Plastic pollution in aquatic ecosystems has emerged as a critical environmental challenge while the conventional methods offer a labor intensive and inefficient solution. To address this issue, we propose an ESP32-based autonomous water-cleaning robot that integrates Internet of Things (IoT) connectivity, edge-deployed YOLO object detection, and a conveyor-based mechanical collection mechanism. This robot navigates over water surfaces, capture floating debris, and classify objects in real time. The system employs DroidCam as a live video source, processed using a Python-based pipeline with TensorFlow and a YOLO-trained deep learning model to classify floating objects as plastic or non-plastic. Upon classification, decision signals are transmitted to an ESP32 microcontroller, which governs the conveyor belt and servo mechanisms to either collect or release objects. The ESP32 also handles wireless connectivity through the Blynk IoT platform, enabling remote monitoring, manual override, and operational data logging. Experimental evaluation demonstrates reliable detection accuracy and efficient debris handling under constrained hardware resources. The proposed design represents a scalable cyberphysical IoT system that leverages edge-assisted AI, embedded actuation, and mobile IoT integration to address sustainable plastic waste management in aquatic environments.