Scalable Low-Power Wi-Fi for the Internet of Things

Date: 26 April 2019

Venue: Stadscampus, Kapel van de Grauwzusters - Lange Sint-Annastraat 7 - 2000 Antwerpen (route: UAntwerpen, Stadscampus)

Time: 4:00 PM

Organization / co-organization: Department of Mathematics and Computer Science

PhD candidate: Le Tian

Principal investigator: Jeroen Famaey & Steven Latré

Short description: PhD defence Le TIAN - Faculty of Science, Department of Mathematics and Computer Science


The Internet of Things (IoT) introduces a novel dimension to the world of information and communication technology where connectivity is available anytime, anywhere for anything. To make this into reality, a large number of battery powered smart things need to be connected to the Internet in an energy efficient manner. The newly released IEEE 802.11ah Wi-Fi standard is considered as a very promising technology for the IoT. One of the new features of IEEE 802.11ah, named Restricted Access Window (RAW), aims to increase efficiency in face of a large number of densely deployed and energy constrained stations. It divides stations into groups, limiting simultaneous channel access to one group, therefore reducing the collision probability and increasing scalability. The IEEE 802.11ah standard, however, does not specify how to configure the RAW grouping parameters. Therefore, this thesis aims to dynamically optimize RAW configurations in real time to adapt to the network conditions.

Firstly, the implementation of IEEE 802.11ah is detailed, along with experimental results to validate it.  Subsequently, the Traffic-Aware RAW Optimization Algorithm (TAROA) is proposed. TAROA introduces a traffic estimation method to predict the packet transmission interval of each station. By using the simulation results under saturated state, TAROA assigns stations to RAW groups according to the estimated traffic conditions, in order to maximize the throughput.  A further step is made by applying surrogate modelling. A surrogate model is an efficient mathematical representation of a black box system, it is based on supervised learning, and can be accurately trained with a few labeled sample data points. This research trains models for homogeneous networks to estimate RAW performance under a wide range of network and traffic conditions.

Based on the trained models, the Model-Based RAW Optimization Algorithm (MoROA) is proposed. MoROA inherits the traffic estimation method of TAROA, using the trained model to determine the optimal RAW configuration in real time through multi-objective optimization.  Finally, an advanced surrogate model is presented that can predict performance of heterogeneous networks. As heterogeneous networks have more parameters leading to an enormous design space, the training methodology is well designed, to speed up the training process and maintain relative high model accuracy.

In summary, the research develops solutions for real-time RAW optimization to support large scale IoT networks. An open-source IEEE 802.11ah simulator, a patent, multiple journal and conference papers have been produced as a result.