DRL Based Channel Estimation to Improve Sensor Power Efficiency for IoT Applications

초록

It is expected that the Wireless-Powered Crowd Sensing (WPCS) is a promising approach for IoT applications. We considered the application scenario is Autonomous Valet Parking (AVP) systems and for higher power transfer efficiency, the ability to implement WPCS in low-power wide area networks and in vehicles on-board units (OBU) is crucial. It can be achieved by using the large-scale array antenna based Wireless Power Transfer (WPT). Therefore, channel estimation at a sensor node should be achieved with minimal power consumption. So, it is important that the sensor must operate with self-powering via WPT from the dedicated energy sources i.e., power beacon, small-cell access point, and ambient RF sources. Deep Reinforcement Learning (DRL) can be used for time-varying wireless channels by sending the received power measurements from the OBU to RSU at DRL as an input, which will enable the channel estimation for WPT at OBU efficiently.