MuJoCo Warp API

Public API for MJWarp.

step(m: Model, d: Data)

Advance simulation.

class Model

Model definition and parameters.

nq

number of generalized coordinates

Type:

int

nv

number of degrees of freedom

Type:

int

nu

number of actuators/controls

Type:

int

na

number of activation states

Type:

int

nbody

number of bodies

Type:

int

noct

number of total octree cells in all meshes

Type:

int

njnt

number of joints

Type:

int

ntree

number of kinematic trees

Type:

int

nM

number of non-zeros in sparse inertia matrix

Type:

int

nC

number of non-zeros in sparse body-dof matrix

Type:

int

ngeom

number of geoms

Type:

int

nsite

number of sites

Type:

int

ncam

number of cameras

Type:

int

nlight

number of lights

Type:

int

nflex

number of flexes

Type:

int

nflexvert

number of vertices in all flexes

Type:

int

nflexedge

number of edges in all flexes

Type:

int

nflexelem

number of elements in all flexes

Type:

int

nflexelemdata

number of element vertex ids in all flexes

Type:

int

nflexelemedge

number of element edge ids in all flexes

Type:

int

nmesh

number of meshes

Type:

int

nmeshvert

number of vertices for all meshes

Type:

int

nmeshnormal

number of normals in all meshes

Type:

int

nmeshface

number of faces for all meshes

Type:

int

nmeshgraph

number of ints in mesh auxiliary data

Type:

int

nmeshpoly

number of polygons in all meshes

Type:

int

nmeshpolyvert

number of vertices in all polygons

Type:

int

nmeshpolymap

number of polygons in vertex map

Type:

int

nhfield

number of heightfields

Type:

int

nhfielddata

size of elevation data

Type:

int

nmat

number of materials

Type:

int

npair

number of predefined geom pairs

Type:

int

nexclude

number of excluded geom pairs

Type:

int

neq

number of equality constraints

Type:

int

ntendon

number of tendons

Type:

int

nwrap

number of wrap objects in all tendon paths

Type:

int

nsensor

number of sensors

Type:

int

nmocap

number of mocap bodies

Type:

int

nplugin

number of plugin instances

Type:

int

ngravcomp

number of bodies with nonzero gravcomp

Type:

int

nsensordata

number of elements in sensor data vector

Type:

int

opt

physics options

Type:

mujoco_warp._src.types.Option

stat

model statistics

Type:

mujoco_warp._src.types.Statistic

qpos0

qpos values at default pose (*, nq)

Type:

array(ndim=2, dtype=float32)

qpos_spring

reference pose for springs (*, nq)

Type:

array(ndim=2, dtype=float32)

body_parentid

id of body’s parent (nbody,)

Type:

array(ndim=1, dtype=int32)

body_rootid

id of root above body (nbody,)

Type:

array(ndim=1, dtype=int32)

body_weldid

id of body that this body is welded to (nbody,)

Type:

array(ndim=1, dtype=int32)

body_mocapid

id of mocap data; -1: none (nbody,)

Type:

array(ndim=1, dtype=int32)

body_jntnum

number of joints for this body (nbody,)

Type:

array(ndim=1, dtype=int32)

body_jntadr

start addr of joints; -1: no joints (nbody,)

Type:

array(ndim=1, dtype=int32)

body_dofnum

number of motion degrees of freedom (nbody,)

Type:

array(ndim=1, dtype=int32)

body_dofadr

start addr of dofs; -1: no dofs (nbody,)

Type:

array(ndim=1, dtype=int32)

body_treeid

id of body’s tree; -1: static (nbody,)

Type:

array(ndim=1, dtype=int32)

body_geomnum

number of geoms (nbody,)

Type:

array(ndim=1, dtype=int32)

body_geomadr

start addr of geoms; -1: no geoms (nbody,)

Type:

array(ndim=1, dtype=int32)

body_pos

position offset rel. to parent body (*, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

body_quat

orientation offset rel. to parent body (*, nbody, 4)

Type:

array(ndim=2, dtype=quatf)

body_ipos

local position of center of mass (*, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

body_iquat

local orientation of inertia ellipsoid (*, nbody, 4)

Type:

array(ndim=2, dtype=quatf)

body_mass

mass (*, nbody,)

Type:

array(ndim=2, dtype=float32)

body_subtreemass

mass of subtree starting at this body (*, nbody,)

Type:

array(ndim=2, dtype=float32)

body_inertia

diagonal inertia in ipos/iquat frame (*, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

body_invweight0

mean inv inert in qpos0 (trn, rot) (*, nbody, 2)

Type:

array(ndim=2, dtype=vec2f)

body_gravcomp

antigravity force, units of body weight (*, nbody)

Type:

array(ndim=2, dtype=float32)

body_contype

OR over all geom contypes (nbody,)

Type:

array(ndim=1, dtype=int32)

body_conaffinity

OR over all geom conaffinities (nbody,)

Type:

array(ndim=1, dtype=int32)

oct_child

octree children (noct, 8)

Type:

array(ndim=1, dtype=vector(length=8, dtype=int32))

oct_aabb

octree axis-aligned bounding boxes (noct, 2, 3)

Type:

array(ndim=2, dtype=vec3f)

oct_coeff

octree interpolation coefficients (noct, 8)

Type:

array(ndim=1, dtype=vector(length=8, dtype=float32))

jnt_type

type of joint (JointType) (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_qposadr

start addr in ‘qpos’ for joint’s data (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_dofadr

start addr in ‘qvel’ for joint’s data (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_bodyid

id of joint’s body (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_limited

does joint have limits (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_actfrclimited

does joint have actuator force limits (njnt,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

jnt_actgravcomp

is gravcomp force applied via actuators (njnt,)

Type:

array(ndim=1, dtype=int32)

jnt_solref

constraint solver reference: limit (*, njnt, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

jnt_solimp

constraint solver impedance: limit (*, njnt, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

jnt_pos

local anchor position (*, njnt, 3)

Type:

array(ndim=2, dtype=vec3f)

jnt_axis

local joint axis (*, njnt, 3)

Type:

array(ndim=2, dtype=vec3f)

jnt_stiffness

stiffness coefficient (*, njnt)

Type:

array(ndim=2, dtype=float32)

jnt_range

joint limits (*, njnt, 2)

Type:

array(ndim=2, dtype=vec2f)

jnt_actfrcrange

range of total actuator force (*, njnt, 2)

Type:

array(ndim=2, dtype=vec2f)

jnt_margin

min distance for limit detection (*, njnt)

Type:

array(ndim=2, dtype=float32)

dof_bodyid

id of dof’s body (nv,)

Type:

array(ndim=1, dtype=int32)

dof_jntid

id of dof’s joint (nv,)

Type:

array(ndim=1, dtype=int32)

dof_parentid

id of dof’s parent; -1: none (nv,)

Type:

array(ndim=1, dtype=int32)

dof_treeid

id of dof’s tree (nv,)

Type:

array(ndim=1, dtype=int32)

dof_Madr

dof address in M-diagonal (nv,)

Type:

array(ndim=1, dtype=int32)

dof_solref

constraint solver reference: frictionloss (*, nv, NREF)

Type:

array(ndim=2, dtype=vec2f)

dof_solimp

constraint solver impedance: frictionloss (*, nv, NIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

dof_frictionloss

dof friction loss (*, nv)

Type:

array(ndim=2, dtype=float32)

dof_armature

dof armature inertia/mass (*, nv)

Type:

array(ndim=2, dtype=float32)

dof_damping

damping coefficient (*, nv)

Type:

array(ndim=2, dtype=float32)

dof_invweight0

diag. inverse inertia in qpos0 (*, nv)

Type:

array(ndim=2, dtype=float32)

tree_bodynum

number of bodies in tree (incl. root) (ntree,)

Type:

array(ndim=1, dtype=int32)

tree_dofadr

start address of tree’s dofs (ntree,)

Type:

array(ndim=1, dtype=int32)

tree_dofnum

number of dofs in tree (ntree,)

Type:

array(ndim=1, dtype=int32)

geom_type

geometric type (GeomType) (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_contype

geom contact type (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_conaffinity

geom contact affinity (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_condim

contact dimensionality (1, 3, 4, 6) (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_bodyid

id of geom’s body (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_dataid

id of geom’s mesh/hfield; -1: none (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_matid

material id for rendering (*, ngeom,)

Type:

array(ndim=2, dtype=int32)

geom_group

geom group inclusion/exclusion mask (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_priority

geom contact priority (ngeom,)

Type:

array(ndim=1, dtype=int32)

geom_solmix

mixing coef for solref/imp in geom pair (*, ngeom,)

Type:

array(ndim=2, dtype=float32)

geom_solref

constraint solver reference: contact (*, ngeom, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

geom_solimp

constraint solver impedance: contact (*, ngeom, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

geom_size

geom-specific size parameters (*, ngeom, 3)

Type:

array(ndim=2, dtype=vec3f)

geom_aabb

bounding box, (center, size) (*, ngeom, 2, 3)

Type:

array(ndim=3, dtype=vec3f)

geom_rbound

radius of bounding sphere (*, ngeom,)

Type:

array(ndim=2, dtype=float32)

geom_pos

local position offset rel. to body (*, ngeom, 3)

Type:

array(ndim=2, dtype=vec3f)

geom_quat

local orientation offset rel. to body (*, ngeom, 4)

Type:

array(ndim=2, dtype=quatf)

geom_friction

friction for (slide, spin, roll) (*, ngeom, 3)

Type:

array(ndim=2, dtype=vec3f)

geom_margin

detect contact if dist<margin (*, ngeom,)

Type:

array(ndim=2, dtype=float32)

geom_gap

include in solver if dist<margin-gap (*, ngeom,)

Type:

array(ndim=2, dtype=float32)

geom_fluid

fluid interaction parameters (ngeom, mjNFLUID)

Type:

array(ndim=2, dtype=float32)

geom_rgba

rgba when material is omitted (*, ngeom, 4)

Type:

array(ndim=2, dtype=vec4f)

site_type

geom type for rendering (GeomType) (nsite,)

Type:

array(ndim=1, dtype=int32)

site_bodyid

id of site’s body (nsite,)

Type:

array(ndim=1, dtype=int32)

site_size

geom size for rendering (nsite, 3)

Type:

array(ndim=1, dtype=vec3f)

site_pos

local position offset rel. to body (*, nsite, 3)

Type:

array(ndim=2, dtype=vec3f)

site_quat

local orientation offset rel. to body (*, nsite, 4)

Type:

array(ndim=2, dtype=quatf)

cam_mode

camera tracking mode (CamLightType) (ncam,)

Type:

array(ndim=1, dtype=int32)

cam_bodyid

id of camera’s body (ncam,)

Type:

array(ndim=1, dtype=int32)

cam_targetbodyid

id of targeted body; -1: none (ncam,)

Type:

array(ndim=1, dtype=int32)

cam_pos

position rel. to body frame (*, ncam, 3)

Type:

array(ndim=2, dtype=vec3f)

cam_quat

orientation rel. to body frame (*, ncam, 4)

Type:

array(ndim=2, dtype=quatf)

cam_poscom0

global position rel. to sub-com in qpos0 (*, ncam, 3)

Type:

array(ndim=2, dtype=vec3f)

cam_pos0

global position rel. to body in qpos0 (*, ncam, 3)

Type:

array(ndim=2, dtype=vec3f)

cam_mat0

global orientation in qpos0 (*, ncam, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

cam_projection

projection type (ProjectionType) (ncam,)

Type:

array(ndim=1, dtype=int32)

cam_fovy

y field-of-view (ortho ? len : deg) (*, ncam)

Type:

array(ndim=2, dtype=float32)

cam_resolution

resolution: pixels [width, height] (ncam, 2)

Type:

array(ndim=1, dtype=vec2i)

cam_sensorsize

sensor size: length [width, height] (ncam, 2)

Type:

array(ndim=1, dtype=vec2f)

cam_intrinsic

[focal length; principal point] (*, ncam, 4)

Type:

array(ndim=2, dtype=vec4f)

light_mode

light tracking mode (CamLightType) (nlight,)

Type:

array(ndim=1, dtype=int32)

light_bodyid

id of light’s body (nlight,)

Type:

array(ndim=1, dtype=int32)

light_targetbodyid

id of targeted body; -1: none (nlight,)

Type:

array(ndim=1, dtype=int32)

light_type

spot, directional, etc. (mjtLightType) (*, nlight)

Type:

array(ndim=2, dtype=int32)

light_castshadow

does light cast shadows (*, nlight)

Type:

array(ndim=2, dtype=warp._src.types.bool)

light_active

is light active (*, nlight)

Type:

array(ndim=2, dtype=warp._src.types.bool)

light_pos

position rel. to body frame (*, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

light_dir

direction rel. to body frame (*, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

light_poscom0

global position rel. to sub-com in qpos0 (*, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

light_pos0

global position rel. to body in qpos0 (*, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

light_dir0

global direction in qpos0 (*, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

flex_dim

1: lines, 2: triangles, 3: tetrahedra (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_vertadr

first vertex address (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_vertnum

number of vertices (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_edgeadr

first edge address (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_edgenum

number of edges (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_elemadr

first element address (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_elemnum

number of elements (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_elemedgeadr

first element address (nflex,)

Type:

array(ndim=1, dtype=int32)

flex_vertbodyid

vertex body ids (nflexvert,)

Type:

array(ndim=1, dtype=int32)

flex_edge

edge vertex ids (2 per edge) (nflexedge, 2)

Type:

array(ndim=1, dtype=vec2i)

flex_edgeflap

adjacent vertex ids (dim=2 only) (nflexedge, 2)

Type:

array(ndim=1, dtype=vec2i)

flex_elem

element vertex ids (dim+1 per elem) (nflexelemdata,)

Type:

array(ndim=1, dtype=int32)

flex_elemedge

element edge ids (nflexelemedge,)

Type:

array(ndim=1, dtype=int32)

flexedge_length0

edge lengths in qpos0 (nflexedge,)

Type:

array(ndim=1, dtype=float32)

flexedge_invweight0

inv. inertia for the edge (nflexedge,)

Type:

array(ndim=1, dtype=float32)

flex_stiffness

finite element stiffness matrix (nflexelem, 21)

Type:

array(ndim=2, dtype=float32)

flex_bending

bending stiffness (nflexedge, 17)

Type:

array(ndim=2, dtype=float32)

flex_damping

Rayleigh’s damping coefficient (nflex,)

Type:

array(ndim=1, dtype=float32)

flexedge_J_rownnz

number of nonzeros in Jacobian row (nflexedge,)

Type:

array(ndim=1, dtype=int32)

flexedge_J_rowadr

row start address in colind array (nflexedge,)

Type:

array(ndim=1, dtype=int32)

flexedge_J_colind

column indices in sparse Jacobian (nJfe,)

Type:

array(ndim=1, dtype=int32)

mesh_vertadr

first vertex address (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_vertnum

number of vertices (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_faceadr

first face address (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_normaladr

first normal address (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_graphadr

graph data address; -1: no graph (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_vert

vertex positions for all meshes (nmeshvert, 3)

Type:

array(ndim=1, dtype=vec3f)

mesh_normal

normals for all meshes (nmeshnormal, 3)

Type:

array(ndim=1, dtype=vec3f)

mesh_face

face indices for all meshes (nface, 3)

Type:

array(ndim=1, dtype=vec3i)

mesh_graph

convex graph data (nmeshgraph,)

Type:

array(ndim=1, dtype=int32)

mesh_quat

rotation applied to asset vertices (nmesh, 4)

Type:

array(ndim=1, dtype=quatf)

mesh_polynum

number of polygons per mesh (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_polyadr

first polygon address per mesh (nmesh,)

Type:

array(ndim=1, dtype=int32)

mesh_polynormal

all polygon normals (nmeshpoly, 3)

Type:

array(ndim=1, dtype=vec3f)

mesh_polyvertadr

polygon vertex start address (nmeshpoly,)

Type:

array(ndim=1, dtype=int32)

mesh_polyvertnum

number of vertices per polygon (nmeshpoly,)

Type:

array(ndim=1, dtype=int32)

mesh_polyvert

all polygon vertices (nmeshpolyvert,)

Type:

array(ndim=1, dtype=int32)

mesh_polymapadr

first polygon address per vertex (nmeshvert,)

Type:

array(ndim=1, dtype=int32)

mesh_polymapnum

number of polygons per vertex (nmeshvert,)

Type:

array(ndim=1, dtype=int32)

mesh_polymap

vertex to polygon map (nmeshpolymap,)

Type:

array(ndim=1, dtype=int32)

hfield_size

(x, y, z_top, z_bottom) (nhfield, 4)

Type:

array(ndim=1, dtype=vec4f)

hfield_nrow

number of rows in grid (nhfield,)

Type:

array(ndim=1, dtype=int32)

hfield_ncol

number of columns in grid (nhfield,)

Type:

array(ndim=1, dtype=int32)

hfield_adr

start address in hfield_data (nhfield,)

Type:

array(ndim=1, dtype=int32)

hfield_data

elevation data (nhfielddata,)

Type:

array(ndim=1, dtype=float32)

mat_texid

texture id for rendering (*, nmat, mjNTEXROLE)

Type:

array(ndim=3, dtype=int32)

mat_texrepeat

texture repeat for rendering (*, nmat, 2)

Type:

array(ndim=2, dtype=vec2f)

mat_rgba

rgba (*, nmat, 4)

Type:

array(ndim=2, dtype=vec4f)

pair_dim

contact dimensionality (npair,)

Type:

array(ndim=1, dtype=int32)

pair_geom1

id of geom1 (npair,)

Type:

array(ndim=1, dtype=int32)

pair_geom2

id of geom2 (npair,)

Type:

array(ndim=1, dtype=int32)

pair_solref

solver reference: contact normal (*, npair, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

pair_solreffriction

solver reference: contact friction (*, npair, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

pair_solimp

solver impedance: contact (*, npair, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

pair_margin

detect contact if dist<margin (*, npair,)

Type:

array(ndim=2, dtype=float32)

pair_gap

include in solver if dist<margin-gap (*, npair,)

Type:

array(ndim=2, dtype=float32)

pair_friction

tangent1, 2, spin, roll1, 2 (*, npair, 5)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

exclude_signature

body1 << 16 + body2 (nexclude,)

Type:

array(ndim=1, dtype=int32)

eq_type

constraint type (EqType) (neq,)

Type:

array(ndim=1, dtype=int32)

eq_obj1id

id of object 1 (neq,)

Type:

array(ndim=1, dtype=int32)

eq_obj2id

id of object 2 (neq,)

Type:

array(ndim=1, dtype=int32)

eq_objtype

type of both objects (ObjType) (neq,)

Type:

array(ndim=1, dtype=int32)

eq_active0

initial enable/disable constraint state (neq,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

eq_solref

constraint solver reference (*, neq, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

eq_solimp

constraint solver impedance (*, neq, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

eq_data

numeric data for constraint (*, neq, mjNEQDATA)

Type:

array(ndim=2, dtype=vector(length=11, dtype=float32))

tendon_adr

address of first object in tendon’s path (ntendon,)

Type:

array(ndim=1, dtype=int32)

tendon_num

number of objects in tendon’s path (ntendon,)

Type:

array(ndim=1, dtype=int32)

tendon_limited

does tendon have length limits (ntendon,)

Type:

array(ndim=1, dtype=int32)

tendon_actfrclimited

does ten have actuator force limit (ntendon,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

tendon_solref_lim

constraint solver reference: limit (*, ntendon, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

tendon_solimp_lim

constraint solver impedance: limit (*, ntendon, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

tendon_solref_fri

constraint solver reference: friction (*, ntendon, mjNREF)

Type:

array(ndim=2, dtype=vec2f)

tendon_solimp_fri

constraint solver impedance: friction (*, ntendon, mjNIMP)

Type:

array(ndim=2, dtype=vector(length=5, dtype=float32))

tendon_range

tendon length limits (*, ntendon, 2)

Type:

array(ndim=2, dtype=vec2f)

tendon_actfrcrange

range of total actuator force (*, ntendon, 2)

Type:

array(ndim=2, dtype=vec2f)

tendon_margin

min distance for limit detection (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_stiffness

stiffness coefficient (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_damping

damping coefficient (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_armature

inertia associated with tendon velocity (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_frictionloss

loss due to friction (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_lengthspring

spring resting length range (*, ntendon, 2)

Type:

array(ndim=2, dtype=vec2f)

tendon_length0

tendon length in qpos0 (*, ntendon)

Type:

array(ndim=2, dtype=float32)

tendon_invweight0

inv. weight in qpos0 (*, ntendon)

Type:

array(ndim=2, dtype=float32)

wrap_type

wrap object type (WrapType) (nwrap,)

Type:

array(ndim=1, dtype=int32)

wrap_objid

object id: geom, site, joint (nwrap,)

Type:

array(ndim=1, dtype=int32)

wrap_prm

divisor, joint coef, or site id (nwrap,)

Type:

array(ndim=1, dtype=float32)

actuator_trntype

transmission type (TrnType) (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_dyntype

dynamics type (DynType) (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_gaintype

gain type (GainType) (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_biastype

bias type (BiasType) (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_trnid

transmission id: joint, tendon, site (nu, 2)

Type:

array(ndim=1, dtype=vec2i)

actuator_actadr

first activation address; -1: stateless (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_actnum

number of activation variables (nu,)

Type:

array(ndim=1, dtype=int32)

actuator_ctrllimited

is control limited (nu,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

actuator_forcelimited

is force limited (nu,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

actuator_actlimited

is activation limited (nu,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

actuator_dynprm

dynamics parameters (*, nu, mjNDYN)

Type:

array(ndim=2, dtype=vector(length=10, dtype=float32))

actuator_gainprm

gain parameters (*, nu, mjNGAIN)

Type:

array(ndim=2, dtype=vector(length=10, dtype=float32))

actuator_biasprm

bias parameters (*, nu, mjNBIAS)

Type:

array(ndim=2, dtype=vector(length=10, dtype=float32))

actuator_actearly

step activation before force (nu,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

actuator_ctrlrange

range of controls (*, nu, 2)

Type:

array(ndim=2, dtype=vec2f)

actuator_forcerange

range of forces (*, nu, 2)

Type:

array(ndim=2, dtype=vec2f)

actuator_actrange

range of activations (*, nu, 2)

Type:

array(ndim=2, dtype=vec2f)

actuator_gear

scale length and transmitted force (*, nu, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

actuator_cranklength

crank length for slider-crank (nu,)

Type:

array(ndim=1, dtype=float32)

actuator_acc0

acceleration from unit force in qpos0 (nu,)

Type:

array(ndim=1, dtype=float32)

actuator_lengthrange

feasible actuator length range (nu, 2)

Type:

array(ndim=1, dtype=vec2f)

sensor_type

sensor type (SensorType) (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_datatype

numeric data type (DataType) (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_objtype

type of sensorized object (ObjType) (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_objid

id of sensorized object (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_reftype

type of reference frame (ObjType) (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_refid

id of reference frame; -1: global frame (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_intprm

sensor parameters (nsensor, mjNSENS)

Type:

array(ndim=2, dtype=int32)

sensor_dim

number of scalar outputs (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_adr

address in sensor array (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_cutoff

cutoff for real and positive; 0: ignore (nsensor,)

Type:

array(ndim=1, dtype=float32)

plugin

globally registered plugin slot number (nplugin,)

Type:

array(ndim=1, dtype=int32)

plugin_attr

config attributes of geom plugin (nplugin, 3)

Type:

array(ndim=1, dtype=vec3f)

M_rownnz

number of non-zeros in each row of qM (nv,)

Type:

array(ndim=1, dtype=int32)

M_rowadr

index of each row in qM (nv,)

Type:

array(ndim=1, dtype=int32)

M_colind

column indices of non-zeros in qM (nM,)

Type:

array(ndim=1, dtype=int32)

mapM2M

index mapping from M (legacy) to M (CSR) (nC)

Type:

array(ndim=1, dtype=int32)

nbranch

number of branches (leaf-to-root paths)

Type:

int

nv_pad

number of degrees of freedom + padding

Type:

int

nacttrnbody

number of actuators with body transmission

Type:

int

nsensorcollision

number of unique collisions for geom distance sensors

Type:

int

nsensortaxel

number of taxels in all tactile sensors

Type:

int

nsensorcontact

number of contact sensors

Type:

int

nrangefinder

number of rangefinder sensors

Type:

int

nmaxcondim

maximum condim for geoms

Type:

int

nmaxpyramid

maximum number of pyramid directions

Type:

int

nmaxpolygon

maximum number of verts per polygon

Type:

int

nmaxmeshdeg

maximum number of polygons per vert

Type:

int

is_sparse

whether to use sparse representations

Type:

bool

has_fluid

True if wind, density, or viscosity are non-zero at put_model time

Type:

bool

has_sdf_geom

whether the model contains SDF geoms

Type:

bool

block_dim

block dim options

Type:

mujoco_warp._src.types.BlockDim

body_tree

list of body ids by tree level

Type:

tuple[array(ndim=1, dtype=int32), …]

body_branches

flattened body ids for all branches

Type:

array(ndim=1, dtype=int32)

body_branch_start

start index in body_branches for each branch (nbranch + 1,)

Type:

array(ndim=1, dtype=int32)

mocap_bodyid

id of body for mocap (nmocap,)

Type:

array(ndim=1, dtype=int32)

body_fluid_ellipsoid

does body use ellipsoid fluid (nbody,)

Type:

array(ndim=1, dtype=warp._src.types.bool)

jnt_limited_slide_hinge_adr

limited/slide/hinge jntadr

Type:

array(ndim=1, dtype=int32)

jnt_limited_ball_adr

limited/ball jntadr

Type:

array(ndim=1, dtype=int32)

dof_tri_row

dof lower triangle row (used in solver)

Type:

array(ndim=1, dtype=int32)

dof_tri_col

dof lower triangle col (used in solver)

Type:

array(ndim=1, dtype=int32)

nxn_geom_pair

collision pair geom ids [-2, ngeom-1]

Type:

array(ndim=1, dtype=vec2i)

nxn_geom_pair_filtered

valid collision pair geom ids [-1, ngeom - 1]

Type:

array(ndim=1, dtype=vec2i)

nxn_pairid

contact pair id, -1 if not predefined,

-2 if skipped

collision id, else -1

Type:

array(ndim=1, dtype=vec2i)

nxn_pairid_filtered

active subset of nxn_pairid

Type:

array(ndim=1, dtype=vec2i)

geom_pair_type_count

count of max number of each potential collision

Type:

tuple[int, …]

geom_plugin_index

geom index in plugin array (ngeom,)

Type:

array(ndim=1, dtype=int32)

eq_connect_adr

eq_* addresses of type CONNECT

Type:

array(ndim=1, dtype=int32)

eq_wld_adr

eq_* addresses of type WELD

Type:

array(ndim=1, dtype=int32)

eq_jnt_adr

eq_* addresses of type JOINT

Type:

array(ndim=1, dtype=int32)

eq_ten_adr

eq_* addresses of type TENDON

Type:

array(ndim=1, dtype=int32)

eq_flex_adr

eq * addresses of type `FLEX

Type:

array(ndim=1, dtype=int32)

tendon_jnt_adr

joint tendon address

Type:

array(ndim=1, dtype=int32)

tendon_site_pair_adr

site pair tendon address

Type:

array(ndim=1, dtype=int32)

tendon_geom_adr

geom tendon address

Type:

array(ndim=1, dtype=int32)

tendon_limited_adr

addresses for limited tendons

Type:

array(ndim=1, dtype=int32)

ten_wrapadr_site

wrap object starting address for sites

Type:

array(ndim=1, dtype=int32)

ten_wrapnum_site

number of site wrap objects per tendon

Type:

array(ndim=1, dtype=int32)

wrap_jnt_adr

addresses for joint tendon wrap object

Type:

array(ndim=1, dtype=int32)

wrap_site_adr

addresses for site tendon wrap object

Type:

array(ndim=1, dtype=int32)

wrap_site_pair_adr

first address for site wrap pair

Type:

array(ndim=1, dtype=int32)

wrap_geom_adr

addresses for geom tendon wrap object

Type:

array(ndim=1, dtype=int32)

wrap_pulley_scale

pulley scaling (nwrap,)

Type:

array(ndim=1, dtype=float32)

actuator_trntype_body_adr

addresses for actuators with body transmission

Type:

array(ndim=1, dtype=int32)

sensor_pos_adr

addresses for position sensors

Type:

array(ndim=1, dtype=int32)

sensor_limitpos_adr

address for limit position sensors

Type:

array(ndim=1, dtype=int32)

sensor_vel_adr

addresses for velocity sensors (excluding limit velocity sensors)

Type:

array(ndim=1, dtype=int32)

sensor_limitvel_adr

address for limit velocity sensors

Type:

array(ndim=1, dtype=int32)

sensor_acc_adr

addresses for acceleration sensors

Type:

array(ndim=1, dtype=int32)

sensor_rangefinder_adr

addresses for rangefinder sensors

Type:

array(ndim=1, dtype=int32)

rangefinder_sensor_adr

map sensor id to rangefinder id (excluding touch sensors) (excluding limit force sensors)

Type:

array(ndim=1, dtype=int32)

sensor_collision_start_adr

address for sensor’s first item in collision

Type:

array(ndim=1, dtype=int32)

collision_sensor_adr

map sensor id to collision id (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_touch_adr

addresses for touch sensors

Type:

array(ndim=1, dtype=int32)

sensor_limitfrc_adr

address for limit force sensors

Type:

array(ndim=1, dtype=int32)

sensor_e_potential

evaluate energy_pos

Type:

bool

sensor_e_kinetic

evaluate energy_vel

Type:

bool

sensor_tendonactfrc_adr

address for tendonactfrc sensor

Type:

array(ndim=1, dtype=int32)

sensor_subtree_vel

evaluate subtree_vel

Type:

bool

sensor_contact_adr

addresses for contact sensors (nsensorcontact,)

Type:

array(ndim=1, dtype=int32)

sensor_adr_to_contact_adr

map sensor adr to contact adr (nsensor,)

Type:

array(ndim=1, dtype=int32)

sensor_rne_postconstraint

evaluate rne_postconstraint

Type:

bool

sensor_rangefinder_bodyid

bodyid for rangefinder (nrangefinder,)

Type:

array(ndim=1, dtype=int32)

taxel_vertadr

tactile sensor vertex address (nsensortaxel,)

Type:

array(ndim=1, dtype=int32)

taxel_sensorid

address for tactile sensors

Type:

array(ndim=1, dtype=int32)

qM_tiles

tiling configuration

Type:

tuple[mujoco_warp._src.types.TileSet, …]

qLD_updates

tuple of index triples for sparse factorization

Type:

tuple[array(ndim=1, dtype=vec3i), …]

qM_fullm_i

sparse mass matrix addressing

Type:

array(ndim=1, dtype=int32)

qM_fullm_j

sparse mass matrix addressing

Type:

array(ndim=1, dtype=int32)

qM_mulm_rowadr

sparse matmul row pointers

Type:

array(ndim=1, dtype=int32)

qM_mulm_col

sparse matmul column indices

Type:

array(ndim=1, dtype=int32)

qM_mulm_madr

sparse matmul matrix addresses

Type:

array(ndim=1, dtype=int32)

class Data

Dynamic state that updates each step.

solver_niter

number of solver iterations (nworld,)

Type:

array(ndim=1, dtype=int32)

ne

number of equality constraints (nworld,)

Type:

array(ndim=1, dtype=int32)

nf

number of friction constraints (nworld,)

Type:

array(ndim=1, dtype=int32)

nl

number of limit constraints (nworld,)

Type:

array(ndim=1, dtype=int32)

nefc

number of constraints (nworld,)

Type:

array(ndim=1, dtype=int32)

time

simulation time (nworld,)

Type:

array(ndim=1, dtype=float32)

energy

potential, kinetic energy (nworld, 2)

Type:

array(ndim=1, dtype=vec2f)

qpos

position (nworld, nq)

Type:

array(ndim=2, dtype=float32)

qvel

velocity (nworld, nv)

Type:

array(ndim=2, dtype=float32)

act

actuator activation (nworld, na)

Type:

array(ndim=2, dtype=float32)

qacc_warmstart

acceleration used for warmstart (nworld, nv)

Type:

array(ndim=2, dtype=float32)

ctrl

control (nworld, nu)

Type:

array(ndim=2, dtype=float32)

qfrc_applied

applied generalized force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

xfrc_applied

applied Cartesian force/torque (nworld, nbody, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

eq_active

enable/disable constraints (nworld, neq)

Type:

array(ndim=2, dtype=warp._src.types.bool)

mocap_pos

position of mocap bodies (nworld, nmocap, 3)

Type:

array(ndim=2, dtype=vec3f)

mocap_quat

orientation of mocap bodies (nworld, nmocap, 4)

Type:

array(ndim=2, dtype=quatf)

qacc

acceleration (nworld, nv)

Type:

array(ndim=2, dtype=float32)

act_dot

time-derivative of actuator activation (nworld, na)

Type:

array(ndim=2, dtype=float32)

sensordata

sensor data array (nworld, nsensordata,)

Type:

array(ndim=2, dtype=float32)

xpos

Cartesian position of body frame (nworld, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

xquat

Cartesian orientation of body frame (nworld, nbody, 4)

Type:

array(ndim=2, dtype=quatf)

xmat

Cartesian orientation of body frame (nworld, nbody, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

xipos

Cartesian position of body com (nworld, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

ximat

Cartesian orientation of body inertia (nworld, nbody, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

xanchor

Cartesian position of joint anchor (nworld, njnt, 3)

Type:

array(ndim=2, dtype=vec3f)

xaxis

Cartesian joint axis (nworld, njnt, 3)

Type:

array(ndim=2, dtype=vec3f)

geom_xpos

Cartesian geom position (nworld, ngeom, 3)

Type:

array(ndim=2, dtype=vec3f)

geom_xmat

Cartesian geom orientation (nworld, ngeom, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

site_xpos

Cartesian site position (nworld, nsite, 3)

Type:

array(ndim=2, dtype=vec3f)

site_xmat

Cartesian site orientation (nworld, nsite, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

cam_xpos

Cartesian camera position (nworld, ncam, 3)

Type:

array(ndim=2, dtype=vec3f)

cam_xmat

Cartesian camera orientation (nworld, ncam, 3, 3)

Type:

array(ndim=2, dtype=mat33(f))

light_xpos

Cartesian light position (nworld, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

light_xdir

Cartesian light direction (nworld, nlight, 3)

Type:

array(ndim=2, dtype=vec3f)

subtree_com

center of mass of each subtree (nworld, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

cdof

com-based motion axis of each dof (rot:lin) (nworld, nv, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

cinert

com-based body inertia and mass (nworld, nbody, 10)

Type:

array(ndim=2, dtype=vector(length=10, dtype=float32))

flexvert_xpos

cartesian flex vertex positions (nworld, nflexvert, 3)

Type:

array(ndim=2, dtype=vec3f)

flexedge_J

edge length Jacobian (nworld, 1, nflexedge*6)

Type:

array(ndim=3, dtype=float32)

flexedge_length

flex edge lengths (nworld, nflexedge, 1)

Type:

array(ndim=2, dtype=float32)

ten_wrapadr

start address of tendon’s path (nworld, ntendon)

Type:

array(ndim=2, dtype=int32)

ten_wrapnum

number of wrap points in path (nworld, ntendon)

Type:

array(ndim=2, dtype=int32)

ten_J

tendon Jacobian (nworld, ntendon, nv)

Type:

array(ndim=3, dtype=float32)

ten_length

tendon lengths (nworld, ntendon)

Type:

array(ndim=2, dtype=float32)

wrap_obj

geomid; -1: site; -2: pulley (nworld, nwrap, 2)

Type:

array(ndim=2, dtype=vec2i)

wrap_xpos

Cartesian 3D points in all paths (nworld, nwrap, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

actuator_length

actuator lengths (nworld, nu)

Type:

array(ndim=2, dtype=float32)

actuator_moment

actuator moments (nworld, nu, nv)

Type:

array(ndim=3, dtype=float32)

crb

com-based composite inertia and mass (nworld, nbody, 10)

Type:

array(ndim=2, dtype=vector(length=10, dtype=float32))

qM

total inertia (nworld, nv, nv) if dense (nworld, 1, nM) if sparse

Type:

array(ndim=3, dtype=float32)

qLD

L’*D*L factorization of M (nworld, nv, nv) if dense (nworld, 1, nC) if sparse

Type:

array(ndim=3, dtype=float32)

qLDiagInv

1/diag(D) (nworld, nv)

Type:

array(ndim=2, dtype=float32)

flexedge_velocity

flex edge velocities (nworld, nflexedge)

Type:

array(ndim=2, dtype=float32)

ten_velocity

tendon velocities (nworld, ntendon)

Type:

array(ndim=2, dtype=float32)

actuator_velocity

actuator velocities (nworld, nu)

Type:

array(ndim=2, dtype=float32)

cvel

com-based velocity (rot:lin) (nworld, nbody, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

cdof_dot

time-derivative of cdof (rot:lin) (nworld, nv, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

qfrc_bias

C(qpos,qvel) (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_spring

passive spring force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_damper

passive damper force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_gravcomp

passive gravity compensation force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_fluid

passive fluid force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_passive

total passive force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

subtree_linvel

linear velocity of subtree com (nworld, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

subtree_angmom

angular momentum about subtree com (nworld, nbody, 3)

Type:

array(ndim=2, dtype=vec3f)

actuator_force

actuator force in actuation space (nworld, nu)

Type:

array(ndim=2, dtype=float32)

qfrc_actuator

actuator force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_smooth

net unconstrained force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qacc_smooth

unconstrained acceleration (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_constraint

constraint force (nworld, nv)

Type:

array(ndim=2, dtype=float32)

qfrc_inverse

net external force; should equal: (nworld, nv) qfrc_applied + J.T @ xfrc_applied + qfrc_actuator

Type:

array(ndim=2, dtype=float32)

cacc

com-based acceleration (nworld, nbody, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

cfrc_int

com-based interaction force with parent (nworld, nbody, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

cfrc_ext

com-based external force on body (nworld, nbody, 6)

Type:

array(ndim=2, dtype=vector(length=6, dtype=float32))

contact

contact data

Type:

mujoco_warp._src.types.Contact

efc

constraint data

Type:

mujoco_warp._src.types.Constraint

nworld

number of worlds

Type:

int

naconmax

maximum number of contacts (shared across all worlds)

Type:

int

naccdmax

maximum number of contacts for CCD (all worlds)

Type:

int

njmax

maximum number of constraints per world

Type:

int

nacon

number of detected contacts (across all worlds) (1,)

Type:

array(ndim=1, dtype=int32)

ncollision

collision count from broadphase (1,)

Type:

array(ndim=1, dtype=int32)

refit_bvh(m: Model, d: Data, rc: RenderContext)

Refit the dynamic BVH structures in the render context.

collision(m: Model, d: Data)

Runs the full collision detection pipeline.

This function orchestrates the broadphase and narrowphase collision detection stages. It first identifies potential collision pairs using a broadphase algorithm (either N-squared or Sweep-and-Prune, based on m.opt.broadphase). Then, for each potential pair, it performs narrowphase collision detection to compute detailed contact information like distance, position, and frame.

The results are used to populate the d.contact array, and the total number of contacts is stored in d.nacon. If d.nacon is larger than d.naconmax then an overflow has occurred and the remaining contacts will be skipped. If this happens, raise the nconmax parameter in io.make_data or io.put_data.

This function will do nothing except zero out arrays if collision detection is disabled via m.opt.disableflags or if d.nacon is 0.

nxn_broadphase(m: Model, d: Data, ctx: CollisionContext)

Runs broadphase collision detection using a brute-force N-squared approach.

This function iterates through a pre-filtered list of all possible geometry pairs and performs a quick bounding sphere check to identify potential collisions.

For each pair that passes the sphere check, it populates the collision arrays in d (d.collision_pair, d.collision_pairid, etc.), which are then consumed by the narrowphase.

The initial list of pairs is filtered at model creation time to exclude pairs based on contype/conaffinity, parent-child relationships, and explicit <exclude> tags.

sap_broadphase(m: Model, d: Data, ctx: CollisionContext)

Runs broadphase collision detection using a sweep-and-prune (SAP) algorithm.

This method is more efficient than the N-squared approach for large numbers of objects. It works by projecting the bounding spheres of all geoms onto a single axis and sorting them. It then sweeps along the axis, only checking for overlaps between geoms whose projections are close to each other.

For each potentially colliding pair identified by the sweep, a more precise bounding sphere check is performed. If this check passes, the pair is added to the collision arrays in d for the narrowphase stage.

Two sorting strategies are supported, controlled by m.opt.broadphase

• SAP_TILE: Uses a tile-based sort.

• SAP_SEGMENTED: Uses a segmented sort.

primitive_narrowphase(m: Model, d: Data, ctx: CollisionContext, collision_table: list[tuple[GeomType, GeomType]])

Runs collision detection on primitive geom pairs discovered during broadphase.

This function processes collision pairs involving primitive shapes that were identified during the broadphase stage. It computes detailed contact information such as distance, position, and frame, and populates the d.contact array.

The primitive geom types: PLANE, SPHERE, CAPSULE, CYLINDER, and BOX.

Additionally, collisions between planes and convex hulls.

To improve performance, it dynamically builds and launches a kernel tailored to the specific primitive collision types present in the model, avoiding unnecessary checks for non-existent collision pairs.

make_constraint(m: Model, d: Data)

Creates constraint jacobians and other supporting data.

deriv_smooth_vel(m: Model, d: Data, out: array(ndim=2, dtype=float32))

Analytical derivative of smooth forces w.r.t. velocities.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• out – qM - dt * qDeriv (derivatives of smooth forces w.r.t velocities).

euler(m: Model, d: Data)

Euler integrator, semi-implicit in velocity.

forward(m: Model, d: Data)

Forward dynamics.

fwd_acceleration(m: Model, d: Data, factorize: bool = False)

Add up all non-constraint forces, compute qacc_smooth.

Parameters:

• m – The model containing kinematic and dynamic information.

• d – The data object containing the current state and output arrays.

• factorize – Flag to factorize inertia matrix.

fwd_actuation(m: Model, d: Data)

Actuation-dependent computations.

fwd_position(m: Model, d: Data, factorize: bool = True)

Position-dependent computations.

Parameters:

• m – The model containing kinematic and dynamic information.

• d – The data object containing the current state and output arrays.

• factorize – Flag to factorize interia matrix.

fwd_velocity(m: Model, d: Data)

Velocity-dependent computations.

implicit(m: Model, d: Data)

Integrates fully implicit in velocity.

rungekutta4(m: Model, d: Data)

Runge-Kutta explicit order 4 integrator.

step1(m: Model, d: Data)

Advance simulation in two phases: before input is set by user.

step2(m: Model, d: Data)

Advance simulation in two phases: after input is set by user.

inverse(m: Model, d: Data)

Inverse dynamics.

create_render_context(mjm: MjModel, nworld: int = 1, cam_res: list[tuple[int, int]] | tuple[int, int] | None = None, render_rgb: list[bool] | bool | None = None, render_depth: list[bool] | bool | None = None, use_textures: bool = True, use_shadows: bool = False, enabled_geom_groups: list[int] = [0, 1, 2], cam_active: list[bool] | None = None, flex_render_smooth: bool = True, use_precomputed_rays: bool = True) → RenderContext

Creates a render context on device.

Parameters:

• mjm – The model containing kinematic and dynamic information on host.

• nworld – The number of worlds.

• cam_res – The width and height to render each camera image. If None, uses the MuJoCo model values.

• render_rgb – Whether to render RGB images. If None, uses the MuJoCo model values.

• render_depth – Whether to render depth images. If None, uses the MuJoCo model values.

• use_textures – Whether to use textures.

• use_shadows – Whether to use shadows.

• enabled_geom_groups – The geom groups to render.

• cam_active – List of booleans indicating which cameras to include in rendering. If None, all cameras are included.

• flex_render_smooth – Whether to render flex meshes smoothly.

• use_precomputed_rays – Use precomputed rays instead of computing during rendering. When using domain randomization for camera intrinsics, set to False.

Returns:

The render context containing rendering fields and output arrays on device.

get_data_into(result: MjData, mjm: MjModel, d: Data, world_id: int = 0)

Gets data from a device into an existing mujoco.MjData.

Parameters:

• result – The data object containing the current state and output arrays (host).

• mjm – The model containing kinematic and dynamic information (host).

• d – The data object containing the current state and output arrays (device).

• world_id – The id of the world to get the data from.

make_data(mjm: MjModel, nworld: int = 1, nconmax: int | None = None, nccdmax: int | None = None, njmax: int | None = None, naconmax: int | None = None, naccdmax: int | None = None) → Data

Creates a data object on device.

Parameters:

• mjm – The model containing kinematic and dynamic information (host).

• nworld – Number of worlds.

• nconmax – Number of contacts to allocate per world. Contacts exist in large heterogeneous arrays: one world may have more than nconmax contacts.

• nccdmax – Number of CCD contacts to allocate per world. Same semantics as nconmax.

• njmax – Number of constraints to allocate per world. Constraint arrays are batched by world: no world may have more than njmax constraints.

• naconmax – Number of contacts to allocate for all worlds. Overrides nconmax.

• naccdmax – Maximum number of CCD contacts. Defaults to naconmax.

Returns:

The data object containing the current state and output arrays (device).

put_data(mjm: MjModel, mjd: MjData, nworld: int = 1, nconmax: int | None = None, nccdmax: int | None = None, njmax: int | None = None, naconmax: int | None = None, naccdmax: int | None = None) → Data

Moves data from host to a device.

Parameters:

• mjm – The model containing kinematic and dynamic information (host).

• mjd – The data object containing current state and output arrays (host).

• nworld – The number of worlds.

• nconmax – Number of contacts to allocate per world. Contacts exist in large heterogenous arrays: one world may have more than nconmax contacts.

• nccdmax – Number of CCD contacts to allocate per world. Same semantics as nconmax.

• njmax – Number of constraints to allocate per world. Constraint arrays are batched by world: no world may have more than njmax constraints.

• naconmax – Number of contacts to allocate for all worlds. Overrides nconmax.

• naccdmax – Maximum number of CCD contacts. Defaults to naconmax.

Returns:

The data object containing the current state and output arrays (device).

put_model(mjm: MjModel) → Model

Creates a model on device.

Parameters:

mjm – The model containing kinematic and dynamic information (host).

Returns:

The model containing kinematic and dynamic information (device).

reset_data(m: Model, d: Data, reset: array | None = None)

Clear data, set defaults; optionally by world.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• reset – Per-world bitmask. Reset if True.

set_const(m: Model, d: Data)

Recomputes qpos0-dependent constant model fields.

This function propagates changes from some model fields to derived fields, allowing modifications that would otherwise be unsafe. It should be called after modifying model parameters at runtime.

Model fields that can be modified safely with set_const:

Field | Notes ———————————|———————————————- qpos0, qpos_spring | body_mass, body_inertia, | Mass and inertia are usually scaled together body_ipos, body_iquat | since inertia is sum(m * r^2). body_pos, body_quat | Unsafe for static bodies (invalidates BVH). body_gravcomp | If changing from 0 to >0 bodies, required. dof_armature | eq_data | For connect/weld, offsets computed if not set. hfield_size | tendon_stiffness, tendon_damping | Only if changing from/to zero. actuator_gainprm, actuator_biasprm | For position actuators with dampratio.

For selective updates, use the sub-functions directly based on what changed:

Modified Field | Call —————-|—————— body_mass | set_const body_gravcomp | set_const_fixed body_inertia | set_const_0 qpos0 | set_const_0

Computes:

• Fixed quantities (via set_const_fixed): - body_subtreemass: mass of body and all descendants - ngravcomp: count of bodies with gravity compensation

• qpos0-dependent quantities (via set_const_0): - tendon_length0: tendon resting lengths - dof_invweight0: inverse inertia for DOFs - body_invweight0: inverse spatial inertia for bodies - tendon_invweight0: inverse weight for tendons - cam_pos0, cam_poscom0, cam_mat0: camera references - light_pos0, light_poscom0, light_dir0: light references - actuator_acc0: acceleration from unit actuator force

Skips: dof_M0, actuator_length0 (not in mjwarp).

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

set_const_0(m: Model, d: Data)

Compute quantities that depend on qpos0.

Computes:

• tendon_length0: tendon resting lengths

• dof_invweight0: inverse inertia for DOFs

• body_invweight0: inverse spatial inertia for bodies

• tendon_invweight0: inverse weight for tendons

• cam_pos0, cam_poscom0, cam_mat0: camera references

• light_pos0, light_poscom0, light_dir0: light references

• actuator_acc0: acceleration from unit actuator force

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

set_const_fixed(m: Model, d: Data)

Compute fixed quantities (independent of qpos0).

Computes:

• body_subtreemass: mass of body and all descendants (depends on body_mass)

• ngravcomp: count of bodies with gravity compensation (depends on body_gravcomp)

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

passive(m: Model, d: Data)

Adds all passive forces.

ray(m: Model, d: Data, pnt: array(ndim=2, dtype=vec3f), vec: array(ndim=2, dtype=vec3f), geomgroup: vec6f | None = None, flg_static: bool = True, bodyexclude: int = -1, rc: RenderContext | None = None) → Tuple[array, array, array]

Returns the distance at which rays intersect with primitive geoms.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• pnt – Ray origin points.

• vec – Ray directions.

• geomgroup – Group inclusion/exclusion mask.

• flg_static – If True, allows rays to intersect with static geoms.

• bodyexclude – Ignore geoms on specified body id (-1 to disable).

• rc – Optional Render context containing BVH information for BVH accelerated ray intersections.

Returns:

Distances from ray origins to geom surfaces, IDs of intersected geoms (-1 if none), and normals at intersection points.

rays(m: Model, d: Data, pnt: array(ndim=2, dtype=vec3f), vec: array(ndim=2, dtype=vec3f), geomgroup: vec6f, flg_static: bool, bodyexclude: array(ndim=1, dtype=int32), dist: array(ndim=2, dtype=float32), geomid: array(ndim=2, dtype=int32), normal: array(ndim=2, dtype=vec3f), rc: RenderContext | None = None)

Ray intersection for multiple worlds and multiple rays.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• pnt – Ray origin points, shape (nworld, nray).

• vec – Ray directions, shape (nworld, nray).

• geomgroup – Group inclusion/exclusion mask. Set all elements to -1 to ignore.

• flg_static – If True, allows rays to intersect with static geoms.

• bodyexclude – Per-ray body exclusion array of shape (nray,). Geoms on the specified body ids are ignored (-1 to disable for that ray).

• dist – Output array for distances from ray origins to geom surfaces, shape (nworld, nray). -1 indicates no intersection.

• geomid – Output array for IDs of intersected geoms, shape (nworld, nray). -1 indicates no intersection.

• normal – Output array for normals at intersection points, shape (nworld, nray).

• rc – Optional Render context containing BVH information for BVH accelerated ray intersections.

render(m: Model, d: Data, rc: RenderContext)

Render the current frame.

Outputs are stored in buffers within the render context.

Parameters:

• m – The model on device.

• d – The data on device.

• rc – The render context on device.

get_depth(rc: RenderContext, camera_index: int, depth_scale: float, depth_out: array(ndim=3, dtype=float32))

Get the depth data output from the render context buffers for a given camera index.

Parameters:

• rc – The render context on device.

• camera_index – The index of the camera to get the depth data for.

• depth_scale – The scale factor to apply to the depth data.

• depth_out – The output array to store the scaled and clamped depth data in with shape (nworld, height, width).

get_rgb(rc: RenderContext, camera_index: int, rgb_out: array(ndim=3, dtype=vec3f))

Get the RGB data output from the render context buffers for a given camera index.

Parameters:

• rc – The render context on device.

• camera_index – The index of the camera to get the RGB data for.

• rgb_out – The output array to store the RGB data in, with shape (nworld, height, width).

energy_pos(m: Model, d: Data)

Position-dependent energy (potential).

energy_vel(m: Model, d: Data)

Velocity-dependent energy (kinetic).

sensor_acc(m: Model, d: Data)

Compute acceleration-dependent sensor values.

sensor_pos(m: Model, d: Data)

Compute position-dependent sensor values.

sensor_vel(m: Model, d: Data)

Compute velocity-dependent sensor values.

camlight(m: Model, d: Data)

Computes camera and light positions and orientations.

Updates the global positions and orientations for all cameras and lights in the model, including special handling for tracking and target modes.

com_pos(m: Model, d: Data)

Computes subtree center of mass positions.

Transforms inertia and motion to global frame centered at subtree CoM. Accumulates the mass-weighted positions up the kinematic tree, divides by total mass, and computes composite inertias and motion degrees of freedom in the subtree CoM frame.

com_vel(m: Model, d: Data)

Computes the spatial velocities (cvel) and the derivative cdof_dot for all bodies.

Propagates velocities down the kinematic tree, updating the spatial velocity and derivative for each body.

crb(m: Model, d: Data)

Computes composite rigid body inertias for each body and the joint-space inertia matrix.

Accumulates composite rigid body inertias up the kinematic tree and computes the joint-space inertia matrix in either sparse or dense format, depending on model options.

factor_m(m: Model, d: Data)

Factorization of inertia-like matrix M, assumed spd.

kinematics(m: Model, d: Data)

Computes forward kinematics for all bodies, sites, geoms, and flexible elements.

This function updates the global positions and orientations of all bodies, as well as the derived positions and orientations of geoms, sites, and flexible elements, based on the current joint positions and any attached mocap bodies.

rne(m: Model, d: Data, flg_acc: bool = False)

Computes inverse dynamics using the recursive Newton-Euler algorithm.

Computes the bias forces (qfrc_bias) and internal forces (cfrc_int) for the current state, including the effects of gravity and optionally joint accelerations.

Parameters:

• m – The model containing kinematic and dynamic information.

• d – The data object containing the current state and output arrays.

• flg_acc – If True, includes joint accelerations in the computation.

rne_postconstraint(m: Model, d: Data)

Computes the recursive Newton-Euler algorithm after constraints are applied.

Computes cacc, cfrc_ext, and cfrc_int, including the effects of applied forces, equality constraints, and contacts.

solve_m(m: Model, d: Data, x: array(ndim=2, dtype=float32), y: array(ndim=2, dtype=float32))

Computes backsubstitution: x = qLD * y.

Parameters:

• m – The model containing inertia and factorization information.

• d – The data object containing factorization results.

• x – Output array for the solution.

• y – Input right-hand side array.

subtree_vel(m: Model, d: Data)

Computes subtree linear velocity and angular momentum.

Computes the linear momentum and angular momentum for each subtree, accumulating contributions up the kinematic tree.

tendon(m: Model, d: Data)

Computes tendon lengths and moments.

Updates the tendon length and moment arrays for all tendons in the model, including joint, site, and geom tendons.

transmission(m: Model, d: Data)

Computes actuator/transmission lengths and moments.

Updates the actuator length and moments for all actuators in the model, including joint and tendon transmissions.

contact_force(m: Model, d: Data, contact_ids: array(ndim=1, dtype=int32), to_world_frame: bool, force: array(ndim=1, dtype=vector(length=6, dtype=float32)))

Compute forces for contacts in Data.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• contact_ids – IDs for each contact.

• to_world_frame – If True, map force from contact to world frame.

• force – Contact forces.

get_state(m: Model, d: Data, state: array(ndim=2, dtype=float32), sig: int, active: array | None = None)

Copy concatenated state components specified by sig from Data into state.

The bits of the integer sig correspond to element fields of State.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output information (device).

• state – Concatenation of state components.

• sig – Bitflag specifying state components.

• active – Per-world bitmask for getting state.

jac(m: Model, d: Data, jacp: array | None, jacr: array | None, point: array(ndim=1, dtype=vec3f), body: array(ndim=1, dtype=int32))

Compute translational and rotational Jacobian for point on body.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state (device).

• jacp – Output translational Jacobian (optional).

• jacr – Output rotational Jacobian (optional).

• point – 3D point in global coordinates.

• body – Body ID for each world.

mul_m(m: Model, d: Data, res: array(ndim=2, dtype=float32), vec: array(ndim=2, dtype=float32), skip: array | None = None, M: array | None = None)

Multiply vectors by inertia matrix; optionally skip per world.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• res – Result: qM @ vec.

• vec – Input vector to multiply by qM.

• skip – Per-world bitmask to skip computing output.

• M – Input matrix: M @ vec.

set_state(m: Model, d: Data, state: array(ndim=2, dtype=float32), sig: int, active: array | None = None)

Copy concatenated state components specified by sig from state into Data.

The bits of the integer sig correspond to element fields of State.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output information (device).

• state – Concatenation of state components.

• sig – Bitflag specifying state components.

• active – Per-world bitmask for setting state.

xfrc_accumulate(m: Model, d: Data, qfrc: array(ndim=2, dtype=float32))

Map applied forces at each body via Jacobians to dof space and accumulate.

Parameters:

• m – The model containing kinematic and dynamic information (device).

• d – The data object containing the current state and output arrays (device).

• qfrc – Total applied force mapped to dof space.

class BiasType

Type of actuator bias.

NONE

no bias

AFFINE

const + kp*length + kv*velocity

MUSCLE

muscle passive force computed by muscle_bias

class BroadphaseFilter

Bitmask specifying which collision functions to run during broadphase.

PLANE

collision between bounding sphere and plane

SPHERE

collision between bounding spheres

AABB

collision between axis-aligned bounding boxes

OBB

collision between oriented bounding boxes

class BroadphaseType

Type of broadphase algorithm.

NXN

Broad phase checking all pairs

SAP_TILE

Sweep and prune broad phase using tile sort

SAP_SEGMENTED

Sweep and prune broad phase using segment sort

class ConeType

Type of friction cone.

PYRAMIDAL

pyramidal

ELLIPTIC

elliptic

class Constraint

Constraint data.

type

constraint type (ConstraintType) (nworld, njmax)

Type:

array(ndim=2, dtype=int32)

id

id of object of specific type (nworld, njmax)

Type:

array(ndim=2, dtype=int32)

J

constraint Jacobian (nworld, njmax_pad, nv_pad)

Type:

array(ndim=3, dtype=float32)

pos

constraint position (equality, contact) (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

margin

inclusion margin (contact) (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

D

constraint mass (nworld, njmax_pad)

Type:

array(ndim=2, dtype=float32)

vel

velocity in constraint space: J*qvel (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

aref

reference pseudo-acceleration (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

frictionloss

frictionloss (friction) (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

force

constraint force in constraint space (nworld, njmax)

Type:

array(ndim=2, dtype=float32)

state

constraint state (nworld, njmax_pad)

Type:

array(ndim=2, dtype=int32)

Ma

M*qacc (nworld, nv)

Type:

array(ndim=2, dtype=float32)

class Contact

Contact data.

dist

distance between nearest points; neg: penetration (naconmax,)

Type:

array(ndim=1, dtype=float32)

pos

position of contact point: midpoint between geoms (naconmax, 3)

Type:

array(ndim=1, dtype=vec3f)

frame

normal is in [0-2], points from geom[0] to geom[1] (naconmax, 3, 3)

Type:

array(ndim=1, dtype=mat33(f))

includemargin

include if dist<includemargin=margin-gap (naconmax,)

Type:

array(ndim=1, dtype=float32)

friction

tangent1, 2, spin, roll1, 2 (naconmax, 5)

Type:

array(ndim=1, dtype=vector(length=5, dtype=float32))

solref

constraint solver reference, normal direction (naconmax, 2)

Type:

array(ndim=1, dtype=vec2f)

solreffriction

constraint solver reference, friction directions (naconmax, 2)

Type:

array(ndim=1, dtype=vec2f)

solimp

constraint solver impedance (naconmax, 5)

Type:

array(ndim=1, dtype=vector(length=5, dtype=float32))

dim

contact space dimensionality: 1, 3, 4 or 6 (naconmax,)

Type:

array(ndim=1, dtype=int32)

geom

geom ids; -1 for flex (naconmax, 2)

Type:

array(ndim=1, dtype=vec2i)

efc_address

address in efc; -1: not included (naconmax, nmaxpyramid)

Type:

array(ndim=2, dtype=int32)

worldid

world id (naconmax,)

Type:

array(ndim=1, dtype=int32)

type

ContactType (naconmax,)

Type:

array(ndim=1, dtype=int32)

geomcollisionid

i-th contact generated for geom (naconmax,) helps uniquely identity contact when multiple contacts are generated for geom pair

Type:

array(ndim=1, dtype=int32)

class DisableBit

Disable default feature bitflags.

CONSTRAINT

entire constraint solver

EQUALITY

equality constraints

FRICTIONLOSS

joint and tendon frictionloss constraints

LIMIT

joint and tendon limit constraints

CONTACT

contact constraints

SPRING

passive spring forces

DAMPER

passive damper forces

GRAVITY

gravitational forces

CLAMPCTRL

clamp control to specified range

WARMSTART

warmstart constraint solver

FILTERPARENT

disable collisions between parent and child bodies

ACTUATION

apply actuation forces

REFSAFE

integrator safety: make ref[0]>=2*timestep

SENSOR

sensors

EULERDAMP

implicit damping for Euler integration

NATIVECCD

native convex collision detection (ignored in MJWarp)

class DynType

Type of actuator dynamics.

NONE

no internal dynamics; ctrl specifies force

INTEGRATOR

integrator: da/dt = u

FILTER

linear filter: da/dt = (u-a) / tau

FILTEREXACT

linear filter: da/dt = (u-a) / tau, with exact integration

MUSCLE

piece-wise linear filter with two time constants

class EnableBit

Enable optional feature bitflags.

ENERGY

energy computation

INVDISCRETE

discrete-time inverse dynamics

MULTICCD

multiple contacts with CCD

class GainType

Type of actuator gain.

FIXED

fixed gain

AFFINE

const + kp*length + kv*velocity

MUSCLE

muscle FLV curve computed by muscle_gain

class GeomType

Type of geometry.

PLANE

plane

HFIELD

heightfield

SPHERE

sphere

CAPSULE

capsule

ELLIPSOID

ellipsoid

CYLINDER

cylinder

BOX

box

MESH

mesh

SDF

sdf

class IntegratorType

Integrator mode.

EULER

semi-implicit Euler

RK4

4th-order Runge Kutta

IMPLICITFAST

implicit in velocity, no rne derivative

class JointType

Type of degree of freedom.

FREE

global position and orientation (quat) (7,)

BALL

orientation (quat) relative to parent (4,)

SLIDE

sliding distance along body-fixed axis (1,)

HINGE

rotation angle (rad) around body-fixed axis (1,)

class Option

Physics options.

timestep

simulation timestep

Type:

array(ndim=1, dtype=float32)

tolerance

main solver tolerance

Type:

array(ndim=1, dtype=float32)

ls_tolerance

CG/Newton linesearch tolerance

Type:

array(ndim=1, dtype=float32)

ccd_tolerance

convex collision detection tolerance

Type:

array(ndim=1, dtype=float32)

gravity

gravitational acceleration

Type:

array(ndim=1, dtype=vec3f)

wind

wind (for lift, drag, and viscosity)

Type:

array(ndim=1, dtype=vec3f)

magnetic

global magnetic flux

Type:

array(ndim=1, dtype=vec3f)

density

density of medium

Type:

array(ndim=1, dtype=float32)

viscosity

viscosity of medium

Type:

array(ndim=1, dtype=float32)

integrator

integration mode (IntegratorType)

Type:

int

cone

type of friction cone (ConeType)

Type:

int

solver

solver algorithm (SolverType)

Type:

int

iterations

number of main solver iterations

Type:

int

ls_iterations

maximum number of CG/Newton linesearch iterations

Type:

int

ccd_iterations

number of iterations in convex collision detection

Type:

int

disableflags

bit flags for disabling standard features

Type:

int

enableflags

bit flags for enabling optional features

Type:

int

sdf_initpoints

number of starting points for gradient descent

Type:

int

sdf_iterations

max number of iterations for gradient descent

Type:

int

impratio_invsqrt

ratio of friction-to-normal contact impedance (stored as inverse square root)

Type:

array(ndim=1, dtype=float32)

ls_parallel

evaluate engine solver step sizes in parallel

Type:

bool

ls_parallel_min_step

minimum step size for solver linesearch

Type:

float

broadphase

broadphase type (BroadphaseType)

Type:

mujoco_warp._src.types.BroadphaseType

broadphase_filter

broadphase filter bitflag (BroadphaseFilter)

Type:

mujoco_warp._src.types.BroadphaseFilter

graph_conditional

flag to use cuda graph conditional

Type:

bool

run_collision_detection

if False, skips collision detection and allows user-populated contacts during the physics step (as opposed to DisableBit.CONTACT which explicitly zeros out the contacts at each step)

Type:

bool

contact_sensor_maxmatch

max number of contacts considered by contact sensor matching criteria contacts matched after this value is exceded will be ignored

Type:

int

class RenderContext

Context for rendering.

nrender

number of actively rendering cameras

Type:

int

cam_res

camera resolution for actively rendering cameras

Type:

array(ndim=1, dtype=vec2i)

cam_id_map

camera id map

Type:

array(ndim=1, dtype=int32)

use_textures

whether to use textures

Type:

bool

use_shadows

whether to use shadows

Type:

bool

use_precomputed_rays

whether to use precomputed rays

Type:

bool

bvh_ngeom

number of geometries in the BVH

Type:

int

enabled_geom_ids

enabled geometry ids

Type:

array(ndim=1, dtype=int32)

mesh_registry

mesh BVH id to warp mesh mapping

Type:

dict

mesh_bvh_id

mesh BVH ids

Type:

array(ndim=1, dtype=uint64)

mesh_bounds_size

mesh bounds size

Type:

array(ndim=1, dtype=vec3f)

mesh_texcoord

mesh texture coordinates

Type:

array(ndim=1, dtype=vec2f)

mesh_texcoord_offsets

mesh texture coordinate offsets

Type:

array(ndim=1, dtype=int32)

mesh_facetexcoord

mesh face texture coordinates

Type:

array(ndim=1, dtype=vec3i)

textures

textures

Type:

array(ndim=1, dtype=warp._src.texture.Texture2D)

textures_registry

texture registry

Type:

list[warp._src.texture.Texture2D]

hfield_registry

hfield BVH id to warp mesh mapping

Type:

dict

hfield_bvh_id

hfield BVH ids

Type:

array(ndim=1, dtype=uint64)

hfield_bounds_size

hfield bounds size

Type:

array(ndim=1, dtype=vec3f)

flex_mesh

flex mesh

Type:

warp._src.types.Mesh

flex_rgba

flex rgba

Type:

array(ndim=1, dtype=vec4f)

flex_bvh_id

flex BVH id

Type:

warp._src.types.uint64

flex_face_point

flex face points

Type:

array(ndim=1, dtype=vec3f)

flex_faceadr

flex face addresses

Type:

array(ndim=1, dtype=int32)

flex_nface

number of flex faces

Type:

int

flex_nwork

total flex work items for refit

Type:

int

flex_group_root

flex group roots

Type:

array(ndim=1, dtype=int32)

flex_elemdataadr

flex element data addresses

Type:

array(ndim=1, dtype=int32)

flex_shell

flex shell data

Type:

array(ndim=1, dtype=int32)

flex_shelldataadr

flex shell data addresses

Type:

array(ndim=1, dtype=int32)

flex_radius

flex radius

Type:

array(ndim=1, dtype=float32)

flex_workadr

flex work item addresses for refit

Type:

array(ndim=1, dtype=int32)

flex_worknum

flex work item counts for refit

Type:

array(ndim=1, dtype=int32)

flex_render_smooth

whether to render flex meshes smoothly

Type:

bool

bvh

scene BVH

Type:

warp._src.types.Bvh

bvh_id

scene BVH id

Type:

warp._src.types.uint64

lower

lower bounds

Type:

array(ndim=1, dtype=vec3f)

upper

upper bounds

Type:

array(ndim=1, dtype=vec3f)

group

groups

Type:

array(ndim=1, dtype=int32)

group_root

group roots

Type:

array(ndim=1, dtype=int32)

ray

rays

Type:

array(ndim=1, dtype=vec3f)

rgb_data

RGB data

Type:

array(ndim=1, dtype=uint32)

rgb_adr

RGB addresses

Type:

array(ndim=1, dtype=int32)

rgb_size

per-camera RGB buffer sizes

depth_data

depth data

Type:

array(ndim=1, dtype=float32)

depth_adr

depth addresses

Type:

array(ndim=1, dtype=int32)

depth_size

per-camera depth buffer sizes

render_rgb

per-camera RGB render flags

Type:

array(ndim=1, dtype=warp._src.types.bool)

render_depth

per-camera depth render flags

Type:

array(ndim=1, dtype=warp._src.types.bool)

znear

near plane distance

Type:

float

total_rays

total number of rays

Type:

int

class SolverType

Constraint solver algorithm.

CG

Conjugate gradient (primal)

NEWTON

Newton (primal)

class State

State component elements as integer bitflags.

Includes several convenient combinations of these flags.

TIME

time

QPOS

position

QVEL

velocity

ACT

actuator activation

WARMSTART

acceleration used for warmstart

CTRL

control

QFRC_APPLIED

applied generalized force

XFRC_APPLIED

applied Cartesian force/torque

EQ_ACTIVE

enable/disable constraints

MOCAP_POS

positions of mocap bodies

MOCAP_QUAT

orientations of mocap bodies

NSTATE

number of state elements

PHYSICS

QPOS | QVEL | ACT

FULLPHYSICS

TIME | PHYSICS | PLUGIN

USER

CTRL | QFRC_APPLIED | XFRC_APPLIED | EQ_ACTIVE | MOCAP_POS | MOCAP_QUAT | USERDATA

INTEGRATION

FULLPHYSICS | USER | WARMSTART

class Statistic

Model statistics (in qpos0).

meaninertia

mean diagonal inertia (per-world)

Type:

array(ndim=1, dtype=float32)

class TrnType

Type of actuator transmission.

JOINT

force on joint

JOINTINPARENT

force on joint, expressed in parent frame

SLIDERCRANK

force via slider-crank linkage

TENDON

force on tendon

BODY

adhesion force on body’s geoms

SITE

force on site