This paper presents the development of a proof-of-concept navigation pipeline for sidewalk robotics, integrating stereo RGB-D sensing with LiDAR point clouds to support robust mapping and dynamic obstacle avoidance. A visual SLAM-based approach is employed, with RTAB-Map selected for its tight integration with the ROS move_base navigation framework. The proposed system enables interoperability with 3D object detection algorithms through a modular integration strategy, allowing perception outputs to be directly incorporated into navigation decisions. In addition, a vision-based sidewalk segmentation method is introduced to constrain traversal paths and improve operational safety in urban environments.

System validation was primarily conducted in a ROS-based simulation environment due to hardware limitations associated with the OAK-D camera and project time constraints, with limited real-world testing performed. Results demonstrate successful integration of perception, localization, mapping, and navigation modules within a unified framework. The proposed architecture is modular and extensible, providing a foundation for future development and real-world deployment. This work establishes a scalable and cost-effective framework for stereovision-based autonomous sidewalk robots, contributing toward the advancement of integrated perception and navigation systems for urban robotic applications.