how the accuracy of the track data to be much closer to the measure؟

[ghamsarimah]

I use my simulation data app to extended kalman filter and unscented kalman filter and particle filter

x_track: 43.638830523686714 x_measure: 43.638799999999996 y_track: 31.357527684009003 y_measure: 31.3575

x_track: 43.638689310588276 x_measure: 43.6363 y_track: 31.3575107432187 y_measure: 31.2329

x_track: 43.638689326309276 x_measure: 43.6557 y_track: 31.35751074942982 y_measure: 31.1227

x_track: 43.638689342029 x_measure: 43.67179999999999 y_track: 31.35751075564258 y_measure: 31.0153

x_track: 43.63868935775006 x_measure: 43.667899999999996 y_track: 31.357510761854254 y_measure: 30.902600000000003

x_track: 43.63868937347023 x_measure: 43.6262 y_track: 31.3575107680666 y_measure: 30.7871

x_track: 43.638689389189906 x_measure: 43.5593 y_track: 31.3575107742786 y_measure: 30.6927

I want their difference to reach 0.001. What should I do?

[sdhiscocks]

Hi @ghamsarimah. Do you have more details on the issue you are having, or what your trying to achieve?

If you are looking to reduce difference between track and measurement, one way is to reduce your measurement model noise, or increase transition model noise; such that Kalman gain is more weighted to measurement for example. But that may not be accurate representation of system/targets.

[ghamsarimah]

thank you for your response. I have target tracking software that uses radar data (GPS) to predict target tracks. The first image shows the ground truth, but the predicted track in the second image is significantly lower, resulting in a large discrepancy between the two.

the radar error = random.gauss(0, 1)

MeasurementModel--> (q_x,q_y) = (2,2) TransitionModel-->(q_x,q_y) = (0.1, 0.1)

I use particle filtering