Struct ParticleFilter 

pub struct ParticleFilter<'a> {
    pub cfg: ParticleConfig,
    pub q_chol: Vec<f64>,
    pub r: Vec<f64>,
    pub f: Box<dyn Fn(&[f64]) -> Vec<f64> + 'a>,
    pub h: Box<dyn Fn(&[f64]) -> Vec<f64> + 'a>,
    pub x0: Vec<f64>,
    pub p0_chol: Vec<f64>,
}

Particle filter (SIR) for nonlinear non-Gaussian state-space models.

State model: x_t = f(x_{t-1}) + noise, noise ~ N(0, Q) Obs model: z_t = h(x_t) + v_t, v_t ~ N(0, R)

Fields

cfg: ParticleConfig

Filter configuration.

q_chol: Vec<f64>

Cholesky factor L_Q (dim_x × dim_x lower triangular) of Q = L_Q L_Q^T.

r: Vec<f64>

Observation noise covariance R (dim_z × dim_z, row-major).

f: Box<dyn Fn(&[f64]) -> Vec<f64> + 'a>

Deterministic part of state transition: x_t = f(x_{t-1}) + noise.

h: Box<dyn Fn(&[f64]) -> Vec<f64> + 'a>

Observation model: z_t = h(x_t).

x0: Vec<f64>

Initial state distribution mean (length dim_x).

p0_chol: Vec<f64>

Cholesky of initial covariance P_0 (lower triangular, dim_x × dim_x).

Implementations

impl<'a> ParticleFilter<'a>

pub fn run(&self, z: &[f64], rng: &mut LcgRng) -> SeqResult<ParticleResult>

Run the particle filter on observations z (T × dim_z, row-major flat).

Returns weighted state estimates, effective sample sizes, and resampling count.