Maintaining real-time data freshness plays a critical role in ensuring the system correctness and optimizing the system performance in networked embedded systems (NESs). To quantitatively measure the freshness of the collected real-time data, the concept of Age of Information (AoI) has been extensively studied in recent years. Most existing studies on AoI optimization focus on the designs of effective update policies to optimize their freshness under different queueing models, assuming that the power supply of the individual nodes is continuous and sufficient. However, a growing trend in NES design is to power the embedded nodes through energy harvesting (EH) sources, which are mostly intermittent and often insufficient to process every real-time data update. This thesis explores how to optimize the freshness of real-time data for EH-based NESs with energy constraints.

In the first part, we explore how to schedule updates to optimize the AoI performance of real-time data in NESs powered by common EH source. We start from the AoI optimization problem in single-source EH-based systems and then extend the study to the multi-source case. Assuming that the EH-based source receives a sequence of real-time data updates, but it would discard some of these updates due to the energy constraint. The objective is to select a subset of the received real-time data updates at the source node and determine their transmission times so that the average AoI of the selected sub-sequence of real-time data updates is optimized, subject to the energy constraint imposed by the EH source.

In the second part, we focus on average AoI optimization in NESs with specific EH sources, radio frequency (RF), which provides key benefits in terms of being wireless, controllable, low cost, and small-form-factor implementation, and is thus paving the way for the emerging wireless powered communication networks (WPCNs). We formulate the average AoI optimization problem with the energy constraint at the sensor node(s). A 5-node WPCN testbed is developed to validate the functional correctness of the proposed solutions. Extensive simulation-based experiments are also conducted for performance evaluation under more comprehensive settings and the experimental results show that the proposed solutions can significantly outperform the state-of-the-art methods.

In the third part, based on a similar system given in the second item, we explore how to minimize worst-case AoI due to the worst-case optimization however is an important consideration in NESs. An optimal time allocation solution is designed to judiciously determine the working cycle and time allocation ratio for individual sensor nodes to minimize the worst-case AoI. We implement the proposed solution in the testbed to validate the functional correctness of the proposed solution. The results show that the proposed solution achieves the best worst-case AoI performance compared with the existing methods.

On the basis of these optimizations, the AoI performance of the networked embedded system can be improved.