Enhanced Robotic Arm Localization Using UKF Based on Joint Angles and Camera Observations

While vision-based methods for self-localization are well-established in mobile robotics, the use of machine vision for end-effector positioning in robotic arms remains comparatively less explored. This work presents a simulation framework for efficient multi-robotic arm collaboration, enabling precise end-effector monitoring across multiple arms. Traditional spline-based trajectory planning is employed to ensure smoother and more efficient control. The framework integrates a YOLO model for fast target recognition and an Unscented Kalman Filter (UKF) to enhance robustness against noise impact, ensuring reliable performance in networked environments. Developed in Webots with UR5e robotic arms, this framework demonstrates high portability across platforms with minimal code modification and achieves millimeter-level accuracy, providing a cost-effective solution for multi-robot perception and collaborative robotic arm research.