AODS: Adaptive Obstacle Detection System for Smart Wheelchairs
Lucy Li, Wenyu Chen, Keyu Chen · 2024
The rising global demand for smart mobility solutions underscores the critical need for advanced technologies in power wheelchairs to ensure the safety and independence of users with mobility challenges. This work introduces a novel segmentation-based pipeline for obstacle detection, specifically tailored for outdoor environments, to enhance navigational safety in smart wheelchairs. The segmentation model within the pipeline addresses prevalent issues in current technologies-such as variable lighting conditions, scale changes, and distance sensitivity-that significantly affect detection accuracy. The proposed model incorporates innovative components, including a Distance Augmentation Module (DAM), Scale Aware Module (SAM), and an Illumination Simulation Module (ISM), ensuring robust performance under diverse operational conditions. Validation on a specially curated dataset from urban settings in Chengdu, China, demonstrates a segmentation accuracy of 86.0%, out-performing existing solutions. The findings suggest substantial improvements in detecting and navigating around obstacles, thereby significantly enhancing the operational efficiency and safety of smart wheelchairs. This work not only contributes a sophisticated algorithmic approach and a tailored dataset to the field but also verifies the practical efficacy of the model through extensive experimental validation.