Simplifying Design of Wireless Sensor Networks with Programming Languages, Compilers, and Synthesis.

Wireless sensor networks have opened opportunities for new applications and attracted users from domains beyond computer system design. Sensor network design is challenging. It is generally an ad hoc process carried out by embedded system experts. In this dissertation, we argue that human efforts necessary to the design of sensor networks can be reduced with the help of high-level specification languages, compilers, and synthesis tools. We designed and implemented a framework to simplify and automate the design of a class of sensor network applications. Our results show that a sensor network novice given only a few pages of instructions, can successfully specify sensing applications within 30 minutes, compared with hours or days required by prior approaches. Within approximately 30 minutes, our modeling and design exploration techniques translate these specifications into implementations, automatically selecting from among 405,790 designs. Moreover, our memory management and compiler-assisted techniques make difficult-to-implement optimizations available to novice programmers, enabling better tolerance of sensor faults and making 39% more usable memory available than would otherwise be the case. We propose a design process that decouples specification from implementation. Application designers specify abstract functionality and design requirements. Compiler and synthesis tools automatically determine implementation details, optimizing design parameter optimization and generating code. First, we develop a design process in which programming novices (e.g., application experts) use high-level, specification languages designed for particular classes of applications. We focus on the class most commonly encountered in sensor network deployment publications. Second, we develop two compiler and runtime techniques to relieve application experts from explicitly dealing with sensor faults and limited memory, two common sources of sensor network design complexity. The first technique automatically generates code for fault detection and error estimation using easy-to-specify hints. The second technique automatically generates code for online memory compression, thereby increasing effective memory. Finally, we develop modeling and optimization techniques to determine high-level design parameters to meet specified design requirements. We present an automated technique that constructs fast and accurate systemlevel models for sensor networks and an optimization technique that uses these models to rapidly search for the optimal design(s). Our evaluation focuses on homogeneous environments.Ph.D.Computer Science & EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/86465/1/lanbai_1.pd