Joint Energy Harvesting, Semantic Transmission Selection, Channel Allocation and Power Control for Resource-Constrained IoT Networks

Nguyen Huu Sang,Nguyen Duc Hai,Nguyen Duc Duy Anh,Nguyen Cong Luong,Shimin Gong,Dusit Niyato
DOI: https://doi.org/10.1109/vtc2024-spring62846.2024.10683154
2024-01-01
Abstract:In this work, we propose the use of energy harvesting and semantic communication for Internet of Things (IoT) systems. The system allows IoT devices to harvest energy from a base station and then uses the harvested energy for extracting and transmitting semantic information, i.e., scene graphs, to the base station. The proposed network thus copes with the energy and network resource constraints of the IoT devices. To maximize the total image data or scene graph transmitted to the base station, we formulate a problem that optimizes the energy harvesting duration, the selection of original image or portions of scene graphs, transmit power, and channel allocation to the IoT devices. Under the high dynamics and uncertainty of the context and size of the collected images as well as the wireless channels and computing resources, we propose two advanced deep reinforcement learning (DRL) algorithms, i.e., advantage Actor-Critic (A2C) and proximal policy optimization (PPO), to solve the problem. Simulation results are implemented on the real dataset clearly showing that the performance achieved by the proposed algorithms is much higher than that achieved by the baseline scheme. This implies that more original images or triplets are transmitted.
What problem does this paper attempt to address?