In recent years, studies have shown that reconfigurable antennas can offer additional performance gains in wireless networks especially networks employing Multiple Input Multiple Output (MIMO) techniques.  Reconfigurable antennas are capable of dynamically re-shaping their radiation patterns in response to the needs of a wireless link or a network, and are gradually making their way into commercial wireless systems. The key to effectively use the reconfigurability offered by these antennas and integrate them in practical wireless systems, is to select an optimal radiation state (in terms of capacity, SNR, SINR, diversity etc.) among all the available states for a wireless transceiver in a given wireless environment. Drexel researchers have developed a real-time online learning method for selecting the state of a reconfigurable antenna. The online learning framework provides certain mathematical performance guarantee and is formulated as a multi-armed bandit learning problem. The selection technique is implemented for an IEEE 802.11x based WiFi protocol using 2x2 MIMO configuration where both the base stations and the client devices have two port highly directional metamaterial Reconfigurable Leaky Wave Antennas. The performance of the selection technique is quantified using a software defined radio testbed for both Line-Of-Sight and Non-Line-Of-Sight links in an indoor environment. This method allows low-resource high-performance state selection that is adaptive to changing conditions and where wireless environment is not known apriori.