National Experimental Platform for Hybrid & Embedded Systems Technology (NEPHEST)

 

 

 

 

The objective of this document is to request a Military Interdepartmental Procurement Request (MIPR) from the Directorate for Computer and Information Science and Engineering (CISE) at NSF for jointly funding with the Information Technology Office (ITO) at DARPA the establishment of a National Experimental Platform for Hybrid and Embedded Systems Technology.

 

Background

 

The PITAC contended that the longstanding “crisis” in software quality and productivity threatens U.S. security and economic viability. A conspicuous example of the enormous opportunities and challenges before us is embedded computing – that is, computing operating with and controlling the physical world. The pervasiveness of this technology is well illustrated by the following facts: (a) the total shipment of microprocessor units (MPU)  and micro control units (MCU) in 1997 was over 4.4 billion units[1], of this about 98% related to embedded applications, (b) between 1994 and 2004 the need for embedded software developers is expected to increase 10 fold[2].

 

The profound technical and economic implications of embedded computing and the well documented difficulties of embedded system and software development present a significant challenge for the research community. The design of embedded systems is extremely hard, because it cannot be based on an idealized model of the real world. The  emerging research discipline on hybrid systems provides the theoretical foundations for modeling and analyzing the behavior of embedded systems and promises a much improved design technology. Recently, there is an increasing recognition in the research community[3] that existing software design techniques are not suitable for building large embedded software systems. The differences are fundamental, which requires a full re-thinking of basic principles. The U. S. Department of Defense, whose development programs are routinely plagued by flaws in embedded software concluded that “There may be some scientific problem which are intrinsic to all military systems that systems developers are not grasping”[4]. A recent study at DARPA[5] argued that today’s embedded software is being written without regard to the interaction of the software with the physical world. As a result a great deal of effort needs to be spent on detailed “by hand” testing and validation. This procedure is expensive and is prone to slippage of time lines since last minute “software glitches” are discovered.

 

Given the staggering impact of  embedded software on both the private sector and the Federal government, DARPA/ITO and NSF/CISE initiated significant investment in fundamental and applied research in hybrid and embedded systems. Currently identified NSF/CISE programs that are relevant to this MOA include elements of the SEL (Software Engineering and Languages) program, the OSC (Operating Systems and Compilers) program, and the DA (Design Automation) Program that are being assembled into an emphasis area in Hybrid and Embedded Systems.   The NSF ITR (Information Technology Research) program is also highly relevant. Currently identified DARPA/ITO programs relevant to this agreement are:  SEC (Software Enabled Control), PCES (Program Composition for Embedded Systems), MOBIES (Model Based Integration of Embedded Systems), Quorum, and NEST (Networked Embedded Systems Technology) representing an approximately $250M investment between FY2000 and FY2005. Future NSF and DARPA programs may also be relevant. These developments indicate a profound shift in the interest of the scientific community, the funding agencies and the industry toward embedded software and system development.

 

PROGRAM RATIONALE AND GOAL

 

Research in hybrid and embedded systems requires a strong experimental component using real-life challenge problems. A well-known problem is that the construction of experimental infrastructure suitable for large scale experiments is extremely expensive, which frequently leads to omitting or marginalizing the experimental tasks from research efforts all together. Unfortunately, this leads to questioning the validity of research results, and depriving the researchers from crucial feedback about practical relevance of the problems they solve and the feasibility of their approach.

 

Recognizing these difficulties, each program in the hybrid and embedded systems at DARPA/ITO uses program wide Open Experimental Platforms and a suite of related challenge problems. The role of the OEP-s is to demonstrate and measure the progress achieved in the program relative to an established baseline in a common, well-defined context. Although the DARPA OEP-s are specific to unique DoD needs, they convincingly showed tremendous advantages of the concept in coalescing the research programs, providing a common evaluation and verification platform for technologies and facilitating technology transfer toward industry.

 

Based on this experience we propose the establishment of a National Experimental Platform for Hybrid and Embedded Systems Technology. The goal of NEPHEST is to provide a common experimental infrastructure for hybrid and embedded systems research. NEPHEST will include the following major components:

 

1.      Open framework for integrating hardware and software components of large scale experiments for hybrid and embedded system research.

2.      Reusable hardware and software components and tools for experimental systems.

3.      An evolving suite of challenge problems, reference solutions and baselines for validation and verification of technologies.

4.      National repository of components, tools, verification results and documents.

 

The proposed NEPHEST will play the following role in federally funded hybrid and embedded system research:

 

1.      Satisfies the need of translation between theory and theoretical advances into experiments and the use of experimental findings in advancing theory.

2.      Provides affordable solution for conducting large-scale experimental research in hybrid and embedded systems.

3.      Facilitates the evaluation and verification of results via common, established and well-understood challenge problems and benchmarks.

4.      Accelerates technology transfer toward Federal programs and private industry.

5.      Creates a tangible, understandable showcase for demonstrating and accumulating the results of the ongoing major Federal research initiative in hybrid and embedded systems.

 

SCIENTIFIC AND NATIONAL SECURITY IMPACT

 

Fusion of  physical and information processes presents one of the most important opportunity and challenge for the scientific community. There are essential differences between embedded software development and software development in general.

 

§         Physicality. Embedded computing devices, viewed from their sensor and actuator interfaces, act like physical processes, with dynamics, noise, fault, size, power and other physical characteristics. Using current software technology, physical properties are not composable, they appear as cross-cutting constraints in the development process. Consequently, we need to change our approach to the design of embedded software: productivity increases must come from modeling methods and tools that directly address the design of the whole system with its many different physical, functional and logical aspects.

§         High Confidence. Embedded computing implements and controls physical interactions that have direct and immediate impact on the physical environment and the people involved. As a result, almost all embedded software is subject to high or extremely high assurance requirements. Advancement in design technology for embedded software and systems requires full integration of modeling and verification techniques for physical and computational systems, and in some cases, such as hybrid systems, the development of new theoretical foundations.

§         Systems with dynamic structure. Embedded systems are increasingly becoming distributed, requiring coordination among multiple information and physical processes. Building highly dependable, robust, distributed embedded applications with hundreds of thousands of nodes is significant software and systems challenge. 

 

The answer to these challenges requires advancement in a broad range of theoretical and applied research areas. In fact, the research will lead to the re-integration of information and physical sciences.

 

Fusion of  physical and information processes presents the single most important opportunity and challenge for DoD. From avionics systems to smart weapons, embedded information processing  is the primary source for superiority in weapon systems. The new wave of inexpensive MEMS-based sensors and actuators and the continued progress in computing and communication technology will further accelerate this trend. Weapon systems will become increasingly “information rich”, where  embedded monitoring, control and diagnostic functions penetrate deeper and with smaller granularity  in physical component structures. Given this trend, the relative separation of  physical and information processing architectures is not sustainable. Strong mutual interdependence requires coherent fusion at fine levels of granularity, i.e. the distribution of information processing among physical components. The coordinated operation of distributed embedded systems makes coordination, distribution and embedding the fundamental technical challenge for software. If successful, research on hybrid and embedded system technology will dramatically simplify the software development task in a wide range of future weapon systems. If  not done, application developers of future weapon systems will need to constantly reinvent  theoretically involved and computationally complex solutions for embedded coordination and synthesis, which can provide sufficient guarantees for predictable behavior of  large-scale systems. 

 

 

STATEMENT OF WORK

 

The jointly funded research between DARPA/ITO and NSF/CISE will complete the following tasks:

 

1.      Overview of federally funded research directions, major industrial R&D and standardization efforts, and available information on existing and predicted COTS products related to the NEPHEST goals.
Deliverable: State-of-the-art Report on hybrid and embedded systems technology.
Deadline: 2 months after start

2.      Architecture design and demonstration of an initial experimental prototype of an open framework for integrating hardware and software components of large scale experiments for hybrid and embedded system research.
Deliverables: Design document and demonstration system
Deadline: Design document 3 months after start; Demonstration system 10 month after start

3.      Reusable hardware and software components and tools for experimental systems.
Deliverables: Design document and examples for integratable hardware and software components.
Deadline: Design document 3 months after start; Demonstration system 10 month after start

4.      Formulating challenge problem concepts for major research directions in hybrid and embedded systems using input from existing DARPA OEP-s.
Deliverables:  Challenge problem specifications, and one example for a prototype challenge problem implementation in the NEPHEST framework.
Deadline: Design document 3 months after start; Demonstration system 10 month after start.

5.      Design of an open, web accessible national repository for components, tools, verification results and documents and demonstration of an early prototype concept.
Deliverables: Design document for the repository concept and demonstration of an early prototype.
Deadline: Design document 3 months after start; Demonstration system 10 month after start.

6.      Evaluation of the feasibility of the NEPHEST concept and design of the  implementation plan.
Deliverables: Evaluation document and Implementation plan.
Deadline: 11 months after start