Unit test cases

[mgrouch]

Plan:

Waves

5 Gersten waves * (exact no noise, noise, noise and bias)

5 Fenton waves * (exact no noise, noise, noise and bias)

5 Jonswap wave patterns * (exact no noise, noise, noise and bias)

===Cases

Wave spacial profiles x/z same scale

Frequency tracking (of heave by KalmanBasic)
Zero crossing, Aranovskiy, KalmANF

Heave tracking
Zero crossing, Aranovskiy, KalmANF
x
KalmanNumStable, KalmanBasic, KalmanZC (by heave ZC)?

turn off warm up for charts.

Wave Direction tracking (only on Jonswap waves)
Degrees, Direction

[mgrouch]

n=5

3n3 =waves=45

sample frequency 200Hz

csv per wave

Test cases

1. profile

2. frequency 3 * heave 3

3. wave dir. 2 but only for Jonswap (5* 3)

[mgrouch]

Parameters

all as inputs. Process noise, measurement noise.
Gains. Etc.

Estimate RMS of error after some warmup time

[mgrouch]

P, Q, R

Q process noise
R measurement noise
P error covariance (uncertainty)

P evolves (scales with dt by filters itself)

Q need to be scaled for sampling frequency

Q ~ (dt) ^ n. n depends on process dynamics for dynamic discrete time processes

Position noise ~ dt
Velocity noise ~ dt^3
Acceleration noise ~ dt^5

R doesn’t scale with dt

P0 is not tied to sampling rate.

Q needs to scale for sampling rate