Maximizing Packets Collection in Wireless Powered IoT Networks With Charge-or-Data Time Slots

This paper studies data collection in a wireless powered Internet of Things (IoT) network with a hybrid access point (HAP). A fundamental problem at the HAP is to determine the number of time slots over a given planning horizon that is used to charge and collect data from devices. To this end, we outline a mixed integer linear program (MILP) to determine (i) the mode (charge or data) of each slot, (ii) the HAP’s transmit power allocation, and (iii) transmitting devices in data slots. Further, we propose a receding horizon approach whereby the HAP solves the said MILP over a time window using channel estimates from a Gaussian mixture model (GMM). We also outline a data-driven approach. In its
offline
stage, an IoT network operator first solves the said MILP over an exhaustive collection of channel power gains. The MILP solution for each channel power gain realization is then stored in neural networks. In the
online
stage, the HAP accesses the trained neural network of each time slot to retrieve its mode. The results show that the rolling horizon and data-driven approach allow the HAP to receive up to 106% and 90% more packets as compared to competing approaches.