Skip to content

Controlling determinism

InertialSim allows users to fully control the level of determinism in sensor simulations. Each sensor class (Gyro, Accelerometer, and IMU) takes an optional rng argument. This argument is passed to the numpy.random.default_rng class with the following behavior.

If None, then fresh, unpredictable entropy will be pulled from the OS. If an int or array_like[ints] is passed, then all values must be non-negative and will be passed to SeedSequence to derive the initial BitGenerator state. One may also pass in a SeedSequence instance. Additionally, when passed a BitGenerator, it will be wrapped by Generator. If passed a Generator, it will be returned unaltered. When passed a legacy RandomState instance it will be coerced to a Generator.

Determinism options

We demonstrate the basic options with an accelerometer.

import numpy as np

from inertialsim.devices import example_imu
from inertialsim.geometry import Vector
from inertialsim.sensors import SensorModel
from inertialsim.sensors.accelerometer import Accelerometer

model = SensorModel()
model.data_interface.simulate_quantization = False
model.data_interface.simulate_sample_rate = False

zero_input = Vector.from_cartesian([[0, 0, 0], [0, 0, 0]], time=[0.0, 0.01])

# Seed is None means "fresh, unpredictable entropy".  None is the default but is
# specified explicitly here for illustration.
seed = None
accelerometer = Accelerometer(model, example_imu, rng=seed)
result = accelerometer.simulate(specific_force=zero_input)
print(f"Run 1 (seed = None): {np.squeeze(result.specific_force.data[0])}")

accelerometer = Accelerometer(model, example_imu, rng=seed)
result = accelerometer.simulate(specific_force=zero_input)
print(f"Run 2 (seed = None): {np.squeeze(result.specific_force.data[0])}")

# Seed is int means repeatable output (a different output per integer).
seed = 7
accelerometer = Accelerometer(model, example_imu, rng=seed)
result = accelerometer.simulate(specific_force=zero_input)
print(f"Run 1 (seed = int):  {np.squeeze(result.specific_force.data[0])}")

accelerometer = Accelerometer(model, example_imu, rng=seed)
result = accelerometer.simulate(specific_force=zero_input)
print(f"Run 2 (seed = int):  {np.squeeze(result.specific_force.data[0])}")

Run 1 (seed = None): [-0.02138932 -0.02035507  0.01191007]
Run 2 (seed = None): [-0.0526421   0.01983978 -0.02136171]
Run 1 (seed = int):  [-0.01568841  0.01027257  0.00906132]
Run 2 (seed = int):  [-0.01568841  0.01027257  0.00906132]

Simulation results

The state attribute of the Gyro and AccelerometerResult classes contains the entire simulation state. The IMU attribute returns both the internal gyro and accelerometer states of the IMU.

The rng field saves the underlying bit generator state numpy.random.PCG64.state when the sensor is initialized. This can be used to recreate a bit generator from a previously saved run.

Other types

Other InertialSim types support a similar rng input with the same behavior described above. In particular most geometry types support generation of random samples. For example, see Vector.from_random.