Multi-timescale event-scheduling in multi-agent immune simulation models

Multi-agent (or MA) -based design approaches have received much research attention lately for modeling immunological systems due to their efficacy in representing non-heterogeneous behaviors in the population under dynamic environmental and topological conditions. The update scheme of a MA model refers to the frequency of agent state updates and how these are related in temporal order. In contrast to verifiable agent behavioral rules at the individual level, the update scheme is a design decision made by the model developer at the systems level that is subject to realism and computational efficiency issues that directly affect the credibility and the usefulness of the simulation results. Previous works have mainly focused on the issue of realism with respect to synchrony of updates but have often overlooked the necessary heterogeneity in update frequencies due to multi-timescales in immunological phenomena. To incorporate such multi-timescales for realism, the efficiency of the update scheme arises as a nontrivial issue. An event-scheduling based asynchronous update scheme has the advantage of allowing arbitrary smaller timescales for realism and avoiding unnecessary execution and delays to achieve efficiency. In this paper we present the application of the event-scheduling update scheme to realistically model the B cell life cycle, and empirically compare its simulation performance with the widely adopted uniform time-step update scheme. The simulation results show a significantly reduced execution time (40 times faster) and also reveal the conditions where the event-scheduling update scheme is superior.