| We consider a Hybrid Access Point (HAP) that is equipped with a Successive Interference Cancellation (SIC) radio, Radio Frequency (RF) energy harvesting devices. The HAP is responsible for charging and collecting data from these devices. A fundamental problem at the HAP is scheduling uplink transmissions. In particular, given a number of transmission schedules where devices are scheduled into one or more uplink time slots, the HAP needs to select the best transmission schedule that yields the highest average sum-rate. To this end, we outline a discrete optimization approach that allows the HAP to learn the best transmission schedule over time without using an channel state information. Our results show that the HAP is able to learn the best transmission schedule with an average throughput that is 50% higher than slotted Aloha. |