CertiCon Applied Research
Solution to Industrial Production
The multi-agent production planning
system can be employed to provide
planning in the environment where the
features like distributiveness, modularity
and easy reconfiguration are required.
The general concept of agent-based
planning tools can be modified for
individual task definition and integrated
into a more complex information ERP
system as one of its subsystems.
This research has been supported in part by European Commission's Information Technologies Programme IST within the frame of the projects:
ExPlanTech – IST-1999-20171, Exploitation of Agent-based Production Planning Using the ProPlant Technology, years 2000–2002;
ExtraPlanT – EUTIST AMI IST-2000-28221, Customization and Validation of the ProPlant Multi-Agent Technology in the Area of Extra-Enterprise
Production Planning, Supply Chain Management and Optimization of Virtual Enterprise Operation, years 2002–2003.
Introduction Multi-Agent Systems
Automated workflow management systems have
Almost all of the project-oriented enterprises are facing proved to be the driving force behind successful
the following problem: How to set the optimal decision making in industry. Multi-agent systems (MAS)
production plan to effectively exploit the company technology offers a convenient platform for workflow
resources while reaching the highest turnover? To keep modelling. An environment where each agent
business running and to avoid such critical situations represents a real information unit of the modelled
when the company production is not balanced (e.g. enterprise is an appropriate model for optimization and
the company does not receive an optimal number of visualization of flows of material, work, and information.
production orders able to manufacture) an extra
attention to a production planning must be paid.
Multi-agent production planning tool could be a choice
for SME’s where standard solution fails or is not
affordable. The solution is suitable for enterprises using
heterogeneous legacy systems.
The used technology is based on the latest research
achievements in the fields of distributed artificial
intelligence, agent based computing and production
Agent (Software agent) – An agent is a computer Intelligent reasoning about the enterprise resources
system, that is situated in some environment, and that is with the aim to produce accurate estimation of
capable of autonomous action in this environment in project’s deadline and costs
order to meet its design objectives. System state update and maintenance
Re-plan, i.e. maintenance of the plans created so far
Multi-agent system – A multi-agent system (MAS) in such a way that they are dynamically updated
consists of a group of agents that can take specific with respect to eventual change in co-operating
roles within an organizational structure. Multi–agent agents (e.g. resource/agent breakdown).
system (multiple interconnected agents) represents the
decentralized nature of the problem. The agents need Besides workflow management, the multi-agent
to interact with one another to meet their individual technology is very much suitable for manufacturing
objectives/goals and common goal as well. processes simulation, production control as much as
scheduling, planning and re-planning.
Agentification – An agent wrapper implements the
agent-to-agent communication and integrates a pre- Integration
existing software system. Such a software system The enterprise management does not need to replace
becomes an agent – a fully-fledged member of the the currently running information systems by a
multi-agent system. completely new technology, while it make the best use
FIPA – The Foundation for Intelligent Physical Agents FIPA of the technology they already have and the modern
is a non-profit organisation aimed at producing multi-agent production planning technology.
software standards for the interoperation of Existing software systems that administer the production
heterogeneous and interacting software agents and process within the enterprise (such as enterprise
agent-based systems. resource planning systems (ERP), material
JADE – Multi-agent platform and en environment for management, human resource systems, CAD/CAM
development of multi-agent applications. JADE comply systems) are exploited by their agentification. An agent
with the latest FIPA 2000 specifications. JADE is wrapper that implements the agent-to-agent
implemented in JAVA 1.2 and has no further communication wraps an agentified software system.
dependency on a third-party software. Such a software system becomes an agent – a fully-
fledged member of the multi-agent system.
The main features and the most interesting properties of
using the agent based systems in contrast with the
conventional software systems are scalability,
modularity, online reconfigurability. Agent approach
adopts principles and advantages of distributed
algorithms (parallel computing).
System Architecture Supply Chain Management
The organizational structure of the multi-agent system
mirrors the organizational structure of the given The multi-agent technology can be used for supply
enterprise. The agents have been divided into two chain management and creating virtual organization.
fundamental super-classes: intra-enterprise (IEA) agents Similarly to agentification of the factory legacy systems
and extra-enterprise (EEA) agents. IEAs represent the or hardware machinery, one can integrate ERP systems
distributed planning system while EEAs provide passive/ of collaborators, suppliers and customers. Integrated
active access to the enterprise data. Special instance of supply chain management requires solution for security
EEA is so called Enterprise-to-Enterprise Agent (E2E) and authentication, trust, long term business processes
which connects two enterprises that may want to optimization, ubiquity and openness.
participate in a supply chain.
Agents technology allows planning algorithms to be
distributed throughout the system. The planning
algorithms work in real-time, so the overall plan is
reprocessed immediately when new task appears or in
case of a crucial change in the production. We
recognize two main levels of the task planning. In the
first level the task is decomposed to the parts (sub-tasks)
that can be processed by second planning level. The
first level controls sub-tasks time ordering and
overlapping. The second level is responsible for
resource allocation on specific departments. Each
department is represented by an agent, which
maintains resources for it and makes demanded
schedules. The scheduling algorithm takes into account
priority of the tasks, appropriate planning strategy (e.g.
eager or lazy) and minimizes possibility of tasks lateness.
Dynamic re-scheduling allows system to react to
production feedback (and to every production-
planning-relevant event) in very short time. Both levels
maximal utilization of the
factory departments and
production constrains (e.g.
Visualization of the plans,
available or booked
resources can remarkably
help a user to reveal and
then solve for example
visualization of the results of
the planning process help
to decide what order is to
be accepted in other words what order is the most
suitable in terms of the balanced production load.
Production data can (about workshops, orders, material
etc.) be accessed via the Internet in an ordinary web-
browser. Apart from the portable or the desktop PC, it is
possible to install the extra-enterprise agents on the
mobile PDA devices that are based on the Windows CE
platform. The extra-enterprise agents’ wrapper is also
adapted for communication with the Wireless Application
Protocol (WAP) for running in the cell-phone devices.
Practical Utilization of Multi-Agent
Production Planning System Methodology
Multi-agent production planning
system is operating in the LIAZ
Pattern Shop (Liberec, Czech
Republic) with the aim to improve
control of manufacturing
processes. LIAZ Pattern Shop is a
manufacturing enterprise that is
engaged in production of
patterns, core boxes, permanent
moulds and dies for foundries
and tooling for high volume
production mostly in the
automotive industry. It has also
contributed to better utilization of
the factory resources, setting of
realistic targets and as well as to
customer service improvement.
The technology has been successfully tested also in the
Hatzapoulos SA, a Greek manufacturing enterprise
providing packaging to transnational chains. Multi-agent
technology has been integrated with a classical three-tier
ERP system Atlantis (Unisoft Software, Altec SA, Greece).
Success Stories – Referential Application
Multi-agent production planning system (ExPlantech/
ExtraPlanT) is integrated within legacy system
(heterogeneous ERP) in the LIAZ Pattern Shop. Planning
tool provides an operator with a global long term view on
the production on factory level. An operator can easily
inspect production on particular workshop as well. The
system is also used as decision support tool during the
new orders’ acceptance phase. The system carries out
on-line re-planning re-scheduling so an operator
immediately knows how the production would be
affected by new incoming orders. At the first round the
system works with the estimations on needed resources.
As soon as the production starts the system collects data
about finished jobs and revises the plans and schedules.
The Gerstner Laboratory for Intelligent CertiCon® a.s.
Decision Making and Control® Václavská 12, 120 00 Praha 2, Czech Republic
Czech Technical University, Faculty of Electrical Engineering a 1
Phone: +420 221 460 200, f x: +420 22 460 211
Technická 2, 166 27 Praha 6, Czech Republic E-mail: firstname.lastname@example.org,
4 7 4 7
Phone: +420 22 35 666, fax: +420 22 35 224 URL: http://www.certicon.cz