Robust Adaptive Coordination in Dynamic Meshes of Networked Devices

The DyMND project, led by Lockheed Martin Space Systems Company, has assembled an outstanding team of academic and industrial researchers who bring unique technical strengths and broad experience "to produce abstract models, techniques, and tools" to enable the DARPA NEST program to meet the challenge of developing abstract models of NEST systems that aid in prediction and analysis of performance "in the large," and developing principled approaches to engineering of such systems. DyMND exploits the strengths of UC Berkeley (S. Sastry) in dynamical systems, autonomous, distributed control-based systems, and the Mote OEP for NEST, and Lockheed Martin (P. Bose) in embedded autonomy, software architectures, agent-based systems, and (E. Byler) distributed robotics. UC Berkeley research contributions on the DyMND project are: > Abstract models, architecture and analysis of DyMND systems for change management. The challenge involves developing abstract formal models, approaches, and tools that enable specification and analysis of architectures of DyMND systems for handling changes in the environment arising from security failures, multiple node faults, active jamming, and other environment changes. In this research area the DyMND project will develop: innovative distributed control based models for adaptive coordination based on the formalization of change; determination of localized influence on desired properties resulting from propagation of such changes; and the use of tools (Petri-net) for formal analysis of robustness properties of such systems. > Programming DyMND systems for adaptive coordination. NEST is developing protocols for coordination of sensing and actuation activities of DyMND nodes for specific tasks. A key challenge is principled approaches to DyMND programming that will enable the systematic and efficient development and analysis of programs for DyMND applications. Our key contributions to address such a challenge are: high-level languages and innovative techniques for modeling DyMND software architectures and specifying desired properties in terms of finite state machine (FSM) models of primitive behaviors that exploit the abstract models for adaptive coordination developed above and use real-time unified modeling language (UML), an industry standard in model-based specification and design of real-time systems; techniques for automated debugging and coding of components for the motes in the UCB OEP from the high level descriptions that exploit our understanding of concurrency of behaviors of components in the UCB OEP (TinyOS concepts) and exploit industrial strength advanced environments for design of realtime systems (e.g. Rationale Rose); and, techniques for instrumented simulation and analysis of largescale DyMND systems using advanced simulation tools.