Simulation-Based Analysis of Cyberphysical Systems

Cyberphysical systems consist of physical parts and control parts both of which can be very complicated. Especially analysis models of physical parts sometimes need lots of details to examine various aspects of the target behaviors accurately. For such complicated systems, dynamic approach, such as simulation, can be relatively easier or simpler to be applied than static one, although the results of simulations must be carefully analyzed having in mind that simulations give only small sets of execution traces. In this paper, we first discuss about our experiences in VLSI circuit designs on various models and their accuracy. Delays, power consumption, thermal things, and others, which are closely interrelated, must be examined to see if the circuit works or not. Then we discuss about acceleration of simulations with hardware, i.e., FPGA. Starting with a pure procedural implementation of the target simulator, we discuss about how stream-based processing can be introduced and how steams help in accelerating simulations. We use Tsunami wave propagation as an example and clarify the approach. In the last part, we discuss about how such implementation of simulation software and hardware can be practically verified. Assertions on floating point variables which describe internal behaviors are first automatically extracted and then they are examined to check with the original problem definition, such as partial differential equations. All of the issues discussed in this paper is based on dynamic analysis with small amount of static one.