Infrastructure-Free Indoor Navigation Via Quorum-Sensing and LT-Style Mapping

This paper presents QSLT-Map, a bearing-only hierarchical mapping and navigation framework that fuses a bacterial quorum-sensing (QS) layer with an LT-style landmark tree on minimalist robots. The method stores compact viewframes and promotes slowly varying bearings to a stable backbone while leaving volatile cues near the leaves, enabling aggressive pruning without breaking homing. QS provides a local scalar intensity derived from single-hop broadcasts (ESP-NOW) that adapts the effective capture threshold, coordinates leader-follower roles, and guides pruning toward regions that matter. A saturated differential-drive controller based on a secant direction from bearing differences, with a wall-following fallback, closes the loop without reliance on metric maps or infrastructure. We implement the full stack on SERB+ESP32 with fixed-size buffers and FreeRTOS tasking, keeping per-agent computation O(1) and memory within a few kilobytes for typical routes. Experiments in Player/Stage and on hardware show reduced time-to-goal and faster re-engagement after visual loss relative to an LT-only baseline, as well as high success under strict pruning and fewer collisions near doorways. The ranking of methods is consistent across testbeds, indicating that QS-aware capture and pruning transfer beyond instrumented arenas while remaining within the compute and communication budgets of small indoor platforms.