Indoor localisation data cleaning

The spatiotemporal data of different people was recorded across different buildings (single and multi-flooried) for a few days. UWB technology was used for the localisation and similar to how it is with gps data, this data is noisy i.e. points getting located outside the facility although no outside the facility localisation was performed, Irregular spikes in the trajectories which makes trajectories cross walls (visually...as blueprint in the form of coordinates of each room corners is available) & bias in few places. Further, data collection was done with fixed in one and variable (in theory motion driven) sampling rate in other buildings. Consequently, jumps between consecutive points are longer in the latter case.

I seek help with data cleaning (and if possible interpolation: not a mandatory requirement) of the above ~ someone experienced in indoor localization & gps data cleaning / mining would be a good match.