Wireless mesh network (WMN) is self-constructing and self-healing. Because of its low cost and easy-deployment features, WMN can be used in many scenarios such as emergency communication, Internet of Things, communication in rural areas and so on. Mesh routers and mesh clients compose WMN, and mesh routers are more stable and have stronger abilities. Routing is very important and non-effective routing will bring more interference and unbalanced loads. Most existing routing methods for WMN are reactive, which does not take advantage of the stability of mesh routers. In addition, load, interference, and energy are not considered comprehensively. In this thesis, we overcome the weaknesses of existing routing methods. As centralized proactive and distributed reactive routing methods are suitable for stable mesh routers and mobile mesh clients respectively, we first give a hybrid routing method for WMN including both static mesh routers and mobile mesh clients. In addition, the centralized proactive routing method among infrastructure WMN composed of stable mesh routers is very important. To improve the proactive routing in the proposed hybrid routing, delay is minimized in the optimization model in our second work. Then, to overcome the neglect of energy constraint in these two works, energy harvesting is considered in the third work. Finally, unlike the third work needing to set weight values for different objectives in the routing problem, our fourth work proposes a multi-objective routing with reinforcement learning. In general, the four main works of the thesis are given as follows. First, a hybrid routing method is proposed. The current weaknesses of neglecting regional conditions and not considering whole proactive path conditions when mesh clients accessing mesh routers are overcome. The load conditions are considered for mesh routers. Mobility and energy are further researched for mesh clients. In addition, the region with heavy load will be avoided. This proposed method brings 10.19% and 33.29% improved performance than two current hybrid routing methods respectively. Second, for the infrastructure WMN part, the centralized proactive routing method is designed. The routing problem is formulated as an optimization problem from a global network view. Minimizing delay is set as the objective, which is derived by considering interference, packet transmission failure, and bandwidth. The overhead of obtaining delay is reduced and load is balanced by an improved genetic algorithm. 29.03% and 50.20% better performance than two current methods is gotten. Third, a centralized routing method with a long-term objective to balance load and energy consumption is further proposed. Load, channel condition, energy cost and energy harvesting are considered. This proposed routing can obtain 1.69%, 10.33% and 26.2% better performance compared with three other routing methods. Fourth, to consider delay and energy efficiency at the same time without setting weight values like the previous third work, a multi-objective routing is proposed. The routing problem with uncertain network conditions can be solved. Dyna-Q is used to solve the multi-objective routing problem for the first time. The performance is 42.12%, 43.96% and 47.54% better than three other state-of-the-art routing methods.
|Date of Award
|1 Aug 2022
- The University of Manchester
|Xiaojun Zeng (Supervisor) & Ke Chen (Supervisor)