Reliability-Guided Depth Fusion for Glare-Resilient Navigation Costmaps
2026-06-02 • Robotics
Robotics
AI summaryⓘ
The authors address the problem of shiny surfaces like floors or glass causing errors in depth sensors used by robots, which leads to fake obstacles on their navigation maps. They created a lightweight neural network called DRM-Net that predicts how trustworthy each depth measurement is when glare is present. Their system uses these predictions to carefully update the robot's map, avoiding mistakes from bad sensor data. Tests on a real robot showed this method reduces false obstacles and keeps the robot's navigation accurate and fast, especially in tricky lighting and reflective environments. Overall, the authors suggest treating glare as an issue of measurement trust rather than trying to fix missing depth data.
specular glareactive stereo RGB-Ddepth reliabilityoccupancy gridDRM-Netweighted fusiondepth measurement errorindoor navigationRealSense D435Jetson Orin Nano
Authors
Shang-En Tsai
Abstract
Specular glare on reflective floors, glass boundaries, and glossy indoor surfaces frequently corrupts active-stereo RGB-D depth measurements, producing holes and spikes that accumulate as persistent phantom obstacles in occupancy-grid costmaps. This paper presents a glare-resilient costmap construction method based on explicit depth-reliability modeling. A lightweight Depth Reliability Map network (DRM-Net) predicts per-pixel measurement trustworthiness under specular interference, and a reliability-guided weighted-and-gated fusion (RGF) mechanism modulates occupancy updates before corrupted measurements are accumulated into the map. To support robust training and evaluation, the method uses pose-aligned multi-view reference-depth construction to reduce circular-supervision bias and is evaluated through fusion-variant ablations, parameter-sensitivity analysis, cross-condition tests, paired navigation comparisons, reliability-map metrics, and embedded runtime profiling. Experiments on a real mobile robotic platform equipped with an Intel RealSense D435 and a Jetson Orin Nano show that the proposed method reduces false obstacle insertion, improves free-space preservation, and maintains real-time throughput under reflective-floor, glass-wall, and natural-light glare conditions. These results support treating glare as a measurement-reliability problem rather than as a dense depth-completion problem for safety-critical indoor navigation.