|
||||||||||||||||
|
Ph.D. ThesesOpportunistic Routing and Middleware Composition for Sensor and Actuator Networks
By Joel W. Branch
Designing a receiver-decided cost-based routing protocol with transmission back-off delay and a local route repair algorithm establishes a novel and efficient wireless sensor network routing protocol that is fault-tolerant and seamlessly accommodates energy-efficient topology-control algorithms. Likewise, designing a new high-level middleware framework creates a foundation for developing extensible environmental control systems that may additionally require coordinating the behavior of coinciding actuators with interfering goals. There exist many wireless sensor network routing protocols that achieve both fault-tolerant and energy-efficient routing. In many of these protocols, a node must maintain its neighbors' states (e.g., active, sleeping, dead) and adjoining links' qualities in order to support routing decisions. This approach is often inefficient for fault-prone environments because it requires additional communication overhead to route packets in a dynamic manner. Receiver-decided cost-based routing is better designed for communication in fault-prone environments since there is no reliance on maintaining neighbors' or links' states. This significantly lowers the overhead required to make dynamic routing decisions. Additionally, this technique is very extensible, easily allowing routing decisions to be guided by a combination of different criteria. The first contribution of this thesis is a receiver-decided cost-based routing protocol that uses transmission back-off delay and an optional local route repair algorithm to support fault-tolerant communication. Additionally, this protocol naturally accommodates out-of-band topology-control algorithms, which are often employed to reduce sensor network energy consumption. We show by simulation that this protocol performs better than typical static routing when dynamic network topologies are encountered. There exist research activities that propose frameworks for supporting the composition of wireless sensor network middleware and applications. However, many produce device-level software that is often platform-specific. There exists a growing need for domain-specific frameworks to promote software extensibility and reusability and integration with larger monitoring systems. Additionally, there is a growing need to coordinate actuators' behavior in response to sensor data and other actuators' behavior. The latter is critical because distributed control applications may manage actuators with individual tasks that are dependent on a shared pool of resources (e.g., treatment supplies or energy), significantly affecting an application's fidelity. Therefore, the second contribution of this thesis is the design of a high-level middleware framework to support the composition of sensor and actuator network systems that require the coordination of coinciding actuators in a distributed manner. An extensible set of manager components, supporting the essential tasks of sensor and actuator control systems, serves as the basis for system composition. A lightweight budgeting and auction-based mechanism is used to coordinate distributed actuators' access to shared and limited pools of resources. We describe the experience of using the framework to compose extensible sensor and actuator network middleware and show the benefits of its coordination-related features.
|
||||||||||||||||
Research
