HomeNews        <>

Autonomous drones revolutionize data collection in challenging environments

24, Nov, 2023

 A groundbreaking initiative, ADANET seeks to revolutionize data collection in challenging environments, such as remote or rural areas, through the development of an Autonomous Drones-Assisted IoT Network. This cutting-edge project deploys low-power IoT nodes, harnessing renewable energy sources, and employs autonomous drones for efficient data collection. Technology like this can be especially useful in precision agriculture or disaster-stricken areas. To achieve success, an interdisciplinary team led by seasoned CISTER researchers in collaboration with industry leaders Tekever and AirMind adresses challenges that include optimizing online flight resource allocation in dynamic conditions, guarding against cyber-physical attacks through game-theoretic models, and implementing cooperative flight resource allocation while addressing privacy concerns.

Recent advancements in energy harvesting have empowered Internet-of-Things (IoT) networks to incorporate numerous nodes powered by renewable energy sources. Equipped with solar panels, wind generators, or wireless power receivers, IoT nodes opportunistically charge their batteries. Large-scale data collection in rural and remote areas, where traditional communication infrastructure is lacking, presents challenges. To address this, a project proposes the use of autonomous drones for data collection, especially in scenarios like disaster-stricken areas or smart farming applications.

The ADANET project aims to design a reliable and secure drones-assisted IoT network utilizing a swarm of autonomous drones. Research focuses on developing deep reinforcement learning-based flight resource allocation, an adversarial deep reinforcement learning framework for enhanced security, and an innovative cooperative flight resource allocation scheme using federated learning. It also involves constructing a testbed for operational validation through real-world experiments.

The anticipated outcome is a robust and secure IoT network where a swarm of autonomous drones optimally controls cruise and communication schedules, minimizing data packet loss. The project integrates deep reinforcement learning, federated learning, wireless communication security, and optimization techniques. With international collaboration from academic and industry partners, it has potential to significantly advance drones-assisted wireless systems in Portugal, impacting areas such as 5G, smart farming, package delivery, and emergency medicine.

Led by the Principal Investigator (PI) Dr. Kai Li (CISTER researcher), in collaboration with experts in various domains, the project aims to conduct high-quality research supported by an international academic-industry consortium, including CISTER Research Centre, Laboratory of Artificial Intelligence and Decision Support (LIAAD), National University of Singapore (NUS), University of Houston (UH), Technical University of Berlin (TUB), Tekever Autonomous Systems, and AirMind LLC.

The ADANET initiative tackles three main challenges in developing an Autonomous Drones-Assisted IoT Network:

  • Online Flight Resource Allocation Optimization: To address dynamic changes in channel conditions and varying energy levels, the project is exploring reinforcement learning, specifically deep Q-learning, to optimize cruise control and prevent data loss.
  • Cyber-Physical Attacks: Recognizing the vulnerability of drones to cyber-physical attacks due to dependence on network state information and employing a game-theoretic model to formulate strategies against adversaries and the drone's reactions to real-time faulty data injection.
  • Cooperative Flight Resource Allocation: Privacy concerns and limited backhaul resources pose challenges in transmitting all flight resource allocation information. The approach involves implementing backhaul-aware federated learning for cooperative cruise control, ensuring minimal latency while maintaining privacy.

The project consists of five key tasks, including online flight resource allocation optimization, practical application of adversarial deep reinforcement learning, development of a cooperative flight resource allocation scheme, construction of an experimental testbed for validation, and overall project management and coordination. These efforts aim to overcome these challenges and advance the development of a secure and efficient Autonomous Drones-Assisted IoT Network.

Ultimately, through this initiative, CISTER looks to achieve significant progress in drone-assisted wireless systems, with an impact on key sectors such as 5G, smart agriculture, parcel delivery and emergency medicine.

Related Projects:

Autonomous Drones Assisted Internet of Things Networks