A stigmergic approach for dynamic routing of active products in FMS
Introduction
This paper is relevant to the context of distributed dynamic control of production processes in flexible manufacturing systems (FMSs). In this context, a stigmergic routing control model, capable of automatically finding efficient routing paths for active products in flexible manufacturing systems undergoing perturbations, is proposed.
Most of the research in the FMS domain focuses on the distributed control of Dynamic Allocation Processes (DAP) (e.g., dynamic scheduling). Little attention has been paid to the distributed control of Dynamic Routing Processes for products (DRP), (e.g., transportation times are assumed constant, paths unique and conveying capacity unlimited) [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]. In addition, products are usually considered to be passive entities: they never communicate, decide or act during the routing process. However, recent technological advances (e.g., RFID, smart cards, embedded systems) have led to research on “active products”, in which products are able to act based on the real state of the system (e.g., resources, production and transportation systems). These “active” capabilities may be embedded in the product itself or may operate from a distance [11].
Given these realities, we have begun to work on the distributed control of DRP using the notions of “active product” and stigmergy to provide efficient, real-time, adaptive product routing. These notions make our model more realistic, able to take such aspects as limited system capacity and system reliability into account.
This paper is structured as follows: Section 2 gives a short presentation of the state-of-the-art in the domain of stigmergy applied to manufacturing control. Section 3 presents the assumptions on which our model is based, as well as the notations and variables that are used in subsequent sections. Section 4 describes the model, and Sections 5 Simulation, 6 Scenarios and results, respectively, introduce the developed simulator and report the results obtained. Section 7 provides a short synthesis and Section 8 describes the ongoing implementation. Section 9 offers our conclusions and perspectives for future research.
Section snippets
Stigmergy: concepts and the state-of-the-art in manufacturing control
FMS routing is a difficult problem because its nature is stochastic and time-variable. Our objective in this study was to build an efficient routing system that is capable of finding the best routing solutions in real-time and of adapting to new traffic situations and changes in the conveying network's connectivity (e.g., jamming, failures or slowdowns on the network arc, topological modifications). Some insect societies that use stigmergy – for example, ant colonies – exhibit these desirable
Assumptions
Our proposed architecture is based upon four assumptions concerning four different system aspects:
- (1)
The topology of the transportation system is assumed to be associated to a strongly connected, directed graph, in which nodes can be both resources and disjunction points, and arcs are the parts of the system that require no decisions during the routing process since the product can only move in one direction towards the next node. Routing times are assumed to be non-negligible compared to
Notations and variables
Let N = {ni} be the set of considered nodes ni (production resources or transportation system disjunction points in both PL and VL). The variable denotes the wth neighbor of ni and , the set of the neighbors of ni: , . N and thus describe then the topology of the FMS, including the transportation system. A possible path between node ni and node nj is written uij = [ni … nj] and corresponds to the ranked list of successive nodes to be visited to
Short description of the simulation tool
The proposed control model is naturally distributed, meaning there is no central memorization and processing system. This property influenced our choice of an agent-based parallel modelling and simulation environment. With Netlogo [22], each modelled entity can be described as an independent agent interacting with its environment. All agents operate in parallel on a grid of patches (i.e., a cellular world), and each agent can read and modify some of the attributes linked to the patches in its
Scenarios and results
Two scenarios are presented, highlighting our model's self-adaptative capacities with respect to perturbations. The second scenario is also used to analyze the optimality of the pheromone-based paths. Using an INTEL Xeon with 3 Gigabytes of Ram, the time needed for the different simulations ranged from 1 to 5 s.
Synthesis
The results described above are consistent with those found by Di Caro using the AntNet algorithm. Di Caro and Dorigo [15] showed that AntNet out-performed most routing algorithms (e.g., Bellman-Ford, OSPF) for heavy perturbations. In our FMS context, the PAPs are able to determine the best path from the departure node to the destination node without any centralized control. In addition, they are able to overcome perturbations by seeking out new paths that bypass the perturbation but still lead
Elements for a real implementation
Given the promising results of the simulation, we decided to test our model on a real implementation. The experimental support for this implementation is the flexible assembly cell of the AIP-PRIMECA Center at the University of Valenciennes (Fig. 13). Its topology corresponds to the upper area of the FMS presented in Fig. 8.
Three types of equipment were used for the implementation:
- -
Conveyer system, which is a Montrac monorail transport system (Montech, 2008 [23]) using self-propelled electrical
Conclusion and prospects
This paper describes an original solution to dynamic routing in FMS, in which products play an active role. The proposed solution uses the concept of stigmergy commonly found in nature. The results of our simulation highlight the robustness and adaptability of our approach. In addition, the first steps in a real implementation have already been validated.
Throughout the paper, several perspectives have been identified. One of the main short-term perspectives is to study different fine-tuning
Yves Sallez is currently assistant professor in ENSIAME Engineering School of the University of Valenciennes and Hainaut-Cambrésis, where he teaches robotics, vision and automated production. He received his PhD in automatics from the Valenciennes University in 1988. He is also member of the LAMIH-CNRS (Automatic Control, Computer Science and Mechanical lab) at the University of Valenciennes, France. His research interests are in the area of self-organized control of manufacturing or transport
References (25)
- et al.
An experimental benchmarking of two multi-agent architectures for production scheduling and control
Computers in Industry
(2000) - et al.
A soft computing approach for task contracting in multi-agent manufacturing control
Computers in Industry
(2003) - et al.
Product-based and resource-based heterarchical approaches for dynamic FMS scheduling
Computer and Industrial Engineering
(2004) A survey of factory control algorithms that can be implemented in a multi-agent heterarchy: Dispatching, scheduling and pull
Journal of Manufacturing Systems
(1998)- et al.
Dynamic shopfloor scheduling in multi-agent manufacturing systems
Expert Systems with Applications
(2006) - et al.
Probabilistic behaviour in ants: a strategy of errors?
Journal of Theoretical Biology
(1983) - et al.
Use of machine learning for continuous improvement of the real time manufacturing control system performances
International Journal of Industrial and Systems Engineering
(2008) - et al.
Agile scheduling of flexible manufacturing systems of production
- et al.
Engineering framework for agent-based manufacturing control
Engineering Applications of Artificial Intelligence
(2006) Survey of resource allocation methods for distributed manufacturing systems
Production, Planning & Control
(2001)
Manufacturing enterprise control and management system engineering: rationales and open issues
IFAC Annual Reviews in Control
The intelligent product in manufacturing control and management
Cited by (69)
Towards adaptive digital twins architecture
2023, Computers in IndustryTowards the deployment of customer orientation: A case study in third-party logistics
2019, Computers in IndustryClassification of cyber-physical production systems applications: Proposition of an analysis framework
2019, Computers in IndustryCitation Excerpt :The machines are for example able to negotiate and establish their own planning in order to decide whether they accept the contract to be established by an order [59]; Products: many works in the last decade were focused on the activity of the product all along the production phase [60–62]. Meanwhile, several parallel directions lead to the concept of “Intelligent Product” [63–67], where the product was the trigger and a true actor of the all the decisions that were dynamically taken during the production thanks to storing and processing capabilities that are physically implemented on the product himself.
Control architecture and design method of reconfigurable manufacturing systems
2016, Control Engineering PracticeCitation Excerpt :Netlogo software has been used to simulate MAS-based models for manufacturing domain. In Sallez, Berger, and Trentesaux (2009), it is used for determining routing products in situations characterized by the occurrence of disturbances. Another manufacturing application is presented in Barbosa, Leitão, Trentesaux, and Adam (2011), in which Netlogo is compared with other similar tools presenting benefits, such as easiness to be programmed.
Yves Sallez is currently assistant professor in ENSIAME Engineering School of the University of Valenciennes and Hainaut-Cambrésis, where he teaches robotics, vision and automated production. He received his PhD in automatics from the Valenciennes University in 1988. He is also member of the LAMIH-CNRS (Automatic Control, Computer Science and Mechanical lab) at the University of Valenciennes, France. His research interests are in the area of self-organized control of manufacturing or transport systems. He is a coordinator of the SCP project “System Controlled by the Product” in the French research CNRS group MACS.
Thierry Berger is currently assistant professor in Automation and in Industrial Computing in the ENSIAME Engineering School at the University of Valenciennes in France. He is also member of the LAMIH-CNRS (Automatic Control, Computer Science and Mechanical lab). He received his PhD degree in automatics from the University of Valenciennes in 1991. The PhD subject dealt with Man Machine System and real-time workload assessment of the Human Operator. Currently, his investigation is related to Flexible Manufacturing System, Self-organized System, Active Product, Transport System, Man–Machine Interactions in Awareness Context and Ambient Intelligence. He is a member of the French research CNRS group MACS and a contributor in the SCP project “System Controlled by the Product”.
Damien Trentesaux received the engineering and PhD degrees respectively in electrical engineering in 1992 and in heterarchical manufacturing control in 1996 from the Institut National Polytechnique of Grenoble (INPG), France. He is currently professor in the LAMIH-CNRS (Automatic Control, Computer Science and Mechanical lab) at the University of Valenciennes and Hainaut-Cambrésis, France. His research areas of interest include multicriteria decision making and heterarchical control of discrete event systems (manufacturing, transport, logistics, and services). He teaches Production Management, Linear Control and Simulation at the ENSIAME Engineering School of the University of Valenciennes and Hainaut-Cambrésis. He is a member of the French research CNRS group MACS and is author of several publications in the manufacturing domain.