Robust Indoor Mapping on a Smart Walking Aid via Visual SLAM

Abstract

The underlying research project aims to enhance independent mobility for elderly users by enabling indoor navigation assistance in a smart walking aid. A robust mapping system is crucial for such a device to ensure safety, effectively avoid obstacles, and create an optimal navigation path for users. Current approaches often rely on expensive sensors (e.g. LIDAR) or suffer from drift in unstructured environments. This paper proposes an optimized Visual SLAM solution based on the RTAB-Map framework for indoor mapping with a walking aid-mounted RGB-D camera. The method leverages appearance-based loop closure detection and parameter optimization to reliably generate both 2D occupancy grids and 3D point clouds under challenging conditions, like rooms located on different levels and transitions between rooms that are difficult to detect due to their layout. Indoor experiments demonstrate that the optimized system successfully handles typical walking aid motions and indoor scene features, achieving accurate reconstruction of corridors and rooms. Mapping accuracy is validated against the floor plan, showing lower drift than the baseline Visual SLAM setup. The results indicate that RTAB-Map, with proper parameter tuning, provides a solid and cost-effective mapping solution for assistive walking aid platforms, facilitating safe navigation support for the elderly.