nabled-sensor

Sensor models and state estimation for the nabled Physical AI stack.

nabled-sensor provides linear Kalman filtering, extended Kalman filter scaffolding, pinhole camera projection, and strapdown IMU integration over ndarray types. Estimation routines compose nabled-linalg linear algebra and stay independent of nabled-control.

Install

[dependencies]
nabled-sensor = "0.0.11"

Key modules

1. kalman: linear Kalman predict/update steps.
2. ekf: EkModel trait and EKF helpers.
3. camera: PinholeIntrinsics and projection utilities.
4. imu: strapdown IMU propagation helpers.

Crate graph

• Depends on: nabled-core, nabled-linalg.
• Used by: nabled-sim, facade nabled (physical-ai).

Optional features

1. blas, lapack-provider: forwarded to nabled-linalg.
2. openblas-system, openblas-static, netlib-system, netlib-static, magma-system.

[dependencies]
nabled-sensor = { version = "0.0.11", features = ["openblas-system"] }

Example

use nabled_sensor::kalman::{predict, KalmanState};
use ndarray::arr2;

let state = KalmanState {
    mean: ndarray::arr1(&[0.0, 0.0]),
    covariance: arr2(&[[1.0, 0.0], [0.0, 1.0]]),
};
let f = arr2(&[[1.0, 0.1], [0.0, 1.0]]);
let q = arr2(&[[0.01, 0.0], [0.0, 0.01]]);
let predicted = predict(&state, &f.view(), &q.view())?;
let _ = predicted.mean;

Docs

1. API docs: https://docs.rs/nabled-sensor
2. Workspace repo: https://github.com/MontOpsInc/nabled
3. Facade feature: nabled with physical-ai