End-to-end prototyping and validation for health management sensor networks

The cost and complexity of aerospace platforms currently under development, from the Joint Strike Fighter (JSF) to NASA\´s Crew Exploration Vehicle (CEV), requires integrated systems that monitor system state, diagnose failures, and correct faults with minimal intervention. Embedded sensor networks with large numbers of integrated low-power processors, sensors, and network transceivers suggest one approach. As sensor technology matures, tighter integration of sensors and power supplies enable "motes-on-a-chip" that subsist indefinitely on power drawn from their environment. As presented in K. S. J. Pister et al. (1999) these large-scale embedded networks will rely on local computation and actuation and short-range networking to maintain system health. The processing constraints of low-power processors, nonlinear effects of small local actuators distributed in a large structure, and a noisy communications environment require a simulation environment that efficiently reproduces both the constraints of the embedded programming environment and the rich effects of the deployment environment. This paper presents a simulation tool called DyMND-EE that balances fidelity and efficiency by combining source-level execution of embedded code using the Emstar runtime based in J. Elson et al. (2003) and sensing and actuation models based on the Ptolemy actor-oriented modeling environment presented in P. Baldwin et al. (2004). DyMND-EE offers a "round-trip" simulation and emulation environment that allows developers to implement prototypes of sensing and networking services as Ptolemy actors then incrementally port them to the embedded environment for execution either in Linux user space or on board a wireless sensing node like the Berkeley mote. The present work explores wireless nodes that combine sensing with local actuation capabilities. Such nodes expand the role of the network from monitoring and diagnosis to fault prevention and recovery. We use DyMND-EE to simulate node loc- - alization and source localization of generic wireless sensors, such as those proposed for use in health management applications, and reconfiguration and health management of reconfigurable phased-array radar consisting of regular tiles