# hacky way; just so its easy to reuse code
measurement_model_radar = kf_independent_fusion.measurement_model_radar
measurement_model_ais = measurement_model_radar
tracks_fused, tracks_ais, tracks_radar = kf_independent_fusion.track(
start_time, measurements_radar, measurements_ais, fusion_rate=1)
# calculate CI
alpha = 0.9
ci_nees = scipy.stats.chi2.interval(alpha, 4)
ci_anees = np.array(scipy.stats.chi2.interval(
alpha, 4 * num_estimates)) / num_estimates
# calculate some metrics
# calculate NEES
nees = calc_metrics.calc_nees(tracks_fused, ground_truth)
# calculate ANEES
anees = calc_metrics.calc_anees(nees)
# calculate RMSE
rmse = calc_metrics.calc_rmse(tracks_fused, ground_truth)
# calculate percentage NEES within CI
percentage_nees_within_CI = calc_metrics.calc_percentage_nees_within_ci(
nees, ci_nees)
metrics_info = "Percentage NEES within CI: " + str(percentage_nees_within_CI) + "\n" + "RMSE: " + str(rmse) + "\n" + \
"ANEES: " + str(anees) + "\n" + "Alpha: " + str(alpha) + "\n" + "CI NEES: " + str(ci_nees) + "\n" + \
"CI ANEES: " + str(ci_anees)
print(metrics_info)
# plot
fig = plt.figure(figsize=(10, 6))
# # calculate NEES
# stats_individual["NEES"] = calc_metrics.calc_nees(tracks_fused_independent[-num_steps_metrics:],
# ground_truth[-num_steps_metrics:])
# # calculate ANEES
# stats_individual["ANEES"] = calc_metrics.calc_anees(stats_individual["NEES"])
# # calculate RMSE
# stats_individual["RMSE"] = calc_metrics.calc_rmse(tracks_fused_independent[-num_steps_metrics:],
# ground_truth[-num_steps_metrics:])
# stats_individual["seed"] = seed
# stats_individual["fusion_type"] = "independent"
# stats_overall.append(stats_individual)
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(
tracks_fused_dependent[-num_steps_metrics:],
ground_truth[-num_steps_metrics:])
# calculate ANEES
stats_individual["ANEES"] = calc_metrics.calc_anees(
stats_individual["NEES"])
# calculate RMSE
stats_individual["RMSE"] = calc_metrics.calc_rmse(
tracks_fused_dependent[-num_steps_metrics:],
ground_truth[-num_steps_metrics:])
stats_individual["seed"] = seed
stats_individual["fusion_type"] = "dependent"
stats_overall.append(stats_individual)
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(
tracks_fused_ais_as_measurement, _ = kf_ais_as_measurement.track()
# fix the length of the fusions to dependent (as it is a bit shorter)
num_tracks = len(tracks_fused_dependent)
# get the num_tracks last elements
tracks_fused_independent = tracks_fused_independent[-num_tracks:]
tracks_fused_ais_as_measurement = tracks_fused_ais_as_measurement[
-num_tracks:]
# remove the first element of ground_truth (because we only fuse the n-1 last with dependent fusion)
ground_truth = ground_truth[-num_tracks:]
# Calculate some metrics
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(
tracks_fused_independent[-num_steps_metrics:],
ground_truth[-num_steps_metrics:])
# calculate ANEES
stats_individual["ANEES"] = calc_metrics.calc_anees(
stats_individual["NEES"])
# calculate RMSE
stats_individual["RMSE"] = calc_metrics.calc_rmse(
tracks_fused_independent[-num_steps_metrics:],
ground_truth[-num_steps_metrics:])
stats_individual["seed"] = seed
stats_individual["fusion_type"] = "independent"
stats_overall.append(stats_individual)
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(
sigma_process=sigma_process,
sigma_meas_radar=sigma_meas_radar,
sigma_meas_ais=sigma_meas_ais)
tracks_fused_ais_as_measurement, _ = kf_ais_as_measurement.track()
# fix the length of the fusions to dependent (as it is a bit shorter)
num_tracks = len(tracks_fused_ais_as_measurement) - 1
# get the num_tracks last elements
tracks_fused_ais_as_measurement = tracks_fused_ais_as_measurement[
-num_tracks:]
# Calculate some metrics
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(
tracks_fused_ais_as_measurement, ground_truth)
# calculate ANEES
stats_individual["ANEES"] = calc_metrics.calc_anees(
stats_individual["NEES"])
# calculate RMSE
stats_individual["RMSE"] = calc_metrics.calc_rmse(
tracks_fused_ais_as_measurement, ground_truth)
stats_individual["seed"] = seed
stats_individual["fusion_type"] = "ais as measurement"
stats_overall.append(stats_individual)
# plot some results
alpha = 0.95
ci_nees = scipy.stats.chi2.interval(alpha, 4)
ci_anees = np.array(scipy.stats.chi2.interval(alpha,
4 * num_tracks)) / num_tracks
start_time,
prior,
sigma_process_radar=0.01,
sigma_process_ais=0.01,
sigma_meas_radar=3,
sigma_meas_ais=1)
tracks_fused_independent, _, _ = kf_independent_fusion.track()
# TODO fix structure
# TODO fix that the size of tracks fused under independence and dependence have different size than ais as
# measurement
# Calculate some metrics
stats_individual = {}
# calculate NEES
stats_individual["NEES"] = calc_metrics.calc_nees(tracks_fused,
ground_truth)
# calculate ANEES
stats_individual["ANEES"] = calc_metrics.calc_anees(
stats_individual["NEES"])
stats_individual["seed"] = seed
stats_overall.append(stats_individual)
# plot some results
num_tracks = len(tracks_fused)
alpha = 0.95
ci_nees = scipy.stats.chi2.interval(alpha, 4)
ci_anees = np.array(scipy.stats.chi2.interval(alpha,
4 * num_tracks)) / num_tracks
# plot ANEES and confidence interval
fig_ci_anees = plt.figure(figsize=(10, 6))