On Balancing Exploration Vs. Exploitation in a Cognitive Engine for Multi-Antenna Systems

In this paper, we define the problem of balancing exploration vs. exploitation in a cognitive engine controlled multi-antenna communication system in terms of the classical multi-armed bandit framework. We then employ the e-greedy strategy and Gittins´ indices methods for addressing the problem in a system with no prior information. Results show that the Gittins´ indices assuming a normal reward process had the best overall performance compared to the Gittins´ indices with a Bernoulli reward process and the e-greedy strategy. The latter was found to be more consistent albeit inefficient for most of the cases except in the case of both a low number of trials and a low SNR in which it was found to have better performance than the other methods. Nevertheless, the Gittins´ indices method should be generally preferred as it is more consistent than the e-greedy strategy across different scenarios.