Apparatus and method for adaptive Markov chain processing over map matching of vehicle trip GPS data
A non-transitory computer readable storage medium has instructions executed by a processor to adaptively increase Markov chain kernel window width while processing global positioning system (gps) samples as a function of the ratio between geodesic distances of a current route to an actual geodesic distance between pairs of gps samples. The total cost of different kernel sequences is found based on an aggregated sum of shortest path runs through graph edges within different kernel sequences. Road weights are modified based on relative distance of prospective snap locations of road segments to the gps points, so that the shortest path runs with respect to all possible snap locations are selected. Selected kernel sequences are presented as a proposed mapped vehicle route.
1. A non-transitory computer readable storage medium with instructions executed by a processor to:
adaptively increase Markov chain kernel window width while processing global positioning system (gps) samples as a function of the ratio between geodesic distances of a current route to an actual geodesic distance between pairs of gps samples;
find the total cost of different kernel sequences based on an aggregated sum of shortest path runs through graph edges within different kernel sequences, wherein road weights are modified based on relative distance of prospective snap locations of road segments to the gps points, so that the shortest path runs with respect to all possible snap locations are selected; and
present selected kernel sequences as a proposed mapped vehicle route.
2. The non-transitory computer readable storage medium of claim 1 further comprising instructions executed by the processor to apply road constraints as filters on the different kernel sequences.
3. The non-transitory computer readable storage medium of claim 1 further comprising instructions executed by the processor to remove noisy data from the gps samples.
4. The non-transitory computer readable storage medium of claim 1 further comprising instructions executed by the processor to apply a Gaussian filter to remove noisy data from the gps samples.
5. The non-transitory computer readable storage medium of claim 1 further comprising instructions executed by the processor to filter folding vertex paths in the gps samples.