Indoor Positioning Based on Active Radar Sensing and Passive Reflectors: Reflector Placement Optimization

TL;DR

This work addresses accurate, low-cost indoor localization for autonomous mobile robots by leveraging a single-channel FMCW radar and passive local radar reflectors (LRPs). It introduces two localization modes—fingerprinting for global estimates and MLAT for absolute positioning—supported by a novel multiobjective PSO (OMOPSO) that optimizes 2D LRP placements in large, arbitrary rooms. The optimization jointly minimizes fingerprinting ambiguity $f_1$ and the GDOP-based MLAT objective $f_2$, under feasibility constraints, and accommodates a variable number of LRPs via up/down mutations and physics-inspired projection models. Simulation results show that carefully optimized placements (e.g., 24 LRPs with two types) significantly improve AMCL tracking accuracy (down to $13.54\,\text{cm}$) compared with random placements using more LRPs, highlighting the method’s potential for scalable indoor positioning with affordable hardware.

Abstract

We extend our work on a novel indoor positioning system (IPS) for autonomous mobile robots (AMRs) based on radar sensing of local, passive radar reflectors. Through the combination of simple reflectors and a single-channel frequency modulated continuous wave (FMCW) radar, high positioning accuracy at low system cost can be achieved. Further, a multi-objective (MO) particle swarm optimization (PSO) algorithm is presented that optimizes the 2D placement of radar reflectors in complex room settings.

Figures (10)

• Figure 1: Our conceptual system. A radar (at the cone vertex) senses LRPs (dots, green dots are detected) above it and uses them for localization.
• Figure 2: Top-down view of the fingerprint calculation. The nearest $N$ (here $N=3$) detected LRPs and corresponding LRP types give the fingerprint for the AMR position.
• Figure 3: By removing the LRP of selected type where most LRPs are visible, we avoid coverage violations after downmutation. The color indicates the number of visible LRPs.
• Figure 4: Our physically inspired models to handle and fix boundary condition violations.
• Figure 5: Iterative boundary condition subroutine. LRPs are drawn towards regions with coverage violations and repulse each other. After a few iterations, a feasible solution is found. This also works if the initial solution is far away from being feasible.