Machine learning-based approaches for handover decision of cellular-connected drones in future networks: A comprehensive review

Abstract

The integration of cellular connectivity in drones signifies a crucial leap forward, offering the potential to revolutionize multiple industries. This comprehensive review examines the latest developments in the field of machine learning based handover (HO) decision-making for connected drones in future networks. The paper reviews a spectrum of machine learning techniques and evaluates their effectiveness in drones’ HO, exploring avenues such as hybrid AI models that combine the strengths of different ML approaches. Notably, combining deep reinforcement learning with other techniques forms promising solutions. The review finds that deep reinforcement learning models, when integrated with other techniques such as dueling double deep Q-network, have shown promising results in realizing optimized HO decisions and improving overall reliability. Additionally, the paper addresses prevalent research challenges, including issues related to high mobility, the three-dimensional nature of drone flight, small-cell deployment, and integration into cellular networks, emphasizing the importance of innovative solutions to achieve a more efficient and seamless handover process. By considering these obstacles and offering a forward-looking perspective outlining potential research directions, the review contributes to guiding future advancements in drones’ HO decision-making, ultimately facilitating the realization of more efficient and reliable drone operations and unlocking the full potential of drone connectivity and mobility within future networks.

Keywords

• Unmanned aerial vehicle
• Cellular-connected drone
• Handover
• Artificial intelligence (AI)
• Machine learning (ML)
• Deep learning (DL)