bam.simulate

Classes

Simulator

Single-axis pendulum simulator used during identification.

Module Contents

class bam.simulate.Simulator(model: bam.model.Model)

Single-axis pendulum simulator used during identification.

Implements the BAM simulation loop: firmware control law → motor torque → friction budget → stopping-torque clipping → Euler integration. Used by bam.fit to roll out a model against recorded logs and compute the identification loss.

Parameters:

model – BAM friction model to simulate.

screen = None

model

reset(q: float = 0.0, dq: float = 0.0)

Resets the simulation to a given state

step(control: None | float, torque_enable: bool, dt: float)

Advance the simulation by one timestep.

Parameters:

• control – Control signal sent to the actuator (voltage, current, or torque depending on the actuator type). None is treated as zero.

• torque_enable – Whether the actuator is powered. When False the motor torque is zero and only gravity and friction act.

• dt – Timestep [s].

rollout_log(log: dict, reset_period: float = None, simulate_control: bool = False)

Roll out the model against a recorded log and return predicted trajectories.

Parameters:

• log – Processed log dict as returned by bam.logs.Logs.make_batch() or loaded directly from a processed JSON file.

• reset_period – If set, re-synchronize the simulator state to the log at this interval [s]. Useful when error accumulation destabilizes long rollouts.

• simulate_control – If True, recompute the control signal from the simulated state using the firmware control law. If False, use the control values recorded in the log.

Returns:

Tuple (positions, velocities, controls) — lists of values at each timestep.