Beacon Positioning via Temporal Clustering of Crowd-Sourced Data

The patent segments the crowd-sourced observations into multiple clusters based on spatial proximity and temporal recency. By dividing the full dataset into smaller spatial clusters and further filtering by time, the system processes manageable subsets rather than the complete observation set, reducing computational complexity while maintaining position accuracy through focused analysis of relevant data groups.

The patent extracts and removes outdated observations from the clustering analysis by applying temporal filters. Only observations within a specified time window relative to the current time are included in the clustering process. This extraction of relevant recent data eliminates the need to process historical observations, significantly reducing computational burden while preserving position determination accuracy.

The patent performs preliminary filtering of observations by time before conducting clustering analysis. By pre-processing the data to remove outdated observations and retain only recent measurements, the system prepares a reduced dataset that requires less computational processing. This preliminary temporal filtering action maintains reliability by ensuring only current relevant observations are analyzed, while dramatically reducing the time required for clustering operations.

The patent implements dynamic temporal filtering where the time window for including observations is adjusted based on current time and beacon movement patterns. The system adaptively determines which historical observations remain relevant, allowing the analysis to focus on dynamically selected recent data rather than static historical datasets. This dynamic approach maintains comprehensive position coverage for moved beacons while optimizing processing efficiency.

The patent segments the analysis into multiple passes or iterations, where each pass handles a specific time window or spatial region. Instead of forming all possible clusters from the complete historical dataset in one computationally intensive operation, the system divides the problem into smaller segmentation tasks processed sequentially or in parallel, reducing the energy required for each individual clustering operation while maintaining complete position information coverage.

The patent applies partial clustering by forming clusters only for the most recent and relevant observations rather than processing the entire historical record. By performing partial clustering on selected subsets of data that are most likely to contain current position information, the system achieves sufficient position accuracy with reduced computational energy, avoiding the excessive action of clustering all possible historical observations.