Energy-Efficient and Fresh Data Collection in IoT Networks by Machine Learning
The Internet-of-Things (IoT) is rapidly changing our lives in almost every field, such as smart agriculture, environmental monitoring, intelligent manufacturing system, etc. How to improve the efficiency of data collection in IoT networks has attracted increasing attention. Clustering-based algorithms are the most common methods used to improve the efficiency of data collection. They group devices into distinct clusters, where each device belongs to one cluster only. All member devices sense their surrounding environment and transmit the results to the cluster heads (CHs). The CHs then send the received data to a control center via single-hop or multi-hops transmission. Using unmanned aerial vehicles (UAVs) to collect data in IoT networks is another effective method for improving the efficiency of data collection. This is because UAVs can be flexibly deployed to communicate with ground devices via reliable air-to-ground communication links. Given that energy-efficient data collection and freshness of the collected data are two important factors in IoT networks, this thesis is concerned with designing algorithms to improve the energy efficiency of data collection and guarantee the freshness of the collected data. Our first contribution is an improved soft-k-means (IS-k-means) clustering algorithm that balances the energy consumption of nodes in wireless sensor networks (WSNs). The techniques of “clustering by fast search and find of density peaks” (CFSFDP) and kernel density estimation (KDE) are used to improve the selection of the initial cluster centers of the soft k-means clustering algorithm. Then, we utilize the flexibility of the soft-k-means and reassign member nodes by considering their membership probabilities at the boundary of clusters to balance the number of nodes per cluster. Furthermore, we use multi-CHs to balance the energy consumption within clusters. Extensive simulation results show that, on average, the proposed algorithm can postpone the first node death, the half of nodes death, and the last node death when compared to various clustering algorithms from the literature. The second contribution tackles the problem of minimizing the total energy consumption of the UAV-IoT network. Specifically, we formulate and solve the optimization problem that jointly finds the UAV’s trajectory and selects CHs in the IoT network. The formulated problem is a constrained combinatorial optimization and we develop a novel deep reinforcement learning (DRL) with a sequential model strategy to solve it. The proposed method can effectively learn the policy represented by a sequence-to-sequence neural network for designing the UAV’s trajectory in an unsupervised manner. Extensive simulation results show that the proposed DRL method can find the UAV’s trajectory with much less energy consumption when compared to other baseline algorithms and achieves close-to-optimal performance. In addition, simulation results show that the model trained by our proposed DRL algorithm has an excellent generalization ability, i.e., it can be used for larger-size problems without the need to retrain the model. The third contribution is also concerned with minimizing the total energy consumption of the UAV-aided IoT networks. A novel DRL technique, namely the pointer network-A* (Ptr-A*), is proposed, which can efficiently learn the UAV trajectory policy for minimizing the energy consumption. The UAV’s start point and the ground network with a set of pre-determined clusters are fed to the Ptr-A*, and the Ptr-A* outputs a group of CHs and the visiting order of CHs, i.e., the UAV’s trajectory. The parameters of the Ptr-A* are trained on problem instances having small-scale clusters by using the actor-critic algorithm in an unsupervised manner. Simulation results show that the models trained based on 20- clusters and 40-clusters have a good generalization ability to solve the UAV’s trajectory planning problem with different numbers of clusters, without the need to retrain the models. Furthermore, the results show that our proposed DRL algorithm outperforms two baseline techniques. In the last contribution, the new concept, age-of-information (AoI), is used to quantify the freshness of collected data in IoT networks. An optimization problem is formulated to minimize the total AoI of the collected data by the UAV from the ground IoT network. Since the total AoI of the IoT network depends on the flight time of the UAV and the data collection time at hovering points, we jointly optimize the selection of the hovering points and the visiting order to these points. We exploit the state-of-the-art transformer and the weighted A* to design a machine learning algorithm to solve the formulated problem. The whole UAV-IoT system, including all ground clusters and potential hovering points of the UAV, is fed to the encoder network of the proposed algorithm, and the algorithm’s decoder network outputs the visiting order to ground clusters. Then, the weighted A* is used to find the hovering point for each cluster in the ground IoT network. Simulation results show that the model trained by the proposed algorithm has a good generalization ability to generate solutions for IoT networks with different numbers of ground clusters, without the need to retrain the model. Furthermore, results show that our proposed algorithm can find better UAV trajectories with the minimum total AoI when compared to other algorithms.
machine learning, IoT, UAV, data collection
Doctor of Philosophy (Ph.D.)
Electrical and Computer Engineering