Autonomous River Clean Up

Abstract

Floating plastic waste in rivers poses severe environmental risks and demands continuous monitoring and removal. Existing manual or semi-autonomous cleaning approaches are slow, labor-intensive, and limited in detection accuracy under dynamic water conditions. To address this gap, this work proposes an AI-assisted autonomous river clean-up robot capable of detecting, navigating toward, and collecting floating debris in real time. The system uses a Raspberry Pi with a Coral Edge TPU for high-speed inference, an Arduino microcontroller for deterministic motor and conveyor control, and a conveyor-based mechanism for physical waste retrieval. A custom dataset was prepared using Roboflow and the model was trained on Google Colab to generate an optimized TensorFlow Lite model. Experimental results show high detection accuracy, stable movement, and efficient waste collection. The system demonstrates a practical, low-cost solution for automated floating waste management in inland water bodies.