Method for selecting state of a reconfigurable antenna in a communication system via machine learning
A method for selecting the state of a reconfigurable antenna installed at either the receiver or transmitter of a communication system is provided. The proposed method uses online learning algorithm based on the theory of multi-armed bandit to perform antenna state selection. The selection technique utilizes the Post-Processing Signal-to-Noise Ratio (PPSNR) as a reward metric and maximizes the long-term average reward over time. The performance of the learning based selection technique is empirically evaluated using wireless channel data. The data is collected in an indoor environment using a 2×2 MIMO OFDM system employing highly directional metamaterial Reconfigurable Leaky Wave Antennas. the learning based selection technique shows performance improvements in terms of average PPSNR and regret over conventional heuristic policies.
1. A method of selecting an antenna array state for a multi-element reconfigurable transmitter and/or receiver antenna, comprising the steps of:
a processor executing a learning algorithm that optimizes unknown signal to noise ratios of possible signal paths between at least one reconfigurable transmitter antenna and at least one reconfigurable receiver antenna over time over different antenna array states of said at least one reconfigurable transmitter antenna and said at least one reconfigurable receiver antenna; and
the processor setting the antenna array configuration for the transmitter and/or a receiver antenna based at least in part on the antenna array states determined by said learning algorithm to lead to optimized signal to noise ratios between said at least one reconfigurable transmitter antenna and said at least one reconfigurable receiver antenna over time,
wherein executing the learning algorithm includes formulating selection of an antenna array state as a multi-armed bandit problem for learning the unknown signal to noise ratios of said possible signal paths so as to maximize signal to noise ratio between said at least one reconfigurable transmitter antenna and said at least one reconfigurable receiver antenna over time.
2. The method of claim 1 , wherein executing the learning algorithm includes implementing an arm selection policy and providing a performance measure, referred to as regret which is defined as the difference in an expected reward gained by always selecting a path with an optimal signal to noise ratio and a path that leads to an optimized signal to noise ratio over time, said learning algorithm processing received data to select the antenna array states that minimizes the regret over time.
3. The method of claim 2 , wherein regret after n selections is given by:
μ
*
n
-
μ
i
∑
i
=
1
N
E
[
T
i
(
n
)
]
where,
μ*=max 1≦i≦N μ i .
μ* is the average reward for the optimal arm, μ i is the average reward for arm i, n is number of total trials, E[●] is the expectation operator and T i is the number of times arm i has been sampled.