|Title:||Self-Healing Wireless Sensor Networks|
|Affiliation:||Imperial College London|
PhD project description
The development of small wireless sensors and smart-phones, which include various sound, video, motion and location sensors have facilitated realising new pervasive applications. These include health-care applications to monitor physiological parameters such as blood-pressure, heart-rate, temperature or ECG of at-risk patients as well as determining their activity; monitoring and controlling temperature, humidity and lighting levels in buildings; environmental monitoring and flood warning and even tracking wildlife movement. These pervasive systems are expected to perform in a vast number of environments, ranging from urban to rural, with di�erent requirements and resources. They are deployed in harsh environments and application requirements may change dynamically requiring flexible adaptation. Users may be non-technical so the systems need to be self-managing. Some applications such as health-care may be life-critical and devices may be inaccessible for repairs so self-healing with respect to faults and errors is important.
Pervasive computing system rely on Wireless Sensor Networks (WSNs) for receiving feedback from the environment they interact with. Sensor networks use devices, known as `motes', that have limited processing and power resources. They have to cater for deterioration of sensor accuracy over time due to physical phenomena such as overheating or chemical fouling as well as external factors such as low battery levels or physical damage. The quality of wireless links may vary particularly with mobile systems and devices may completely fail.
Pervasive computing involves large number of components often integrated into the environment. Frequent replacement of devices and manual recalibration are impractical. They hinder adoption and use of such systems by non-expert users with limited technical skills. A self-healing framework that supports fundamental reusable components for failure detection and recovery is required to boost the implementation and adoption pace of pervasive applications. Self-healing implies, �rstly, a taxonomy of fault classes to identify the characteristics of faults enabling their detection; secondly a range of mechanisms supporting recovery for the various classes of faults and �finally a flexible infrastructure to facilitate context aware adaptation in the face of sub-components failures. We focus on data faults that manifest on the sensing devices as a result of inaccuracies, noise or drift and expand on fail-stop functional failures of network nodes.