torchfilter.filters._particle_filter

Private module; avoid importing from directly.

Module Contents

Classes

ParticleFilter

Generic differentiable particle filter.
class torchfilter.filters._particle_filter.ParticleFilter(*, dynamics_model: DynamicsModel, measurement_model: ParticleFilterMeasurementModel, num_particles: int = 100, resample: bool | None = None, soft_resample_alpha: float = 1.0, estimation_method: str = 'weighted_average')[source]

Bases: torchfilter.base.Filter

Inheritance diagram of torchfilter.filters._particle_filter.ParticleFilter

Generic differentiable particle filter.

dynamics_model

Forward model.

Type:

torchfilter.base.DynamicsModel

measurement_model

Observation model.

Type:

torchfilter.base.ParticleFilterMeasurementModel

num_particles

Number of particles to represent our belief distribution. Defaults to 100.

Type:

int

resample

If True, we resample particles & normalize weights at each timestep. If unset (None), we automatically turn resampling on in eval mode and off in train mode.

Type:

bool

soft_resample_alpha

Tunable constant for differentiable resampling, as described by Karkus et al. in “Particle Filter Networks with Application to Visual Localization”: https://arxiv.org/abs/1805.08975 Defaults to 1.0 (disabled).

Type:

float

estimation_method

Method of producing state estimates. Options include:

  • ‘weighted_average’: average of particles weighted by their weights.

  • ‘argmax’: state of highest weighted particle.

Type:

str

particle_states :torch.Tensor

Discrete particles representing our current belief distribution. Shape should be (N, M, state_dim).

Type:

torch.Tensor

particle_log_weights :torch.Tensor

Weights corresponding to each particle, stored as log-likelihoods. Shape should be (N, M).

Type:

torch.Tensor

initialize_beliefs(self, *, mean: types.StatesTorch, covariance: types.CovarianceTorch) → None[source]

Populates initial particles, which will be normally distributed.

Parameters:

  • mean (torch.Tensor) – Mean of belief. Shape should be (N, state_dim).

  • covariance (torch.Tensor) – Covariance of belief. Shape should be (N, state_dim, state_dim).

forward(self, *, observations: types.ObservationsTorch, controls: types.ControlsTorch) → types.StatesTorch[source]

Particle filter forward pass, single timestep.

Parameters:

  • observations (dict or torch.Tensor) – observation inputs. should be either a dict of tensors or tensor of shape (N, ...).

  • controls (dict or torch.Tensor) – control inputs. should be either a dict of tensors or tensor of shape (N, ...).

Returns:

torch.Tensor – Predicted state for each batch element. Shape should be (N, state_dim).