Movable Antenna-Enabled Integrated Sensing and Communication in Low-Altitude UAV Networks

Technical Summary: Movable Antenna-Enabled Integrated Sensing and Communication in Low-Altitude UAV Networks

Problem Statement
This paper addresses the optimization of a multi-Unmanned Aerial Vehicle (UAV) assisted Integrated Sensing and Communication (ISAC) system operating in low-altitude airspace. The system employs UAVs equipped with Movable Antenna (MA) arrays to serve ground users with dynamic mobility while simultaneously performing radar sensing tasks. The core challenge lies in the high-dimensional, non-convex nature of jointly optimizing UAV trajectories, user association, beamforming, antenna positioning, and power allocation. These variables are tightly coupled and operate within a dynamic environment characterized by unknown state transitions due to user mobility. Furthermore, traditional fixed-position antenna (FA) arrays are ill-suited for such dynamic scenarios, often leading to beam misalignment and suboptimal spatial utilization. The optimization is constrained by UAV flight dynamics (speed, collision avoidance, altitude limits), transmission power budgets, mechanical limits of antenna movement, and a minimum sensing Signal-to-Interference-plus-Noise Ratio (SINR) requirement.

Methodology
To tackle the complexity of the formulated optimization problem, the authors decompose it into two sequential stages:

1. User Association via HDBSCAN:
   The first stage addresses the air-ground association problem. Instead of relying on simple distance metrics which fluctuate rapidly with UAV movement, the authors employ the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm. Ground users are periodically clustered based on their spatial distribution. UAVs are then assigned to these clusters (using the Hungarian algorithm or nearest-neighbor assignment). This strategy ensures that users associated with a specific UAV share similar spatial characteristics, stabilizing the association and facilitating efficient MA array adjustments.

2. Joint Optimization via Soft Actor-Critic (SAC):
   The second stage utilizes a Deep Reinforcement Learning (DRL) approach, specifically the Soft Actor-Critic (SAC) algorithm, to solve the joint optimization of UAV trajectories, antenna positions, beamforming vectors, and power allocation.

   • MDP Formulation: The problem is modeled as a Markov Decision Process (MDP). The state space includes channel information, UAV and user locations, and current antenna deployments. The action space encompasses beamforming design, flight speed/direction, antenna displacement, and power allocation ratios.
   • Reward Function: The reward is defined as the total communication sum rate, penalized for violating sensing SINR thresholds, collision avoidance constraints, and speed limits.
   • Algorithm Choice: SAC is selected for its off-policy nature, sample efficiency, and ability to balance exploration and exploitation through entropy regularization, which is crucial for navigating the high-dimensional continuous action space and avoiding local optima.

Key Contributions
The paper outlines three primary contributions:

1. System Architecture: The integration of MA arrays into a low-altitude multi-UAV ISAC system. Unlike traditional fixed-antenna deployments, this architecture exploits the spatial degrees of freedom of MA arrays to dynamically adapt to dual communication and sensing requirements. The authors jointly optimize MA deployment, beamforming, user association, and UAV trajectories under strict sensing and flight constraints.
2. Two-Stage Optimization Framework: A novel solution combining HDBSCAN and SAC. HDBSCAN is used for dynamic user clustering to optimize air-ground association, ensuring spatial consistency for subsequent MA adjustments. SAC is then applied to solve the complex joint optimization of trajectory, antenna configuration, and resource allocation.
3. Experimental Validation: The proposed framework is experimentally validated through simulations. The results demonstrate that the MA-enabled system significantly outperforms traditional fixed-antenna systems in terms of total communication rate while reliably meeting sensing SINR constraints. The study also quantifies the specific performance gains contributed by the MA array and the clustering mechanism.

Results
Numerical simulations were conducted in a 500m x 500m area with 3 UAVs, 9 communication users, and 3 sensing targets. Key findings include:

• Performance Gain: The proposed scheme (MA + HDBSCAN + SAC) achieved the highest total communication rate compared to four baseline schemes, including those using Fixed Antennas (FA) or lacking the clustering mechanism.
• Role of Clustering: The HDBSCAN-based association improved early-stage exploration efficiency and overall performance by reducing the exploration pressure on the agent and stabilizing user-UAV links.
• MA vs. FA: Systems with MA arrays consistently outperformed those with FA arrays. However, the performance advantage of MA diminished slightly as the number of antennas increased, attributed to UAV load constraints and increased action dimensionality.
• Algorithm Comparison: The SAC algorithm demonstrated superior performance over DDPG, PPO, and CMA-ES. It achieved higher cumulative rewards, faster convergence (effective strategy after ~130 epochs), and greater training stability.
• Scalability and Sensitivity: The system maintained stability even as the number of users increased, though the per-user rate decreased due to resource sharing. Increasing sensing SINR thresholds naturally reduced communication performance as more resources were diverted to sensing.

Significance
The paper claims that its work fills a gap in existing research, which predominantly focuses on fixed-antenna arrays or single-UAV scenarios. By integrating MA arrays into a multi-UAV ISAC framework and proposing a specific two-stage optimization strategy, the study demonstrates that MA technology can significantly enhance system flexibility and performance in dynamic, low-altitude environments. The work validates that optimizing antenna positions in conjunction with UAV trajectories and beamforming allows for better utilization of spatial diversity, leading to improved communication rates without compromising sensing reliability. The proposed method offers a practical approach to handling the complex, coupled decision-making required in next-generation 6G ISAC networks.