The problem
After an earthquake or flood the network is down. Drones can act as flying edge servers, but their battery is limited: the path, the hovering time and the computing decisions must be planned together, and conditions change all the time.
Who needs it
Emergency services, telecom operators (rapid network recovery), drone service companies.
Idea of the approach
- Build a 2-D simulation: ground users with tasks, one or more UAVs with battery and CPU limits, a channel model.
- Formulate the joint trajectory and offloading decision as a Markov decision process.
- Train a DRL agent (DQN for discrete moves, then PPO/DDPG); compare with greedy and fixed-path baselines.
- Visualize the learned trajectories.
What you will build and deliver
- Simulator and trained agents, with an animation.
- Paper.
Skills needed
Python, reinforcement learning, mathematics (optimization basics).
Research contribution & where to publish
Contribution: energy-aware joint trajectory/offloading learning with task deadlines. Venues: IEEE VTC, IEEE ICC, IEEE Transactions on Vehicular Technology, IEEE IoT Journal.
Key references — start reading here
- Y. Zeng, R. Zhang, T. J. Lim. Wireless Communications with Unmanned Aerial Vehicles: Opportunities and Challenges. IEEE Communications Magazine, 2016.
- S. Jeong, O. Simeone, J. Kang. Mobile Edge Computing via a UAV-Mounted Cloudlet: Optimization of Bit Allocation and Path Planning. IEEE Transactions on Vehicular Technology, 2018.
- V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, et al.. Human-level control through deep reinforcement learning. Nature, 2015.
- H. Djigal, J. Xu, L. Liu, Y. Zhang. Machine and Deep Learning for Resource Allocation in Multi-Access Edge Computing: A Survey. IEEE Communications Surveys & Tutorials, 2022.
Interested?
Send the topic title and the courses you have taken.
Contact Dr. Hamza Djigal