Unplanned Event Impact Assessment in Multiproject (Re)scheduling through Agent-based Simulation

LAURA R. TOSSELLI; VERÓNICA BOGADO; ERNESTO MARTÍNEZ

San Miguel del Tucumán

Congreso; ARGENCOM 2018; 2018

IEEE

he concurrent execution of multiple projects might be affected by the occurrence of unplanned events, leaving obsolete the initial schedule resulting from the individual actions of the different agents involved. Therefore, computational tools that allow impact assessment of unplanned events or significant disturbances are necessary in order to anticipate the outcome of the interactions between selfish agents, generating a compromise solution to face the uncertainty and the dynamics of the environment. In this work, an interaction mechanism based on an agent-based simulation model to evaluate how individual selfish learning may affect the overall solution to the scheduling problem is presented. This mechanism includes an assessment process in terms of violations to project constraints, such as time and cost, where each agent learns to select their next action which represents the best response for achieving its planned objectives. Thus, each project is able to adapt its initial schedule to the changes caused by unforeseen events. The proposed mechanism is implemented and incorporated into a Multi-agent System (MAS) for the Multi-Project (Re)scheduling Problem, where decentralized interaction protocols and decoupled learning rules are used based on learning by trial-and Error approach. The Agent-based Simulation Model allows obtaining a new equilibrium schedule in a decentralized way, where selfish agents decide the response to other agents´ actions using schedule indicators that allow evaluating the impact of each unplanned event.