Joint Trajectory, RIS, and Computation Offloading Optimization via Decentralized Model-Based PPO in Urban Multi-UAV Mobile Edge Computing
A new model-based MARL system coordinates drone fleets and smart surfaces to overcome urban signal blockages.
A team of researchers has published a paper detailing a novel AI framework designed to solve a critical bottleneck in urban drone networks. The system tackles the joint optimization of multiple drone (UAV) flight paths, computation offloading schedules, and the configuration of smart signal-reflecting surfaces (Reconfigurable Intelligent Surfaces or RIS). This coordination is vital in dense cities where buildings constantly block direct line-of-sight signals, crippling drone-based services like mobile edge computing. The proposed solution is a decentralized, model-based Multi-Agent Reinforcement Learning (MARL) approach, where each drone acts as an intelligent agent.
Instead of relying on slow, purely trial-and-error model-free learning, each drone agent learns a local model of its environment dynamics. It then uses this model to perform short, branched rollouts to plan its actions, updating its policy via the Proximal Policy Optimization (PPO) algorithm. Crucially, drones only need to coordinate with a few neighboring drones, and a lightweight central controller aggregates their proposals to configure the RIS panels. This architecture balances decentralized efficiency with necessary coordination. Simulations show the framework achieves performance close to a theoretically optimal centralized system, significantly improving network throughput and energy efficiency for scalable urban drone operations.
- Uses a decentralized, model-based MARL framework with PPO for stable, sample-efficient learning in dynamic environments.
- Coordinates UAV trajectories, computation offloading, and RIS phase shifts to mitigate urban signal blockages.
- Achieves near-centralized performance in simulations, boosting throughput and energy efficiency for scalable networks.
Why It Matters
Enables reliable, large-scale drone networks for real-time urban applications like delivery, surveillance, and emergency response.