Efficient and sustainable weed control remains a significant challenge in modern agriculture. This paper presents the development of an Autonomous Weeding Robot (AWR) that detects and removes weeds using a vision-guided system integrated with sensor fusion for precise field navigation. The robot employs a pre-trained InceptionV3 model for feature extraction and a Random Forest algorithm for accurate weed classification. To ensure robust localization and control, sensor data from GPS, IMU, and wheel encoders are combined using an Extended Kalman Filter (EKF) within the Robot Operating System (ROS) framework. The system was evaluated through both simulation in Gazebo and physical trials across crops like cabbage, wheat, and sunflower. The AWR achieved approximately 85% classification accuracy in real-field conditions and demonstrated reliable autonomous navigation and weed removal. The results indicate that the proposed robot offers a practical and eco-friendly alternative to manual and chemical weed management, supporting the shift toward smarter and more sustainable agricultural practices.