This thesis proposes a novel approach for enhancing the teleoperation of robotic arms through shared control mechanisms, integrating advanced machine learning techniques and classical control methods. The primary objective is to develop a robust system capable of autonomously guiding the robotic arm towards predicted objects while maintaining operator oversight. The framework employs a Long Short-Term Memory (LSTM) classification model to predict the location and nature of objects within the robot's workspace. This predictive capability enables the system to anticipate the operator's intentions and adaptively plan trajectories. Furthermore, artificial potential fields are utilized to generate guidance commands that assist the operator in maneuvering the robotic arm towards the predicted objects.