Agent-Based Modeling and Simulation of Intelligent Distributed Scheduling Systems

Agent-Based Modeling and Simulation of Intelligent Distributed Scheduling Systems

Milagros Rolón (INGAR (CONICET-UTN), Argentina) and Ernesto Martínez (INGAR (CONICET-UTN), Argentina)
DOI: 10.4018/978-1-4666-2098-8.ch002
OnDemand PDF Download:
No Current Special Offers


For responsiveness and agility, disruptive events must be managed locally to avoid propagating the effects along the value chain. In this work, a novel approach based on emergent distributed scheduling is proposed to overcome the traditional separation between task scheduling and execution control. An interaction mechanism designed around the concept of order and resource agents acting as autonomic managers is described. The proposed Manufacturing Execution System (MES) for simultaneous distributed (re)scheduling and local execution control is able to reject disturbances and successfully handle unforeseen events by autonomic agents implementing the monitor-analyze-plan-execution loop while achieving their corresponding goals. For detailed design of the autonomic MES and verification of its emergent behaviors, a goal-oriented methodology for designing interactions is proposed. Encouraging results obtained for different operating scenarios using a generative simulation model of the interaction mechanism implemented in Netlogo are presented.
Chapter Preview


There exists a definitive trend towards introducing agility, adaptability, autonomy and flexibility in production systems to face successfully highly dynamic and uncertain environments (ElMaraghy, 2009). Conventional production systems typically works following a two-tier hierarchy comprising of monolithic schedule generation (upper layer) and execution control (lower layer). Scheduling and planning systems are predominantly centralized systems aiming at one-time global optimization of resource usage and processing performance (Valckenaers et al, 2007; Verstraete et al, 2008). Manufacturing execution system (MES) attempts to follow a given schedule as closely as possible. The MES performs this task in a reactive manner, filling-in missing details, providing alternatives for unfeasible assignments, handling auxiliary tasks, and so on based on shop-floor information and real-time control (Kletti, 2007). This quasi-standard of rigid, hierarchical planning and control architectures in today’s industry has been unable to cope with the new challenges of agility and self-configuration successfully, since the production schedules and plans are known to become ineffective after a short time on the shop floor. Centralized production planning and control systems are therefore vulnerable to abrupt changes and unforeseen events in production processes. Furthermore, there is an increasing trend towards inter-firm integration through enterprise networking (Ueda, 1992; Warnecke, 1993; Van Brussel et al, 1998; Canavesio and Martínez, 2007; Wiendahl et al, 2007) which gives rise to the need for responsive (re)scheduling using distributed decision-making and local control systems.

For effectiveness, decentralized MES must be designed so as to address disruptive events seeking robustness rather than optimality (Valckenaers et al, 2007). In decentralized MES, production control is not carried out by a central control unit but it is rather an emergence from the actions and interactions of local controllers in the system (Wang and Usher, 2004).

This chapter proposes an entirely new methodology for design and verification of autonomic MES. In each autonomic unit, the agent playing its manager role implements the monitor-analyze-plan-execute (MAPE) loop which comprise of both scheduling and control function for a given order or resource. For detailed design and behavior verification of the autonomic MES, an extension of the Prometheus-Hermes methodology (Cheong and Winikoff, 2006) is proposed by including generative simulation and behavior verification. The design methodology highlights the goal hierarchy, action sequences and a number of failure recovery procedures to provide design guidelines when specifying goal-oriented interactions. The autonomic MES is made up of a society of agents, each one having cognitive capabilities such learning, reasoning and planning which allow the agent to know what it is doing (Brachman, 2002). For behavior verification in a case study, a generative simulated model in Netlogo has been created and some results obtained for normal and disturbed operating scenarios are presented. A remarkable result is the stability of the autonomic MES despite the dynamic complexity resulting from goal-oriented interactions among a number of autonomic agents. Also, emergent behaviors of the interaction mechanism in abnormal scenarios highlight the importance of generative simulation in designing autonomic MES.

Complete Chapter List

Search this Book: