Method for selecting state of a reconfigurable antenna in a communication system via machine learning

Overview

This technology is an algorithm for selecting the state of a reconfigurable antenna array installed at either the transmitter or receiver of a communication system. The learning algorithm can be deployed remotely based on the theory of multi-armed bandit, which aims to optimize the wireless link between transmitter and receiver over time by selecting the best antenna array state. The selection is based on Post-Processing Signal-to-Noise Ratio (PPSNR) as a reward metric and aims to maximize the long-term average reward. The learning algorithm is tested using wireless channel data collected in an indoor environment, employing highly directional metamaterial Reconfigurable Leaky Wave Antennas (RLWAs). The method shows improvements in terms of average PPSNR and regret compared to conventional heuristic policies.

Market Applications

Mobile devices, such as smartphones, tablets, laptops, and wearable devices
Wireless communication systems, including MIMO (Multiple Input Multiple Output) systems
Satellite communication, such as LEO, MEO, and GEO satellites
Military and defense, such as radar, jamming, and stealth systems
Wireless networks, such as cellular, Wi-Fi, Bluetooth, and IoT
Autonomous vehicles and drones

Key Advantages

Simple and efficient way to select the optimal state of a reconfigurable antenna in real-time
Can adapt to changing communication environments
Improves communication performance by selecting the antenna state that maximizes the PPSNR
Enables the implementation of the algorithm on low-cost and low-power devices
Adapts to changing wireless channel conditions, antenna orientation, and node mobility
Utilizes a multi-armed bandit framework for learning and decision-making
Provides a general framework that can be applied to various types of reconfigurable antennas and communication systems

Problems Solved

Reducing the complexity and overhead by avoiding the need for channel estimation and feedback
Addressing the challenges of unknown optimal states for different links, balancing exploration and exploitation, and reducing training overhead
Low computational overhead and minimal feedback requirements for practical implementation
Improving wireless link performance, leading to higher throughput and signal-to-noise ratio
Providing an adaptive, practical, and hands-off selection technique for reconfigurable antennas