Questions: Kalman Filter and State Estimation

2 questions to test your understanding

Score: 0 / 2
Question 1 Multiple Choice

A scalar Kalman filter estimates the current robot position x given GPS measurements z. The motion model predicts x_pred = x_prev + v·dt (constant velocity). The measurement z = x + noise_measurement has variance σ_z². The filter maintains state estimate x_est and error covariance P. After prediction, P increases (uncertainty grows due to model imperfection); after measurement update, P decreases. Why does the filter trade these two sources?

AThe prediction is always correct, and the measurement is always wrong; the filter ignores measurements and trusts predictions
BThe measurement is always correct, and the prediction is always wrong; the filter ignores predictions and trusts measurements
CBoth prediction and measurement have errors. The filter computes the Kalman gain K = P/(P + σ_z²), which weights the measurement update based on relative uncertainties: if measurement noise is large, K is small and the update is small; if prediction uncertainty is large, K is large and the update is large
DThe filter randomly chooses between prediction and measurement
Question 2 True / False

A particle filter is an alternative to the Kalman filter for state estimation. Unlike the Kalman filter, which assumes Gaussian distributions, the particle filter represents the belief as a set of weighted samples (particles). When would a particle filter be preferable to a Kalman filter?

TTrue
FFalse