Source-linked AI summary
Joint Antenna Position and Transmit Power Optimization for Pinching Antenna-Assisted ISAC Systems
Yunhui Qin, Yaru Fu, Haijun Zhang
TL;DR
The paper addresses joint antenna-position and transmit-power optimization for pinching-antenna-assisted ISAC under sensing and energy constraints. It proposes maximum entropy-based reinforcement learning, which balances exploration and exploitation by maximizing reward and policy entropy. MERL outperforms benchmark schemes in cumulative reward, normalized data rate, sensing SNR, and stability, including 20.3% and 44.4% data-rate gains over TD3 and DDPG.
Problem
The paper asks how to jointly optimize pinching-antenna positions and transmit power to maximize communication data rate while satisfying sensing requirements and system energy constraints.
Method
MERL solves the constrained nonconvex optimization problem by maximizing cumulative reward and policy entropy to balance exploration and exploitation.
Results
MERL achieves higher cumulative reward, normalized total data rate, normalized sensing SNR, and stability than benchmark schemes, with normalized total-data-rate improvements of 20.3% over TD3 and 44.4% over DDPG.
Takeaways & Limitations
The results support MERL as a robust strategy for joint antenna positioning and transmit-power allocation in the studied pinching-antenna-assisted ISAC system.
Abstract
from arXiv · showhide
This letter explores how pinching antennas, an advanced flexible-antenna system, can enhance the performance of integrated sensing and communication (ISAC) systems by leveraging their adaptability, cost-effectiveness, and ability to facilitate line-of-sight transmission. To achieve this, a joint antenna positioning and transmit power optimization problem is formulated to maximize the total communication data rate while meeting the target sensing requirements and the system energy constraint. To address the complex non-convex optimization problem, we propose a maximum entropy-based reinforcement learning (MERL) solution. By maximizing cumulative reward and policy entropy, this approach effectively balances exploration and exploitation to enhance robustness. Numerical results demonstrate that the proposed MERL algorithm surpasses other benchmark schemes in cumulative reward, total data rate, sensing signal-to-noise ratio, and stability.
I. INTRODUCTION
ISAC needs flexible, directional antennas that can adapt to changing environments while sharing communication and sensing resources. This work formulates joint pinching-antenna positioning and power optimization and proposes MERL to improve performance under these requirements.
- ISAC shares hardware, spectrum, and energy between radar sensing and wireless communication to improve operational efficiency and sustainability.
- Traditional fixed-position antennas struggle to meet the directional, flexible, and adaptive requirements of dynamic ISAC environments.
- Existing flexible antennas generally offer only wavelength-scale position variation, limiting their ability to counteract large-scale path loss.
- Pinching antennas provide flexible deployment with almost no additional cost as their number increases, while antenna positions and transmit power affect wireless-system performance.
- The work jointly optimizes pinching-antenna positions and power allocation to improve communication rate and sensing SNR.
- MERL maximizes cumulative reward and policy entropy to balance exploration and exploitation while solving the constrained optimization problem.
- 20.3% higher normalized total data rate than TD3 and 44.4% higher than DDPG are reported for MERL, alongside a 16.7% normalized sensing-SNR improvement over TD3.
II. SYSTEM MODEL AND PROBLEM FORMULATION
The modeled system uses pinching antennas on a waveguide to serve mobile users and sense targets through newly established LoS links. The optimization jointly selects antenna positions and user transmit powers to maximize total data rate subject to sensing, power, spacing, and energy constraints.
- The system contains a base station with N pinching antennas, M single-antenna mobile users, and K sensing targets connected through pinching-antenna LoS links.
- Pinching antennas are positioned on a common waveguide, whose transmitted signals are phase-shifted versions of one another.
- Users can cancel the known, offline-generated sensing signal before decoding the information signal.
- The communication model uses TDMA, with user transmit power determining each mobile user's data rate under additive white Gaussian noise.
- Sensing treats each transmitted communication symbol as a radar pulse and represents target performance through an SNR requirement Γ_k,t − Γ_th ≥ 0.
- The objective maximizes total data rate by jointly optimizing antenna positions Ψ and user powers p while satisfying sensing, individual-power, energy, and antenna-spacing constraints.
- The resulting problem is nonconvex and further complicated by user mobility, changing antenna positions, and dynamic channel conditions.
III. MAXIMUM ENTROPY-BASED REINFORCEMENT LEARNING SOLUTION
This section introduces MERL to address the complex, nonconvex optimization problem. The method is designed as a reinforcement-learning solution to reconstruct and solve the formulation.
- MERL reconstructs the optimization problem and addresses its complex, nonconvex nature through reinforcement learning.
A. The Problem Reconstruction
The problem is reconstructed as a sequential decision process for jointly adapting pinching-antenna positions, user locations, and power allocation while respecting sensing and energy requirements.
- The decision process is represented by (S, A, R, γ), where γ is the discount factor.
- The state includes pinching-antenna, user, and sensing-target locations plus the system’s total energy constraints.
- Actions represent changes in pinching-antenna locations, user locations, and users’ power allocation.
- The reward combines the original optimization objective with the target-sensing SNR requirement.
B. The MERL Solution
MERL optimizes a stochastic policy by maximizing cumulative reward together with policy entropy, using neural networks and replay-based training to support the decision process.
- MERL maximizes cumulative reward and policy entropy to balance exploration and exploitation.
- The entropy term introduces randomness into the optimal policy and can be adaptively optimized through the temperature parameter ρ.
- The policy network generates actions from states, while two Q-networks reduce overestimation during training.
- Experience tuples are stored in replay memory and sampled in mini-batches to optimize the neural networks.
- The Q-networks are updated using target values, and the policy network is updated separately.
IV. SIMULATION EXPERIMENTS AND ANALYSIS
Simulations compare MERL with DDPG, TD3, and random positioning/power allocation using reward, normalized communication and sensing performance, and learning-rate analyses. MERL achieves stronger late-training rewards, communication and sensing outcomes, and relative stability across learning rates.
- Benchmark schemes: The simulations compare MERL against DDPG, TD3, and a random scheme for antenna positions and transmit power.The random scheme samples both quantities within their predefined regions.
- Reward performance: MERL achieves higher cumulative rewards than all benchmarks in later training stages and shows better stability and robustness than DDPG.All algorithms use a learning rate of LR = 1 × 10^-5 in this comparison.
- Communication and sensing performance: MERL improves normalized total data rate by 20.3% over TD3 and by 44.4% over DDPG.These communication results are reported from the normalized performance comparison.
- Communication and sensing performance: MERL improves normalized sensing SNR by 16.7% over TD3 and far surpasses DDPG.The comparison also reports better sensing stability than the benchmark schemes.
- Learning-rate analysis: With LR = 1×10^-5, MERL yields the best later-stage cumulative rewards and remains relatively stable across varying learning rates.The simulations randomly distribute communication users and sensing targets within a 150 m × 150 m square area.
V. CONCLUSION
The conclusion presents joint optimization of pinching-antenna positions and user power allocation for ISAC, solved with MERL under sensing and energy constraints. MERL maximizes reward and policy entropy and outperforms benchmarks across reward, data rate, sensing SNR, and stability.
- Conclusion: The study jointly optimizes pinching-antenna positions and user power allocation to improve total data rate and sensing SNR.The optimization addresses target sensing requirements and the system energy constraint.
- Conclusion: MERL balances exploration and exploitation by maximizing cumulative reward and policy entropy for antenna positioning and transmit-power allocation.The conclusion characterizes this as producing a robust strategy.
- Conclusion: MERL outperforms the benchmarks in cumulative discounted reward, total data rate, sensing SNR, and stability.The paper identifies further study of NOMA-assisted ISAC networks and off-policy reinforcement learning as future work.