SAC-MS: Joint Slice Resource Allocation, User Association and UAV Trajectory Optimization with No-Fly Zone...

Area of Science:

• Network Engineering
• Wireless Communications
• Artificial Intelligence in Networks

Background:

• Space-air-ground integrated networks (SAGINs) face increasing user demands and resource limitations.
• No-fly zone (NFZ) constraints add complexity to network management and resource allocation.
• Existing solutions struggle to efficiently manage diverse services and connectivity in SAGINs.

Purpose of the Study:

• To propose a joint optimization approach for SAGINs addressing NFZ constraints.
• To maximize system utility by optimizing user association, UAV trajectory, and slice resource allocation.
• To develop an efficient algorithm for solving the complex, non-convex optimization problem.

Main Methods:

• Decomposition of the non-convex problem into user association, UAV trajectory, and slice resource allocation subproblems.
• Development of the iterative SAC-MS algorithm combining matching game theory, sequential convex approximation (SCA), and soft actor-critic (SAC) reinforcement learning.
• Simulation-based comparison against TD3-MS, DDPG-MS, DQN-MS, and hard slicing methods.

Main Results:

• The proposed SAC-MS algorithm demonstrates superior performance in enhancing system utility.
• SAC-MS achieved significant improvements: 10.53% over TD3-MS, 13.17% over DDPG-MS, 31.25% over DQN-MS, and 45.38% over hard slicing.
• The joint optimization approach effectively balances resource allocation and network performance under NFZ constraints.

Conclusions:

• The SAC-MS algorithm provides an effective solution for joint optimization in SAGINs with NFZ constraints.
• This approach significantly improves system utility by integrating advanced optimization and reinforcement learning techniques.
• The findings offer a pathway for more efficient and robust SAGIN resource management in complex operational environments.