Original example shows that using fused measurements can result in lower error than single measurement (0.391 vs 0.523). However, the code examples on master branch output error of 1.647 for fused measurements instead. It seems that the reason is that here FilterPy's saver is used, which has no ground truth data and instead calculates the error regarding measure.
P.S. It also may be better to correct the example with a single measurement either by preserving wheel sensor measurements or mocking the fused sensor by calling the randn() one more time each iteration of the loop.
Original example shows that using fused measurements can result in lower error than single measurement (0.391 vs 0.523). However, the code examples on master branch output error of 1.647 for fused measurements instead. It seems that the reason is that here FilterPy's saver is used, which has no ground truth data and instead calculates the error regarding measure.
P.S. It also may be better to correct the example with a single measurement either by preserving wheel sensor measurements or mocking the fused sensor by calling the
randn()one more time each iteration of the loop.