, User Interaction
Last Updated	2026-02-15 04:00 GMT

Overview

Implements interactive visualization functions for MuJoCo, including coordinate space transformations, camera control, object selection via ray casting, and user perturbation force application.

Description

This file provides the interactive visualization layer that bridges user input with the physics simulation. It includes room-to-model and model-to-room coordinate transformations (mjv_room2model, mjv_model2room) that account for scene scaling and transformation, camera movement functions (mjv_moveCamera) for orbit, pan, and zoom, object selection via ray casting (mjv_select) that identifies bodies, geoms, and other objects under the mouse cursor, and perturbation force/torque application (mjv_applyPerturbPose, mjv_applyPerturbForce). The camera control supports both free and tracking modes with configurable sensitivity. The perturbation system allows interactive dragging and rotation of bodies during simulation.

Usage

These functions are called by GUI applications (such as the MuJoCo simulate viewer) in response to mouse and keyboard input to manipulate the camera, select objects, and apply perturbation forces to the simulation.

Code Reference

Source Location

Repository: Google_deepmind_Mujoco
File: src/engine/engine_vis_interact.c
Lines: 1-977

Key Functions

// Coordinate transformations
void mjv_room2model(mjtNum* modelpos, mjtNum* modelquat,
                    const mjtNum* roompos, const mjtNum* roomquat,
                    const mjvScene* scn);
void mjv_model2room(mjtNum* roompos, mjtNum* roomquat,
                    const mjtNum* modelpos, const mjtNum* modelquat,
                    const mjvScene* scn);

// Camera control
void mjv_moveCamera(const mjModel* m, int action, mjtNum reldx, mjtNum reldy,
                    const mjvScene* scn, mjvCamera* cam);

// Object selection
int mjv_select(const mjModel* m, const mjData* d, const mjvOption* vopt,
               mjtNum aspectratio, mjtNum relx, mjtNum rely,
               const mjvScene* scn, mjtNum selpnt[3], int geomid[1],
               int skinid[1]);

// Perturbation
void mjv_applyPerturbPose(const mjModel* m, mjData* d,
                          const mjvPerturb* pert, int flg_paused);
void mjv_applyPerturbForce(const mjModel* m, mjData* d,
                           const mjvPerturb* pert);

Import

#include "engine/engine_vis_interact.h"

I/O Contract

Inputs

Name	Type	Required	Description
m	const mjModel*	Yes	Model containing geometry and physics parameters
d	mjData*	Yes	Simulation data with current state
scn	const mjvScene*	Yes	Scene with current transform and scale
action	int	Yes	Camera action type (orbit, pan, zoom)
reldx, reldy	mjtNum	Yes	Relative mouse displacement
pert	const mjvPerturb*	Yes	Perturbation specification (body, force, torque)

Outputs

Name	Type	Description
modelpos, modelquat	mjtNum*	Transformed position and orientation in model space
cam	mjvCamera*	Updated camera state after movement
selpnt	mjtNum[3]	3D selection point on selected object
return value	int	Selected body ID, or -1 if no selection

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment