| We consider an Access Point (AP) equipped with multiple antenna elements and an in-band full-duplex radio. In each time slot, it needs to poll clients for uplink and/or carry out downlink transmissions, and to allocate its antenna elements for data transmissions and interference cancellation. We outline a Markov Decision Process (MDP) and use it to find the optimum polling policy that maximizes the average number of data streams. We also outline two novel Degree of Freedom (DoF) based antenna allocation algorithms. The results show our policy is able to activate respectively up to 58% more data streams and within 83% of the theoretical upper bound of half-duplex and full-duplex polling based methods that require perfect information of queue occupancy and interference at clients. |