How to improve accuracy?

Attached is an example of deltas(difference between the observed and the truth). How can I improve by making these deltas or residuals smaller(approach to zero)

[bio lim]

You need to give us more information. What are you observing?

[AbelM Kusemererwa]

The attached file is the deltas for GPS measurement

Answer 1

[Star Strider]

All the values are positive, leading me to believe they are absolute distances from some reference. You likely cannot improve the accuracy of any measurement, but you can reduce the error by averaging the ‘raw’ GPS readings (not the derived distances) over a small window, then deriving the distance and calculating the error. That will likely reduce it. You can never eliminate it entirely.

[AbelM Kusemererwa]

They are root mean square error(rmse) of plan position. With your explanation can you please add figure to it to make it clearer?

[Star Strider]

I would use a moving-average filter with a short window on each of the (x,y,z) elements in your GPS data.

This code snippet illustrates the idea. You can use either filter (core MATLAB) or filtfilt (Signal Processing Toolbox) to do the moving-average filtering:

L = 100;                                    % Data Record Length
v = 1:L;
xyz = randn(L, 3);                          % Hypothetical GPS Data (Lx3)

N = 5;                                      % Moving Average Filter Length
filt_b = ones(1, N);
filt_a = N;

xyz_avg = filter(filt_b, filt_a, xyz);

xyz_dist = sqrt(sum(xyz.^2, 2));            % Sum Across Rows
xyz_avg_dist = sqrt(sum(xyz_avg.^2, 2));    % Sum Across Rows

figure(1)
plot(v, xyz_dist,  v, xyz_avg_dist)
grid
xlabel('Observation #')
ylabel('Distance From Reference')
legend('Raw Data', 'Moving Average Of Data')