IUPUI

Presented by Dr. Jules White of the Department of Electrical Engineering and Computer Science at Vanderbilt University

ABSTRACT: A cyber-physical system (CPS) involves close coordination between the system's computational and physical elements. CPS applications are essential in mission- and safety-critical domains, such as aerospace, automotive, chemical processes, civil infrastructure, energy, entertainment, healthcare, manufacturing, and transportation. Software for CPS applications has traditionally been tightly-coupled to the underlying hardware, which makes it hard to evolve and validate. Recent trends in CPS application development are decoupling software from the underlying hardware and providing much more flexibility in software component deployment and configuration. The deployment and configuration of CPS components can significantly can significantly impact the cost, weight, power consumption, and performance of a system. Optimizing the deployment and configuration of CPS software is hard, however, due to the need to meet complex real-time scheduling, memory, network, and cost constraints. This talk presents search-based techniques for optimizing the deployment and configuration of CPS software and shows how deployment and configuration can be transformed into multi-dimensional multiple-choice knapsack (MMKP) problems to facilitate automation with heuristic algorithms. The talk also shows how evolutionary and swarm intelligence algorithms can optimize the deployment and configuration of software components to minimize consumed network bandwidth while respecting CPS application scheduling constraints. Empirical results are presented based on applying these algorithms to several representative CPs applications, including a production avionics system. The results show how these algorithms substantially reduced required processing hardware and consumed network bandwidth.

Dr. Jules White is a Research Assistant Professor at Vanderbilt University. He received his BA in Computer Science from Brown University, his M.S. from Vanderbilt University, and his Ph.D. from Vanderbilt University. His research focuses on applying search-based optimization techniques to the configuration of distributed, real-time, and embedded systems. In conjunction with Siemens AG, Lockheed Martin, AFRL, IBM and others, Jules has developed scalable constraint and heuristic techniques for software deployment and configuration. Jules is the project-lead of the Eclipse Foundation's Generic Eclipse Modeling System (GEMS--http://www.eclipse.org/gmt/gems).

This seminar will be held on Friday, November 6th, 2009 at 11:00 am in SL 150.