Leveraging biologically-inspired self-adaptation architectures in wireless sensor networks.

Wireless sensor networks (WSNs) have been used to detect events and/or collect data in various domains such as environmental observation, structural health monitoring and human health monitoring. Due to their deeply-embedded pervasive nature, WSNs have a potential to revolutionize the way that humans understand and construct complex natural/physical system. Generally speaking, sensor nodes convey similar requirements and constraints. For example, they are deployed in an unattended area and expected to function in a long period of time. Also, each sensor node has limited resources, e.g., processing power, network bandwidth, and power availability. Due to the requirement and constraint, WSNs face several issues such as autonomy, adaptability and simplicity. WSNs are required to operate without the aid from human administrators. WSNs are also required to adapt their configuration to the changes in network/physical condition in order to maintain their performance. WSN applications need to be simple in its design and small in its footprint because of limited availability of CPU power and memory. To address the aforementioned issues, this dissertation investigates BiSNET, a biologically-inspired application architecture for resource-constraint WSNs, and BiSNET/e, a biologically-inspired application architecture with evolutionary adaptation mechanism. BiSNET implements a series of biologically-inspired mechanisms to support autonomous and adaptive applications. BiSNET/e is an adaptive application architecture for WSNs which introspectively understand the performance objectives such as latency and battery consumption, find optimal applications' configurations under given constraints and autonomously adapt to dynamics of the network such as node/link failures. BiSNET/e allows applications to improve their performance against the network condition. The evaluation results of BiSNET and BiSNET/e indicate that they allow WSNs to be autonomous, adaptive and simple.