isaaclab_newton.actuators

Newton-native actuator drives.

DriveBam

BAM voltage-domain servo drive.

BAM Servo Drive

class isaaclab_newton.actuators.DriveBam

Bases: DriveBase

BAM voltage-domain servo drive.

Each step delays the position command, sags the supply, runs the firmware proportional loop to a PWM duty cycle, converts it to motor torque, and publishes the gearbox friction budget that MJWarp applies as a constraint. Only the position target is consumed.

The drive clamps its own effort: authoring a registered clamping schema next to the unregistered NewtonBamDriveAPI token hides the drive from Newton’s schema discovery.

Attributes:

PER_DOF_PARAMS	Per-DOF parameter arrays, shape (N,) each.
backlash_pos_indices	Play-hinge coordinate indices, shape (N,).
external_torque	External gearbox load of the previous MJWarp solve [N.m], shape (N,).
env_dof_stride	DOFs per environment, which share one supply and one command-delay stream.
startup_settings	BAM settings the start-up sampling ran with, or None if it has not run.
friction_budget	Velocity-independent friction budget of the last step [N.m], shape (N,).
effective_vin	Supply voltage after sag of the last step [V], shape (N,).
motor_torque	Motor torque of the last step, before friction [N.m], shape (N,).

Classes:

State

Double-buffered per-DOF state of the BAM drive.

Methods:

resolve_arguments(args)	Resolve shared settings and per-DOF coefficients from a USD actuator prim.
__init__(*[, has_backlash, stribeck, ...])	Initialize the drive from Newton's resolved arguments.

PER_DOF_PARAMS: ClassVar[tuple[str, ...]] = ('kp_fw', 'kp_scale', 'kd_scale', 'vin', 'sag_gain', 'friction_scale', 'kt', 'resistance', 'error_gain', 'max_pwm', 'max_current', 'friction_base', 'friction_stribeck', 'dtheta_stribeck', 'alpha', 'load_friction_motor', 'load_friction_external', 'load_friction_motor_stribeck', 'load_friction_external_stribeck', 'load_friction_motor_quad', 'load_friction_external_quad', 'max_effort')

Per-DOF parameter arrays, shape (N,) each.

class State

Bases: State

Double-buffered per-DOF state of the BAM drive.

Attributes:

prev_motor_torque	Unclamped motor torque of the previous step, used for supply sag [N.m], shape (N,).
prev_applied_torque	Clamped effort of the previous step, used for friction [N.m], shape (N,).
delay_ring	Past position commands [rad or m], shape (N, max(max_delay, 1)).
delay_lag	Current command lag [physics steps], shape (N,).
delay_fill	Number of valid entries in delay_ring, shape (N,).
delay_step_count	Steps since the last reset, shape (N,).
delay_phase	Offset of the lag-resampling period per environment, shape (N,).
delay_rng_seed	Episode seed per environment, shape (N,).
delay_update_period	Lag-resampling period [physics steps].
env_dof_stride	DOFs sharing one command-delay stream.
delay_seed	Base seed of the lag and phase draws.
reset_count	Number of resets applied, which decorrelates successive episodes.

Methods:

assign(other)	Copy another state into this one, keeping this state's array storage.
__init__([prev_motor_torque, ...])

prev_motor_torque: wp.array[float] | None = None

Unclamped motor torque of the previous step, used for supply sag [N.m], shape (N,).

prev_applied_torque: wp.array[float] | None = None

Clamped effort of the previous step, used for friction [N.m], shape (N,).

delay_ring: wp.array2d[float] | None = None

Past position commands [rad or m], shape (N, max(max_delay, 1)).

delay_lag: wp.array[wp.int32] | None = None

Current command lag [physics steps], shape (N,).

delay_fill: wp.array[wp.int32] | None = None

Number of valid entries in delay_ring, shape (N,).

delay_step_count: wp.array[wp.int32] | None = None

Steps since the last reset, shape (N,).

delay_phase: wp.array[wp.int32] | None = None

Offset of the lag-resampling period per environment, shape (N,).

delay_rng_seed: wp.array[wp.int32] | None = None

Episode seed per environment, shape (N,).

delay_update_period: int = 0

Lag-resampling period [physics steps].

env_dof_stride: int = 1

DOFs sharing one command-delay stream.

delay_seed: int = 0

Base seed of the lag and phase draws.

reset_count: int = 0

Number of resets applied, which decorrelates successive episodes.

assign(other: State) → None

Copy another state into this one, keeping this state’s array storage.

__init__(prev_motor_torque: wp.array[float] | None = None, prev_applied_torque: wp.array[float] | None = None, delay_ring: wp.array2d[float] | None = None, delay_lag: wp.array[wp.int32] | None = None, delay_fill: wp.array[wp.int32] | None = None, delay_step_count: wp.array[wp.int32] | None = None, delay_phase: wp.array[wp.int32] | None = None, delay_rng_seed: wp.array[wp.int32] | None = None, delay_update_period: int = 0, env_dof_stride: int = 1, delay_seed: int = 0, reset_count: int = 0) → None

classmethod resolve_arguments(args: dict[str, Any]) → dict[str, Any]

Resolve shared settings and per-DOF coefficients from a USD actuator prim.

Parameters:

args – Authored attribute values, keyed by snake-case name.

Returns:

Shared settings and per-DOF coefficients.

Raises:

ValueError – If names are unknown, coefficients are missing, or delays are invalid.

__init__(*, has_backlash: int = 0, stribeck: int = 0, load_dependent: int = 0, quadratic: int = 0, vin_min: float = -inf, min_delay: int = 0, max_delay: int = 0, delay_hold_prob: float = 0.0, delay_update_period: int = 0, delay_seed: int = 0, **per_dof: warp.array)

Initialize the drive from Newton’s resolved arguments.

Parameters:

• has_backlash – Whether the encoder reads through a passive play hinge.

• stribeck – Whether the Stribeck friction terms are active.

• load_dependent – Whether gearbox friction grows with the transmitted torque.

• quadratic – Whether the quadratic load-coupling term is active.

• vin_min – Lower bound on the sagged supply voltage [V]. The default -inf keeps only the 0 V floor.

• min_delay – Minimum command delay [physics steps].

• max_delay – Maximum command delay [physics steps]. 0 disables the delay.

• delay_hold_prob – Probability of keeping the current lag instead of resampling it.

• delay_update_period – Physics steps between lag resamples. 0 resamples every step.

• delay_seed – Base seed of the lag and phase draws.

• per_dof – One (N,) array per entry of PER_DOF_PARAMS.

backlash_pos_indices: wp.array[wp.uint32] | None

Play-hinge coordinate indices, shape (N,). Bound before stepping for backlash drives.

external_torque: wp.array[float] | None

External gearbox load of the previous MJWarp solve [N.m], shape (N,). Bound by the MJWarp bridge.

env_dof_stride: int

DOFs per environment, which share one supply and one command-delay stream.

startup_settings: tuple | None

BAM settings the start-up sampling ran with, or None if it has not run.

friction_budget: wp.array[float] | None

Velocity-independent friction budget of the last step [N.m], shape (N,).

effective_vin: wp.array[float] | None

Supply voltage after sag of the last step [V], shape (N,).

motor_torque: wp.array[float] | None

Motor torque of the last step, before friction [N.m], shape (N,).