Distributing Scenario-based Models: A Replicate-and-Project Approach

Abstract

In recent years, scenario-based modeling has been proposed to help mitigate some of the underlying difficulties in modeling complex reactive systems, by allowing modelers to specify system behavior in a way that is intuitive and directly executable. This modeling approach simplifies the specification of systems that include events occurring in distinct system components. However, when these system components are physically distributed, executing the scenario-based model requires inter-component coordination that may negatively affect system performance or robustness. We describe a technique that aims to reduce the amount of joint eventselection decisions that require coordination and synchronization among distributed system components. The technique calls for replicating the entire scenario-based executable specification in each of the components, and then transforming it in a component-specific manner that induces the required differences in execution while reducing synchronization requirements. In addition to advantages in streamlining design and improving performance, our approach captures the fact that in certain "smart" distributed systems it is often required that components know what rules govern the behavior of other components. Our evaluation of the technique shows promising results.

Scenario-Assisted Deep Reinforcement Learning

“…Thus, it is often possible to apply formal verification to ensure that a scenario-based model satisfies various criteria, either as a stand-alone model or as a component within a larger system. Automated analysis techniques can also be used to execute scenariobased models in distributed architectures (Harel et al, 2015a;Steinberg et al, 2018;Steinberg et al, 2017;Greenyer et al, 2016b;Harel et al, 2013a), to automatically repair these models Harel et al, 2012a;Katz, 2021b), and to augment them in various ways, e.g., as part of the Wise Computing initiative (Harel et al, 2018;Marron et al, 2016;Harel et al, 2016a;Harel et al, 2016b).…”

Guarded Deep Learning using Scenario-based Modeling

“…These aspects include the automatic repair (Harel et al, 2012a), verification (Harel et al, 2015b), synthesis (Greenyer et al, 2016a) and optimization (Harel et al, 2015a;Greenyer et al, 2016b;Steinberg et al, 2017;Steinberg et al, 2018;Harel et al, 2020) of models.…”

“…These aspects include the automatic repair (Harel et al, 2012a), verification (Harel et al, 2015b), synthesis (Greenyer et al, 2016a) and optimization (Harel et al, 2015a;Greenyer et al, 2016b;Steinberg et al, 2017;Steinberg et al, 2018;Harel et al, 2020) Harel et al, 2018), which seeks to transform the computer into a proactive team member, capable of developing complex models alongside human engineers.…”

Towards Repairing Scenario-Based Models with Rich Events

“…This compatibility stems from the fact that scenario objects in a scenariobased (SB) model interact through the well defined interface of the event selection mechanism, making it possible to automatically construct a model of the full, composite system, and then analyze it. Tools and techniques have been devised to model-check and repair safely and liveness violations in SB models (Harel et al, 2011;Harel et al, 2011), to automatically optimize and distribute these models (Harel et al, 2013a;Harel et al, 2015a;Steinberg et al, 2017), and to identify emergent properties thereof (Harel et al, 2018).…”