Optimistic Parallel Simulation of Tightly Coupled Agents in Continuous Time

Agent-based simulations relying on synchronous state updates using a fixed time step size are considered attractive candidates for parallel execution in order to reduce simulation running times for large and complex scenarios. However, if the underlying models are formulated with respect to continuo...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2021 IEEE/ACM 25th International Symposium on Distributed Simulation and Real Time Applications (DS-RT) s. 1 - 9
Hlavní autoři: Andelfinger, Philipp, Uhrmacher, Adelinde
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 27.09.2021
Témata:
agent-based simulation
Approximation algorithms
Concurrent computing
Markov processes
Memory management
multi-agent simulation
optimistic synchronization
parallel simulation
Real-time systems
Semantics
Topology
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Agent-based simulations relying on synchronous state updates using a fixed time step size are considered attractive candidates for parallel execution in order to reduce simulation running times for large and complex scenarios. However, if the underlying models are formulated with respect to continuous time, a time-stepped execution may only approximate the strict model semantics. To simulate continuous-time agent-based models, parallel discrete event algorithms can be applied. Traditionally those are based on logical processes exchanging time-stamped events, which clashes with the properties of models in which tightly coupled agents frequently access each other's states. To illustrate the challenges of such models and to derive a solution, we consider the domain-specific modeling language ML3, which allows modelers to succinctly express transitions and interactions of linked agents based on a continuous-time Markov chain (CTMC) semantics. We propose an optimistic synchronization scheme tailored towards simulations of fine-grained interactions among tightly coupled agents in highly dynamic topologies. By restricting the progress per round to at most one state change per agent, the synchronization scheme enables efficient direct read and write accesses among agents. To maintain concurrency given actions that depend on dynamically updated macro-level properties, we introduce a simple relaxation scheme with guaranteed error bounds. Using an extended variant of the classical susceptible-infected-recovered network model, we demonstrate that the proposed synchronization scheme accelerates simulations even under challenging model configurations.
DOI:10.1109/DS-RT52167.2021.9576156