| To ensure that a system is robust and will continue operation even when facing disruptive or traumatic events, we have created a methodology for system architects and designers which may be used to locate risks and hazards in a design and enable the development of more robust and resilient system architectures. It uncovers design vulnerabilities by conducting a complete exploration of a systems' component operational state space by observing the system from multi-dimensional perspectives and conducts a quantitative design space analysis by means of probabilistic risk assessment using Bayesian Networks. Furthermore, we developed a tool which automated this methodology and demonstrated its use in an assessment of the OCTT PHM communication system architecture.;To boost the robustness of a wireless communication system and efficiently allocate bandwidth, manage throughput, and ensure quality of service on a wireless link, we created a wireless link management architecture which applies sensor fusion to gather and store platform networked sensor metrics, uses time series forecasting to predict the platform position, and manages data transmission for the links (class based, packet scheduling and capacity allocation). To validate our architecture, we developed a link management tool capable of forecasting the link quality and uses cross-layer scheduling and allocation to modify capacity allocation at the IP layer for various packet flows (HTTP, SSH, RTP) and prevent congestion and priority inversion.;Wireless sensor networks (WSN) are vulnerable to a plethora of different fault types and external attacks after their deployment. To maintain trust in these systems and increase WSN reliability in various scenarios, we developed a framework for node fault detection and prediction in WSNs. Individual wireless sensor nodes sense characteristics of an object or environment. After a smart device successfully connects to a WSN's base station, these sensed metrics are gathered, sent to and stored on the device from each node in the network, in real time. The framework issues alerts identifying nodes which are classified as faulty and when specific sensors exceed a percentage of a threshold (normal range), it is capable of discerning between faulty sensor hardware and anomalous sensed conditions. Furthermore we developed two proof of concept, prototype applications based on this framework. |
