Improving Indoor Localization Accuracy by Using an Efficient Implicit Neural Map Representation
Year: 2025
Type: preprint
Abstract: A review of ENM-MCL, a global localization method that replaces the occupancy grid used by classical Monte Carlo Localization with an Efficient Neural Map: a dense feature grid paired with a shallow MLP that predicts both a non-projective signed distance and a direction-aware projective signed distance for any queried position and ray direction. The summary reconstructs the method's formulation, reports its measured accuracy and runtime against four baselines across five indoor sequences, and discusses why continuous direction-aware distance fields disambiguate corridor-like environments that occupancy grids cannot. Figures are the author's own: the model diagram is redrawn, and the results figure is plotted from the values reported in the reviewed work.
Source: arXiv (Cornell University)
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4416413113
Cites:
Cited by:
Related to: