DRL-Based Beam Split Alleviation for Movable Antenna-Enabled Near-Field Wideband Communications.

Area of Science:

• Wireless Communication Engineering
• Signal Processing
• Antenna Theory

Background:

• Near-field communication (NFC) is crucial for future 6G wireless systems.
• Wide bandwidth operation in NFC suffers from beam split effect, degrading beamforming gain due to analog phase shifters.
• Existing solutions involve true-time-delay hardware, which can be complex.

Purpose of the Study:

• To alleviate the beam split effect in wideband NFC systems using movable antenna technology.
• To jointly optimize analog phase shifts and antenna positions for maximum minimum beamforming gain.
• To address the non-convex optimization challenges in near-field wideband systems.

Main Methods:

• A movable antenna-enabled near-field wideband uplink system with analog beamforming was considered.
• A deep reinforcement learning (DRL) framework using the soft actor-critic algorithm was developed.
• The DRL framework handles continuous action spaces and non-smooth max-min objectives for joint optimization.

Main Results:

• The proposed approach effectively alleviates the beam split effect.
• The system achieves a higher minimum beamforming gain compared to conventional methods.
• The DRL framework successfully handles the complex, non-convex optimization problem.

Conclusions:

• Movable antenna technology offers a viable solution for beam split alleviation in wideband NFC.
• Deep reinforcement learning provides an effective optimization strategy for complex near-field communication systems.
• The proposed method enhances beamforming performance for future 6G wireless applications.