test_filters module

Tests for a general KF

class test_filters.MSD_UKF(init_dict, params_dict)[source]

Bases: BaseUnscentedKalmanFilter

Mass spring damper system for testing EKF implementation

k

Spring constant of system

Type:: float

m

Mass of system

Type:: float

b

Damping coefficient of system

Type:: float

_abc_impl = <_abc_data object>

dyn_model(x, u, predict_dict=None)[source]

Full dynamics model

Parameters:

x (UKF belief state under unscented transformation) –
u (Input vector) –
predict_dict (Dict) – Additional update parameters, not used in current implementation

Returns:

x_new – predicted state, dimension 2 x 1

Return type:

np.ndarray

linearize_measurements(update_dict=None)[source]

Linearization of measurement model

Parameters:: update_dict (Dict) – Additional update parameters, not used in current implementation
Returns:: H – Jacobian of measurement model, dimension 2 x 1
Return type:: np.ndarray

measure_model(x, update_dict=None)[source]

Full measurement model

Parameters:

update_dict (Dict) – Additional update parameters, not used in current implementation
x (np.ndarray) – State for measurement model

Returns:

y – measurement, dimension 2 x 1

Return type:

np.ndarray

class test_filters.MsdEkf(init_dict, params_dict)[source]

Bases: BaseKalmanFilter

Mass spring damper system for testing EKF implementation

k

Spring constant of system

Type:: float

m

Mass of system

Type:: float

b

Damping coeeficient of system

Type:: float

_abc_impl = <_abc_data object>

get_B(predict_dict=None)[source]

Map from control space to state space . No control signals in this example

Parameters:: predict_dict (Dict) – Additional predict parameters, not used in current implementation
Returns:: B – 2 x 1 array of all zero elements
Return type:: np.ndarray

linearize_dynamics(predict_dict=None)[source]

Linearization of dynamics model

Parameters:: predict_dict (Dict) – Additional predict parameters, not used in current implementation
Returns:: A – Linear dynamics model for MSD
Return type:: np.ndarray

linearize_measurements(update_dict=None)[source]

Linearization of measurement model

Parameters:: update_dict (Dict) – Additional update parameters, not used in current implementation
Returns:: H – Jacobian of measurement model, dimension 2 x 1
Return type:: np.ndarray

test_filters.msd_ekf_params()[source]

Return dictionary of initial parameters for MSD KF implementation

Returns:: ekf_init_dict – Dictionary with KF state dimension, initial state and covariance
Return type:: Dict

test_filters.msd_exact_sol(times)[source]

Compute exact solution for mass-spring-damper position evolution

Parameters:: times (np.ndarray) – Vector of linearly spaced time values for MSD
Returns:: exact_sol – Vector of exact position solutions for MSD system
Return type:: np.ndarray

test_filters.msd_filter_sol(times, x_exact, init_dict, params_dict, q, r, filter_type)[source]

Run filter for all time steps

Parameters:

times (np.ndarray) – Vector containing time instances for exact state evolution
x_exact (np.ndarray) – Vector containing exact state positions
init_dict (Dict) – Dictionary of filter initialization parameters
params_dict (Dict) – Dictionary of MSD parameters
q (float) – Process noise covariance values on diagonal term
r (float) – NavData noise covariance values on diagonal term
filter_type (string) – Type of filter to test, currently ekf

Returns:

x_filter (np.ndarray) – Position estimated by filter, dimension T x 1
P_pre (np.ndarray) – State covariance after prediction step, dimension T x 2 x 2
P_post (np.ndarray) – State covariance after update step, dimension T x 2 x 2

test_filters.msd_params()[source]

Return dictionary of additional parameters for MSD KF implementation

Returns:: params – Dictionary with MSD mass, spring constant and damping coefficient
Return type:: Dict

test_filters.msd_times()[source]

Return time vector for MSD system evolution

Returns:: msd_times – Vector of linearly spaced time values from 0 to 20 seconds
Return type:: np.ndarray

test_filters.msd_ukf_params()[source]

Return dictionary of initial parameters for MSD UKF implementation

Returns:: ekf_init_dict – Dictionary with UKF state dimension, initial state and covariance
Return type:: Dict

test_filters.test_exact_sol(times, x_exact, init_dict, params_dict, filter_type)[source]

Compare solution from KF to exact solution of the system

Parameters:

times (np.ndarray) – Vector containing time instances for exact state evolution
x_exact (np.ndarray) – Vector containing exact state positions
init_dict (Dict) – Dictionary of filter initialization parameters
params_dict (Dict) – Dictionary of MSD parameters
filter_type (string) – Type of filter to test, currently ekf

test_filters.test_filter_cov_tests(times, x_exact, init_dict, params_dict, q, r, filter_type)[source]

Test that covariance is PSD, decreases after update and increases after prediction step

Parameters:

times (np.ndarray) – Vector containing time instances for exact state evolution
x_exact (np.ndarray) – Vector containing exact state positions
init_dict (Dict) – Dictionary of EKF initialization parameters
params_dict (Dict) – Dictionary of MSD parameters
q (float) – Process noise covariance values on diagonal term
r (float) – Measurement noise covariance values on diagonal term
filter_type (string) – Type of filter to test, currently ekf