Realization of dynamical electronic systems

¹ Dept. of Electronics and Telecommunications, Norwegian University of Science and Technology, 7491 Trondheim, Norway
² imec, 3001 Heverlee, Belgium and University of Leuven, 3000 Leuven, Belgium
³ Dept. of Bioengineering and Dept. of Electrical Engineering, Arizona State University, Tempe, AZ 85287-9709, USA
⁴ imec/Holst Center, The Netherlands

^a e-mail: elena.hammari@iet.ntnu.no

Published online: 10 April 2014

Abstract

This article gives an overview of a methodology for building dynamical electronic systems. As an example a part of a system for epileptic seizure prediction is used, which monitors EEG signals and continuously calculates the largest short-term Lyapunov exponents. In dynamical electronic systems, the cost of exploitation, for instance energy consumption, may vary substantially with the values of input signals. In addition, the functions describing the variations are not known at the time the system is designed. As a result, the architecture of the system must accommodate for the worst case exploitation costs, which rapidly exceed the available resources (for instance battery life) when accumulated over time. The presented system scenario methodology solves these challenges by identifying at design time groups of possible exploitation costs, called system scenarios, and implementing a mechanism to detect system scenarios at run time and re-configure the system to cost-efficiently accommodate them. During reconfiguration, the optimized system architecture settings for the active system scenario are selected and the total exploitation cost is reduced. When the dynamic behavior is due to input data variables with a large number of possible values, current techniques for bottom-up scenario identification and detection becomes too complex. A new top-down technique, based on polygonal regions, is presented in this paper. The results for the example system indicate that with 10 system scenarios the average energy consumption of the system can be reduced by 28% and brought within 5% of the theoretically best solution.