This research delves into the potential of Imitation Learning and Generative Adversarial Networks (GANs) for enhancing robotic manipulation. Historical data-driven models, such as Behavior Cloning (BC) and Hierarchical Behavior Cloning (HBC), showcased remarkable performance on expert human datasets. However, challenges including overfitting and the necessity for exhaustive policy testing were evident. HBC emerged asa promising solution, particularly for complex, hierarchical tasks, but raised concernsregarding scalability due to increased memory usage during training. Generative Adversarial Imitation Learning (GAIL) displayed steady progress, particularly for tasks like ”Lift”, highlighting the balance between its potential and challenges. Despite GAIL’s ability to replicate expert behavior, its extensive computational demands and prolonged training durations present hurdles. Emphasis is also placed on the invaluable role of human data, especially for intricate tasks, underscoring the importance of large humandatasets for complex operations. The research underscores the need for further exploration in areas such as Quality of Experience (QoE) in telecontrol, broader applicability of techniques in diverse scenarios, and the challenges of translating findings to real-worldapplications. As robotic manipulators become more integrated into human environments,the focus shifts towards ensuring safety, scalability, and adaptability. This research sets afoundation, but the path ahead offers ample opportunities for refinement and broadened applicability.