Autonomous Coastal Litter Collection via SLAM-Based Navigation and Deep Learning: System Design, Field Validation, and Operational Impact
Kar Ho Tso
Abstract
Coastal litter accumulation represents a significant environmental, economic, and public health challenge affecting thousands of beaches globally. Conventional manual cleaning is labour-intensive (5–10 hours per site) and inefficient (30 –50% recyclables re covery). This paper presents EZBeachbot, an autonomous robotic system integrating SLAM-based navigation with LiDAR-Radar sensor fusion, deep learning-based waste detection, and real-time ESG measurement. We establish theoretical foundations for navigation and perception in unstructured coastal environments and validate the system through 12 -day field deployment at Deep Water Bay Beach, Hong Kong. Key contributions include: (1) an Extended Kalman Filter -based SLAM-LiDAR-Radar fusion algorithm with <0.87% drift over 2 km traverses; (2) deep learni ng detector (Faster R -CNN) achieving 91.4% F1 -score on 102,400 -image training set; (3) comprehensive field validation demonstrating 50% time reduction, 60% labour reduction, 78% recyclables diversion, and 82% carbon emissions reduction. The system is production -ready for municipal deployment at single or multi - site scale. Results provide empirical evidence for technical feasibility and commercial viability of autonomous coastal remediation.
Cite this article
Kar Ho Tso (2026). Autonomous Coastal Litter Collection via SLAM-Based Navigation and Deep Learning: System Design, Field Validation, and Operational Impact. Australian Journal of Business and Social Science, 1(5), 69-76. https://doi.org/10.66397/260510.20261024