System design trade-offs in a nextgen embedded wireless platform -

System design trade-offs in a nextgen embedded wireless platform

Over the course of the past decade, the evolution of advanced low-energy microcontrollers has raised three questions which this paper outlines and addresses.

The first question is: Can a 32-bit platform be constructed that provides advanced features but fits within the energy constraints of a wireless sensor network?

We answer this in the affirmative by presenting the design and preliminary evaluation of Storm – one such system based on an ARM Cortex-M4 that achieves 2.3µA idle current with a 1.5µS wake up time.

The second question we answer is: Can this platform simultaneously meet the very different demands of both monitoring-type applications and cyber-physical systems?

We demonstrate that this is indeed possible and present the design trade-offs that must be made to achieve this, yielding a module with a rich set of exported peripherals that fits in a 16mm x 26mm form factor.

The final question explored by this paper is: If such a platform is possible, what new opportunities and challenges would it hold for embedded operating systems?

We answer this by showing that the usage of modern 32 bit microcon- trollers requires reconsidering system architecture governing power management, clock selection and inter-module dependencies, as well as offering opportunities for supervisory code and the coordination of common tasks without CPU intervention. In particular, we address three basic questions:1. Can we now utilize full-featured, 32-bit microcon- trollers with enough memory and flash to support so- phisticated applications with the power profile of a mote, i.e., idle power of a few uWs, fast wake up, and efficient active operation?

2. Can the platform serve the distinct needs of the two dominant usage models: wireless monitoring, with a few sensors and predictable behavior and cyber- physical systems with rich I/O, actuation, and dynamic variation?

3. If so, does such a platform introduce qualitatively new operating system challenges and opportunities?

We show by developing a new platform around specific offerings in the Cortex-M family that the answer to the first two questions is affirmative and by examining aspects of this solution we outline a new suite of important system opportubities and challenges.

Indeed, the building blocks are finally of a state where the integration into a system-on-a-chip is likely to produce extremely general, cost-effective solutions.

Addressing the first question requires not just an analysis of data sheets; a quantitative, empirical study of the complex- ities and implications of utilizing next-generation hardware in sensor networks requires the careful design of a physical platform.

Storm is an example reference platform based upon best-in-class next-generation components. The process of mapping the model of a representative wireless embedded system into a physical instantiation by evaluation of available components and selective design trade-offs is discussed.

A physical module design is presented that extends and serves the range of usage models from simple sen-sor networks for monitoring to sophisticated cyber-physical systems.

The Storm platform is then used as a representative for next-generation wireless platforms in general for an exploration of new systems opportunities and challenges.

We identify five primary factors – modular power management, multiple clock domains, inter-module compatibility, chaining of multiple overlapping transfers and in creased supervisory control – which lead to a whole-system optimization framework for real time embedded operating systems, such as TinyOS.

Such intricacies naturally pose new problems for the architecture of any embedded operating system aiming to abstract device-specific complexity from users by utilizing layering and modularity.

To read this external content in full, download the complete paper from the open online author archives at the University of California, Berkeley.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.