Use this URL to cite or link to this record in EThOS:
Title: Self-healing in wireless sensor networks
Author: Bourdenas, Themistoklis
ISNI:       0000 0004 2718 3580
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
Wireless sensor networks (WSNs) and pervasive systems are increasingly used for applications such as building monitoring and control, health-care and environmental monitoring. The users are frequently non-technical and devices may not be easily accessible, thus their management and complexity should be transparent to the users. To this extent, the systems need to be self-healing, able to respond to failures. We extend previous work on self-managed cell (SMC), which introduced an infrastructure for autonomous pervasive systems, with fault detection and recovery services. We present a middleware for constrained platforms, which supports dynamic adaptation of network components imposing small overheads. It provides an event-driven paradigm for expressing system behaviour based on policies. We identify and define sensor readings' fault models extracted from long-running, real-world sensor deployments. We describe a fault detection mechanism for sensor readings based on heuristic and Bayesian probabilistic approaches that accurately identifies error occurrences in readings and minimises false positives. We implemented a recovery mechanism that responds to sensor and communication link degradation to dynamically reorganise the original role and task allocation among sensor nodes without disrupting service operations. Finally, we present a case study on a production-quality, multi-hop routing middleware, ITA Sensor Fabric, where we prototyped an adaptive routing mechanism, which set-ups virtual circuits for sensor data subscriptions avoiding recurring communication link and traffic congestion patterns that appear in the network. Evaluation of the framework shows that the embedded policy management system is lightweight for power constrained nodes. The self-healing service accurately identifies erroneous sensors and is capable to effectively reconfigure network assets to improve quality of information while maintaining long life expectancy of the system.
Supervisor: Sloman, Morris ; Lupu, Emil Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral