Multimodal processes prototyping subject to gridlike network and fuzzy operation time constraints
 581 Downloads
 1 Citations
Abstract
This paper describes a Gridlike Material Transportation Network (GMTN) in which several heterogeneous means of transportation (Automated Guided Vehicles (AGVs), hoists, lifts, etc.) interact with each other via common shared workstations to provide a variety of demandresponsive material handling operations. Different material handling transport modes provide movement of workpieces between workstations along their manufacturing routes in the GMTN and they can be seen as processes realized with synergic utilization of various local periodically acting unimodal processes. The main contribution of this research is the solution of a constraint satisfaction problem addressing AGVs fleet matchup scheduling subject to GMTN and fuzzy operation time constraints. In the presented case both production rate (production takt) and operations execution time are described by imprecise (fuzzy) data. In other words, the research’s objective concerns assessing gridlike networks of periodically acting local transportation modes from the perspective of possible masscustomized oriented requirements imposed on scheduling of multimodal flows of jobs assigned to certain technological routes passing through common shared workstations.
Keywords
AGVs fleet scheduling Gridlike network Multimodal process Fuzzy constraints Declarative modelling1 Introduction
The productivity of an AGVserved flow shop repetitively producing a set of different products depends on both the job flow sequencing and the material handling system’s ability to achieve a prespecified throughput. Normally implemented modelling frameworks of this type assume a deterministic character of such systems, where demand is known in advance and the processing times of each job on each machine are known constants. In order to improve systems’ [e.g. Flexible Manufacturing Systems (FMS)] productivity the AGVs fleet size and the scheduling/dispatching influence on the system throughput are usually examined. More formally, such problems belong to a class of deterministic cyclic scheduling problems for which throughput rate targets are established and nonidentical, multistage job routings and machine assignments are fixed. A goal is to sequence and position job operations in a cyclic fashion so as to minimize a weighted sum of the job flow times subject to transportation, blocking and time window constraints. The general problem is known to be NPhard (Levner et al. 2010; Groenemeyer 2012).
In contrast to the above mentioned, the problem considered in this research assumes a given transportation system encompassing a network of AGVs periodically circulating along cyclic routes while servicing workpieces load/unload operations. The considered performance measures such as a production flow takt time, a flow time, a flow rate and a job shop productivity depend on uncertainty requirements caused by availability of workers employed for workstations servicing, e.g. for jig/tool fastening and pick and place operation performance. For the sake of further consideration the following definitions are used: production flow is defined as the movement of material through segregated stages of the workshop, flow rate (throughput) is the number of flow units (e.g. workpieces) going through the workshop (seen as the jobshop) per unit time, flow time (total processing time) is the amount of time a flow unit spends in the job shop from beginning to end, job shop productivity is the output of the workshop in a specific period of time, and takt time (production rate) is the rate at which a finished product needs to be completed.
In that context the investigated question concerns AGVs fleet matchup scheduling subject to productivity constraints with reference to the schedule of the production flow imposed by the uncertainty implied by workers’ availability for operation execution. Consequently it can be seen as an extension of the problems formulated within the framework of a gridlike multimodal transportation network in which an arbitrarily chosen Elementary Transport Networks (ETNs) supported by different transportation modes interact with each other via distinguished subsets of common shared workstations as to provide a variety of transportation/handling services (Bocewicz et al. 2014a, b), and then in (Bocewicz et al. 2015). Moreover, such commonly found regular structures, e.g. gridlike topology networks (Qiu et al. 2002; Balakrishnan and Chun 2007; Perera and Ratnayake 2015), are widely observed in different application domains (such as manufacturing, intercity freight transportation supply chains, multimodal passenger transport), provide additional motivation for this study.
The realistic examples supporting the adopted topological assumptions follow from numerous reports concerning meshlike or gridlike as well as fractallike structures of GMTNs. Such topologies have been considered for over twenty years see Qiu et al. (2002). At the same time fractallike topologies are investigated as well see Silva (2010). Drawing upon the research listed above, our topological assumptions are motivated by the expected advantages gained from the GMTN layout flexibility being vital to improve manufacturing flexibility. This is especially the case when faced with shorter lead times, variable product mix and volumes.
Since, most manufacturing processes observed in their steady state are periodically recurring ones, cyclic scheduling methods can be considered. This kind of representation enables one to show which behaviours (consisting of many cyclic concurrent threads corresponding with routes of AGVs, routes of transported work pieces, etc.) can exist in any given structure (limited by the number of resources and their capacities). Many models and methods aimed at cyclic scheduling of concurrently flowing jobs have been considered to date. Among those, the mathematical programming approach (Von Kampmeyer 2006), maxplus algebra (Polak et al. 2004), Petri nets (Baruwa and Piera 2016), tabu search (Zheng et al. 2014) constraint logic programming (Bocewicz et al. 2014a, b) frameworks are the more frequently used. Most of these methods aim at finding a minimal cycle or maximal throughput while assuming deadlockfree processes flow. Very rarely, however, they concern production systems consisting of an AGV fleet arranged to transport workpieces between workstations (i.e. following a given production route) while moving in a carousel mode along cyclic path topologies. In that context the main question behind the study is: Can the AGV fleet assignment reach its goal subject to the constraints assumed on the concurrent multiproduct manufacturing system?
The remainder of the paper is organized as follows: Sect. 2 provides a brief review of related works. Section 3 introduces a framework aimed at GMTN modeling in terms of System of Concurrently executed Cyclic Processes (SCCP) and provides the problem formulation concerning AGV fleet scheduling with a multiproduct manufacturing flow schedule imposed by fuzzy operation time constraints. After discussion of some issues regarding multimodal processes prototyping, the solution to the considered constraint satisfaction problem is submitted in Sect. 4. An illustrative example providing results of computational experiments is given in Sect. 5. Concluding remarks and further work are presented in Sect. 6.
2 Literature review
Numerous papers have been published to address Vehicle Routing Problems (VRP) (Carić and Gold 2008), AGVs fleet assignment (Abara 1989; Hall et al. 2001), integrated machine scheduling and vehicle routing with time windows (Ullrich 2013) as well as multimodal network flow models (Haghani and Oh 1996). Because of the NPhard nature of these problems, the proposed solution methods range from dynamic programming (Maneengam et al. 2012; Azi et al. 2012), through constraint programming (Kao and Chen 2013; Kilby and Urli 2016) to metaheuristicdriven evolutionary programming (Ullrich 2013; Reed et al. 2014).
Since VRPs assume searching for shortest job flows time subject to transportation, blocking and time window constraints it can be also seen as an extension of the Heterogeneous Fleet Vehicle Routing Problem with Time Windows (HFVRPTW), where workpieces transportation following production routes must occur in certain time windows, i.e. the interval is treated as part of the respective production takts, which varies from production order to production order. The goal of the HFVRPTW is to find transportation times for each AGV stream such that hard time window constraints on both kind of windows hold for a cyclic steady state of AGV flow.
Other perspectives provided by the integration of fleet assignment and maintenance planning (Moudani and MoraCamino 2000; Lu et al. 2013) to AGV fleet scheduling is presented in Bzdyra et al. (2015) where a declarative framework enabling one to determine conditions employed in a decision support system aimed at small and medium size enterprises involved in a unique, multi projectlike and mass customized oriented production is discussed. The unique production orders are grouped into a set of portfolio orders where each production order is treated as an activity network of common shared resources, known in advance, however imprecise operation times are allotted. The job routings may be arbitrary partial orders (i.e., they are not restricted to be serial), and the jobs may be reentrant, meaning that the same machine may be visited multiple times before completion. The problem concerns scheduling a newly inserted project portfolio taking into account both: imprecise operations imposed by a multiproject environment and material handling constraints imposed by AGV availability. A declarative model is sought, enabling one to state a constraint satisfaction problem aimed at: Whether a given portfolio can be completed within the assumed time period in the given manufacturing system?
The material handling system of a flexible manufacturing system where vehicles serve workstations located along different modes of cyclic guided paths in a 3D environment composed of rails/lifts/hoists is considered in (Bocewicz et al. 2016). A production flow scheduling problem subject to fuzzy processing time constraints treated as a fuzzy constraint satisfaction problem is studied. The main objective is to provide sufficient conditions guaranteeing avoidance of timeconsuming computersimulationbased calculations of AGV fleet schedules, while keeping the throughput at the maximum achievable level.
Nielsen et al. (2016) provides some generalization of the problem, stating the resourceconstrained multiproduct scheduling problem for an AGVserved flow shop, where multiple material handling transport modes provide movement of workpieces between machining centers.
Consequently, the problem considered in this research can be seen as an extension and/or refining of the problems formulated in our former works (Bocewicz and Banaszak 2013; Bocewicz et al. 2014a, b, 2015, 2016; Nielsen et al. 2016) focused on a solution to a constraint satisfaction problem aimed at AGVs fleet matchup scheduling while taking into account fuzzy operation times constraints. In order to delimit the work addressed in this research we will first present contributions concerning some of the main issues. The schedulability analysis of the processes executed in the socalled systems of concurrent cyclic processes executed within a declarative modeling framework is provided in Bocewicz and Banaszak (2013). The considered representation provides a unified way for evaluating the performance of local cyclic processes as well as of the multimodal processes supported by these. We propose a recursive design approach to create sufficient conditions that guarantee the schedulability of both local and multimodal cyclic processes. In turn the solution to a constraint satisfaction problem aimed at AGV fleet matchup scheduling while taking into consideration the itineraries of concurrently manufactured product types is provided in Bocewicz et al. (2014b). An algebralike driven approach to cyclic scheduling based on stepbystep composition of multimodal transportation network substructures is proposed.
The considered network of repetitively acting local transportation modes provides a framework for multimodal processes scheduling treated in terms of AGV fleet scheduling problems subject to fuzzy operation time constraints. The aim of the research is to show the universality of the employed declarative modeling driven approach. This approach makes it possible to treat the above mentioned scheduling problem in terms of the following two routine queries formulated either in: a direct way–What schedule of AGV fleet following a given set of operation times maximizes operation availability and productivity?, or a reverse way–What set of operation times guarantee that the resultant schedule of a given AGV fleet maximizes both operation availability and productivity?
3 Gridlike transportation network modelling
Note, that a specific network topology of GMTN, e.g. as shown in Fig. 1a), assumes that it can be considered as composed of multiple Elementary Transport Networks (ETNs). Each ETN can be seen as an autonomous cell (or grid) encompassing cyclically circulating AGVs transporting workpieces among workstations. Workpieces moved along different ETNs while following different production routes can be viewed as multimodal processes passing along locally used transportation paths and workpiece processing operations.
The assumption behind these statements supposes a regular composition of ETNs, encompassing local cyclic steady states, implies a cyclic steady state of whole GMTN, i.e. guaranteeing congestion and deadlockfree flow of AGVs. In a mixed floor and overhead material handling transport system equipped with many unidirectional AGV and overhead hoist transport modes guaranteeing congestion and deadlocksfree operations plays a pivotal role in achieving a wellorganized material flow. Our approach focuses on GMTN steady state behaviour following assumed steady state task requirements, enables deadlockfree scheduling and routing of a great number of multimodal transport means.
3.1 Systems of concurrent cyclic processes modelled in GTN framework
Consider a set of workstations deployed along paths in the transportation network, where fleet of AGVs service workpiece movements between workstations. Due to the organization of material handling operations AGVs circulate along transportation paths, loops in the network composed of multiple AGV loops connected through common shared workstations. In order to guarantee each workstation can be solely serviced by exactly one AGV at the moment, a dispatching rule choosing a transportation mean among currently accessing ones has to be assigned. In that contexts the considered transportation system can be seen as composed of a set of concurrently executed cyclic process synchronized by mutual exclusion protocol, i.e. a System of Concurrently executed Cyclic Processes (SCCP). In turn, SCCP can be directly represented in terms of Petri nets formalism being a powerful formal model for the analysis and modelling of concurrent systems, hence a rich theory has developed around them so far can be also easily used. A way in which the GMTN from Fig. 1a can be modelled as a SCCP is shown in Fig. 1b (Bocewicz et al. 2015). In other words, each SCCP consists of a set of periodically performed processes composed of material handling and processing operations. Processes interact with each other via common used resources in the form of either transportation or processing operations. Dispatching rules allotted to common shared resources provide a mutual exclusion protocol synchronizing processes interaction (Bocewicz and Banaszak 2013).
The considered GMTN consists of ETNs with the structure presented in Fig. 2a. In terms of SCCP representation, it will be called a SCCPdriven model of ETN, (SETN) for short. As the considered GMTN consists of a set of ETNs, therefore to each ath ETN the relevant ath SETN denoted as \(^{\left( a \right) }{} \textit{SC}\) belongs as well, see Fig. 2b.
Formally, local P and multimodal mP processes are defined in the following way:
\(P=\left\{ {P_i i=1,\ldots ,n} \right\} \) the set of local processes described by sequences of operations executed on resources R, where each \(P_i \) is specified by the set of streams: \(P_i =\left\{ {P_i^1 ,P_i^2 ,\ldots ,P_i^k ,\ldots ,P_i^{ls\left( i \right) } } \right\} \),
The operations of multimodal processes’ streams require the execution of some local processes. For example, transport operations between resources in \(mP_1 \) (product \(W_1 \), see Fig. 2) require the two streams of local processes: \(P_1 \) and \(P_4 \). This means the routes of multimodal processes are also determined by the subsequences of routes of the local processes through which they have to be processed. It is assumed that processes are synchronized by a mutual exclusion protocol guaranteeing that only one process can be executed on a shared resource at a given time.
In order to describe the system shown in Fig. 1b and SETN in Fig. 2) let us introduce the following notations (Bocewicz et al. 2014a, b):
\(p_i^k =\left( {p_{i,1}^k ,p_{i,2}^k ,\ldots ,p_{i,j}^k ,\ldots ,p_{i,lr\left( i \right) }^k } \right) \) specifies the route of the stream of a local process \(P_i^k \) (the kth stream of the ith local process \(P_i )\), and its components define the resources used in course of process operations execution, where: \(p_{i,j}^k \in R\), and \(R=\left\{ {R_1 ,R_2 ,\ldots ,R_c ,\ldots ,R_m } \right\} \)—denotes the set of resources used for the jth operation in the kth stream of the ith local process; in the remainder of the paper, thejth operation executed on the resource \(p_{i,j}^k \)in the stream \(P_i^k \) will be denoted by \(o_{i,j}^k \); \(lr\left( i \right) \)—is the length of the cyclic process route. For example, in the SETN seen in Fig. 2b) the routes of streams of processes \(P_1 \), \(P_2 \), \(P_3 \) and \(P_4 \) (e.g. executed by relevant AGVs), are defined by the following sequences: \(p_1^1 =\left( {R_1 ,R_5 ,R_2 ,R_8 ,R_4 ,R_{14} ,R_3 ,R_7 } \right) \), \(p_2^1 =\left( {R_{16} ,R_9 ,R_{17} ,R_{15} ,R_4 ,R_8 ,R_2 ,R_6 } \right) \), \(p_3^1 =\left( {R_3 ,R_{12} ,R_4 ,R_{10} } \right) \), \(p_4^1 =\left( {R_{17} ,R_{11} ,R_4 ,R_{13} } \right) \). \(t_i^k =\left( {t_{i,1}^k ,t_{i,2}^k ,\ldots ,t_{i,j}^k ,\ldots ,t_{i,lr\left( i \right) }^k } \right) \)—specifies theithprocess operation times, where \(t_{i,j}^k \) denotes the time of execution of the jth operation \(o_{i,j}^k \) in the stream \(P_i^k \). In the case of the SETN from Fig. 2b) \(t_{i,j}^k \) means either the transportation time assigned to the sector (i.e. resources \(R_5 R_{15} )\) or the load/unload operation time assigned to the workstation (i.e. resources \(R_1 \), \(R_2 \), \(R_3 \), \(R_4 \), \(R_{16} \), \(R_{17} )\).).
\(x_{i,j}^k \left( l \right) \in {\mathbb {N}}\) the moment of operation \(o_{i,j}^k \) beginning in the lth cycle,
In the considered case of the SETN from Fig. 2b), streams of multimodal processes \(mP_1 \), \(mP_2 \) are specified by the following sequences: \(mp_1^k =( ( {R_1 ,R_5 ,R_2 ,R_8 ,R_4 } ) \cap ( {R_4 ,R_{13} ,R_{17} } ) )=\left( {R_1 ,R_5 ,R_2 ,R_8 ,R_4 ,R_{13} ,R_{17} } \right) \), \(mp_2^k =\left( {R_3 ,R_{12} ,R_4 ,R_8 ,R_2 ,R_6 ,R_{16} } \right) \), being the concatenation of subsequences \(p_1^1 \), \(p_4^1 \) and of \(p_3^1 \), \(p_2^1 \), respectively. In the remainder of the paper, the jth operation of the stream\(mP_i^k \) will be denoted by \(mo_{i,j}^k \).
The operation times \(mt_{i,j}^k \) of multimodal processes are: the operation times of AGV moving between the workstations and operation times assigned to workstations \(R_1 R_4 ,R_{16} ,R_{17} \) (in the model the setup times are included in the operation times). The operation time \(mt_{i,j}^k \) assigned to workstation \(R_g \) is equal to: \(mt_{i,j}^k =x_{\beta ,\gamma }^\delta \left( l \right) x_{\varphi ,\phi }^\psi \left( l \right) \), where: \(x_{\beta ,\gamma }^\delta \left( l \right) \)—is the moment the process \(mP_i^k \) leaves \(R_g \)(taken by local the stream \(P_\beta ^\delta )\) and \(x_{\varphi ,\phi }^\psi \left( l \right) \)—is the moment the process \(mP_i^k \) enters \(R_g \) (delivered by the stream \(P_\varphi ^\psi )\).
\(mx_{i,j}^k \left( l \right) \in {\mathbb {N}}\) moments of operation \(mo_{i,j}^k \) beginning in the lth cycle,
\(\varTheta ^{v}=\left\{ {\sigma _1^v ,\sigma _2^v ,\ldots ,\sigma _c^v ,\ldots ,\sigma _m^v } \right\} \) is the set of the priority dispatching rules of local \((v=0)\) and multimodal processes \((v=1)\), where \(\sigma _c^v \) is the sequence components which determine the order in which streams of processes can be executed on resource \(R_c \). Dispatching rules that determine the order in which vehicles access (streams of local processes) common shared resources \(R_2 \), \(R_8 \), \(R_4 \) are the following: \(\sigma _2^0 =\left( {P_1^1 ,P_2^1 } \right) \), \(\sigma _8^0 =\left( {P_1^1 ,P_2^1 } \right) \), \(\sigma _4^0 =\left( {P_1^1 ,P_4^1 ,P_3^1 ,P_2^1 } \right) \), in turn, dispatching rules synchronizing multimodal processes execution are as follows: \(\sigma _2^1 =\sigma _8^1 =\sigma _4^1 =\left( {mP_2^1 ,mP_1^1 } \right) \).
 \(SL=\left( {P,U,T,\varTheta ^{0}} \right) \)

—characterizes the structure of local processes of SCCP, i.e. P—the set of local processes, U—the set of local process routes, T—the set of local process operations times, \(\varTheta ^{0}\)—the set of dispatching priority rules.
 \(SM=\left( {mP,m\varPi ,mT,\varTheta ^{1}} \right) \)

—characterizes the structure of the multimodal processes of the SCCP, i.e. mP—the set of multimodal processes, M—the set of multimodal process routes, mT—the set of multimodal process operation times, \(\varTheta ^{1}\)—the set of dispatching priority rules.
In simple terms the structure of the SCCP can be treated as a set of pipeline cyclic processes (P, mP) interacting via common shared resources R. Potential process conflicts are resolved in advance by the priority dispatching rules \((\varTheta ^{0}\), \(\varTheta ^{1})\). In that context, the pipelinelike processes typical for a straightline production allows one to take into account both onepiece and variable batch/lot size flows.
In the case of GMTN (e.g. Fig. 1b) the structure \(\textit{SC}\) (1) can be seen as a result of multiple compositions of ETNs (Fig. 2b). Consequently, the structure \(\textit{SC}\) (1) consists of the set of SETNs \(\textit{SC}^{{*}}\) such that:

each resource \(R_c \in Rp_a \) of the substructure \(^{\left( a \right) }{} \textit{SC}\) corresponds to exactly one resource \(R_\omega \in Rp_b \) of the structure \(^{\left( b \right) }{} \textit{SC}\),

each process \(P_i \in Pp_a /mP_i \in mPp_a \) (local as well as multimodal) of the substructure \(^{\left( a \right) }{} \textit{SC}\) corresponds to exactly one process \(P_\mu \in Pp_b /mP_\mu \in mPp_b \) of the structure \(^{\left( b \right) }{} \textit{SC}\),

routes \(p_i \in Up_a /mp_i \in mPp_a \) and \(p_\mu \in Up_b \)/\(mp_\mu \in mPp_b \) of the corresponding processes are sequences consisting of corresponding resources,

each operation \(o_{i,j}^k /mo_{i,j}^k \) executed within the substructure \(^{\left( a \right) }{} \textit{SC}\) corresponds to exactly one operation \(o_{\mu ,\nu }^\varphi /mo_{\mu ,\upsilon }^\varphi \) executed within the substructure \(^{\left( b \right) }{} \textit{SC}\),

dispatching rules \(\sigma _c^v \) of the resources are sequences consisting of elements \(s_{c,d}^v\) indicating the corresponding streams of processes.
3.2 Uncertainty requirements

Operation time \({ }^{\left( a \right) } mt_{i,j} \) (the time the operation \({ }^{\left( a \right) } mo_{i,j}^k \) executed in \({ }^{\left( a \right) } \textit{SC})\) allocated to a workstation (resource) \({ }^{\left( a \right) } R_c \) from \({ }^{\left( a \right) } \textit{SC}\) is associated with the so called operation availability specified by the fuzzy set \({ }^{\left( a \right) } s_{i,j} =\mu \left( {mt_{i,j} } \right) \),\({ }^{\left( a \right) } s_{i,j} \in \left[ {0,1} \right] \). The membership function \(\mu \left( {{ }^{\left( a \right) } mt_{i,j} } \right) \) determines workers’ availability for operations handling on the resource \({ }^{\left( a \right) } R_c \). \({ }^{\left( a \right) } s_{i,j} =0\) indicates that the corresponding value \({ }^{\left( a \right) } mt_{i,j} \) is out of worker’s availability, while \({ }^{\left( a \right) } s_{i,j} =1\)—indicates availability, with acceptable operation time value \({ }^{\left( a \right) } mt_{i,j} \). The accepted production takt \({ }^{\left( a \right) } mTc_i \) of the ith multimodal process has to follow the assumed Multimodal Process driven Awaited Productivity\({ }^{\left( a \right) } sT_i \) (MPAP \({ }^{\left( a \right) } sT_i \hbox {}\) for short) measured along \(mP_i \) in the ath SETN, i.e. in the \({ }^{\left( a \right) } \textit{SC}\).

The takt time of the process \(mP_i \), is the time between two consecutive moments of its subsequent streams completion: \({ }^{\left( a \right) } mTc_i =mx_{i,j}^k \left( l \right) mx_{i,j}^{k1} \left( l \right) \). Similar to the operation availability \({ }^{\left( a \right) } s_{i,j} \), the productivity \({ }^{\left( a \right) } sT_i \) is specified by the fuzzy set \({ }^{\left( a \right) } sT_i =\mu \left( {{ }^{\left( a \right) } mTc_i } \right) \), \({ }^{\left( a \right) } sT_i \in \left[ {0,1} \right] \). The membership function \(\mu \left( {{ }^{\left( a \right) } mTc_i } \right) \) determines productivity following \({ }^{\left( a \right) } mTc_i \) in \({ }^{\left( a \right) } \textit{SC}\). \({ }^{\left( a \right) } sT_i =0\) indicates that production performed with takt \({ }^{\left( a \right) } mTc_i \) is not profitable, while \({ }^{\left( a \right) } sT_i =1\) indicates that takt \({ }^{\left( a \right) } mTc_i \) ensures productivity.
Resource operation availabilities for multimodal processes \({ }^{\left( a \right) } mP_1 \), \({ }^{\left( a \right) } mP_2 \) of SETN from Fig. 2b

Awaited productivity measured along multimodal processes \({ }^{\left( a \right) } mP_1 \), \({ }^{\left( a \right) } mP_2\) in a production takt

3.3 Problem formulation
The majority of problems stated within a SCCP framework can be seen as concerning the relationship between its structure \(\textit{SC}\) and behavior (e.g. the schedule). In that context the main question concerns the SCCP’s cyclic behavior, following requirements imposed by operations availability and awaited productivity subject to a given \(\varTheta \)? Since these kind of problems are NPhard in nature, i.e. computationally intractable, hence the optimal solutions can be obtained only for small scale instances that are quite far from reallife cases (Bocewicz et al. 2014a, b; Levner et al. 2010). However, this disadvantage can be removed in case of a SCCP (modelling a GMTN) which can be seen as composed of a set of SETNs. In order to show this let us note that the considered problem can be decomposed into the following two subproblems. The first one can be reduced to finding the SETN and corresponding set of dispatching rules \(\varTheta \) allocation of which guarantees cyclic behavior of the SCCP (consisting of set of isomorphic SETNs). The second subproblem comes down to the question of whether the assumed uncertainty requirements regarding workers’ availability and job shop productivity can be satisfied for the operation times T, mT specifying the considered SCCP?
The obtained pair: (a set of dispatching rules \(\varTheta \), a set of operation times T, mT) and determined by these, the system’s behaviour can be seen as a distributed control procedure provided to the spatially distributed common shared resources while imposing the right initial process allocations. These kinds of control procedures can then be directly implemented in SCADA (Supervisory Control And Data Acquisition) systems that tie together decentralized facilities such as shared by technological processes AGVs and machine tools (under the given uncertainty requirements).
4 Multimodal processes prototyping
The main idea behind the proposed approach follows from the observation that the assessment of possible cyclic behaviors of GMTN (SCCP) can be replaced the by assessment of its ETN (SETN). Therefore, in order to consider possible deadlockfree/collisionfree executions of processes executed in GMTN, let us focus on constraints of a Constraint Satisfaction Problem dedicated to its ETN.
4.1 Constraints
Since each possible SCCP’s behaviour can be specified by a schedule as a sequence of vectors’ coordinates of which define resources currently allotted to operations over discrete time, hence any obtainable cyclic schedule \(X_{\textit{SC}} \) (3) can be seen as the representation of a cyclic steady state reachable in the state space generated by the SCCP structure (Bocewicz et al. 2012, 2015).
Variables \(x_{i,j}^k /mx_{i,j}^k \in {{\mathbb {Z}}}\) determine moments of operations beginning in the lth cycle of the SCCP cyclic steady state: \(x_{i,j}^k \left( l \right) =x_{i,j}^k +l\cdot \alpha \, /\,mx_{i,j}^k \left( l \right) =mx_{i,j}^k +l\cdot m\alpha \).
Since values of \(x_{i,j}^k/mx_{i,j}^k \) follow the system structure parameters, the cyclic behavior \(X_{\textit{SC}} \) is determined by \(\textit{SC}\) (1). The constraints determining the admissible cyclic schedule \(X_{\textit{SC}} \) (3) are the following:
Constraints (4), (5) and (6) describe conditions guaranteeing cyclic execution of the GMTN’s processes subject to arbitrary given uncertainty requirements. A number of admissible schedules \(X_{\textit{SC}} \) (3), denoted by L, depends on \(\textit{SC}\) (1), and especially on the assumed operations times T of local processes. In that context different values of operation times \(t_{i,j}^k \) (i.e., representing transportation and layover times) lead to schedules with different levels of operations’ availability S and productivity E (according to Tables 1 and 2).
An illustration of this fact can be seen in Fig. 5, where the number L of admissible schedules \(X_{\textit{SC}} \) is shown following from (5) and (6) for the SETN from Fig. 2b), while taking into account uncertainty requirements S and E (4). The results observed are obtained for operation times \(t_{i,j}^k \in \left\{ {1\ldots 6} \right\} \). Space \({\mathbb {L}}\) from Fig. 5 consists of 104,946 (among \(4\times 10^{5})\) cyclic schedules \(X_{\textit{SC}} \) following constraints \(S>0\) and \(E>0\).
4.2 CSPdriven multimodal processes scheduling
Figure 6 presents details specifying each substructure, \({ }^{\left( a \right) } \textit{SC}\), following the SCCP from Fig. 1b. Processes occurring in each substructure, e.g. \(^{\left( a \right) }{} \textit{SC}\),\(^{\left( b \right) }{} \textit{SC}\), \(^{\left( c \right) }{} \textit{SC}\), ...,\(^{\left( f \right) }{} \textit{SC}\) see Fig. 6b, are implemented in the same manner, this means the operations are performed along similar routes, the same dispatching rules are applied, etc.
In the considered system (i.e. composed of one kind of SETN—see Fig. 6) it is enough to know the schedule \({}^{\left( a \right) }X^{{\prime }}\) of substructure \({}^{\left( a \right) }{} \textit{SC}\) to determine the schedule \({X}^{\prime }\). However, to make the composition (8) possible, it is necessary to ensure that the operations executed according to \({}^{\left( a \right) }X^{{\prime }}\), does not lead to deadlocks. Moreover, the schedule \({}^{\left( a \right) }X^{{\prime }}\) is calculated for the “convoluted” structure \({}^{\left( a \right) }{\overline{\textit{SC}}} \) (see example shown in Fig. 6c). The convoluted structure \({}^{\left( a \right) }{\overline{\textit{SC}}}\) is obtained from \({}^{\left( a \right) }{} \textit{SC}\) as a result of merging corresponding resources distinguished in neighbouring structures. In the example shown in Fig. 6c the structure \({}^{\left( a \right) }{\overline{\textit{SC}}} \) has been obtained by way of merging respectively: \({ }^{\left( a \right) } R_1 \) and \({ }^{\left( a \right) } R_{17} \); \({ }^{\left( a \right) } R_3 \) and \({ }^{\left( a \right) } R_{16} \); \({ }^{\left( a \right) } R_5 \) and \({ }^{\left( a \right) } R_{14} \); \({ }^{\left( a \right) } R_6 \) and \({ }^{\left( a \right) } R_{15} \). It should be noted, that local processes executed in the considered resourceconsolidated structures \({}^{\left( a \right) }{} \textit{SC}\) are both collisionfree and deadlockfree. This means the substructure \({}^{\left( a \right) }{} \textit{SC}\) from which the structure \({}^{\left( a \right) }{\overline{\textit{SC}}} \) was born also interact with its neighboring structures (e.g. \(^{\left( a \right) }{} \textit{SC}\),\(^{\left( b \right) }{} \textit{SC}\),\(^{\left( c \right) }{} \textit{SC}\),\(^{\left( e \right) }{} \textit{SC})\) in a deadlockfree way.
Consider the structure \(\textit{SC}\) (2) modeling a given GMTN, composed of ETNs specified by “convoluted” structures \({}^{\left( a \right) }{\overline{\textit{SC}}} \). The periodicity of SETN specified by cyclic, i.e. specified by cyclic schedule \(X_{\textit{SC}} \), behavior of the whole \(\textit{SC}\).
In that context the question corresponding with considered problem is reduced to the following: Which dispatching rules \({ }^{\left( a \right) } \varTheta \) and operation times \({ }^{\left( a \right) } T\), \({ }^{\left( a \right) } mT\), guarantee that in the relevant structure \({}^{\left( a \right) }{\overline{\textit{SC}}}\) there exists the cyclic schedule \({}^{\left( a \right) }X^{{\prime }}\) such that following conditions hold: \({ }^{\left( a \right) } S\ge S_{TH} \) and \({ }^{\left( a \right) } E\ge E_{TH} \)?
Consequently, the positive answer of this question allows one to determine \(\varTheta \, T\) and mT (based on \({ }^{\left( a \right) } \varTheta \),\({ }^{\left( a \right) } T\), \({ }^{\left( a \right) } mT)\) guaranteeing required values of S and E.
\({}^{\left( a \right) }X^{{\prime }},{ }^{\left( a \right) } \varTheta ,{}^{\left( a \right) }T^{\prime },{}^{\left( a \right) }\alpha ^ {{\prime }}\)—decision variables, \(^{\left( a \right) }X^{\prime }\)—the cyclic schedule (8) of substructure \(^{\left( a \right) }{} \textit{SC}\), \({ }^{\left( a \right) } \varTheta \)— dispatching rules of \(^{\left( a \right) }{} \textit{SC}\),\({}^{\left( a \right) }T^{\prime }=\left( {{ }^{\left( a \right) } T,{ }^{\left( a \right) } mT} \right) \)—the sequence of local and multimodal operation times, \({}^{\left( a \right) }\alpha ^{\prime }=\left( {{}^{\left( a \right) }\alpha ,{}^{\left( a \right) }m\alpha } \right) \)—the period of local and multimodal processes for \(^{\left( a \right) }{} \textit{SC}\),
\(D_X ,D_{\Theta } ,D_T ,D_\alpha \) domains determining admissible values of decision variables: \(D_X : mx_{i,j}^k \), \(x_{i,j}^k \in {{\mathbb {Z}}}\); \(D_T :mt_{i,j}^k \), \(t_{i,j}^k \in \left\{ {1\ldots 10} \right\} \),
\(\hbox {C}\)—the set of constraints describing SCCP behavior, the set C includes the constraints describing the execution of local and multimodal processes (4), (5), (6) as well as uncertainty requirements constraints \(S\ge S_{TH} \), \(E\ge E_{TH} \).
The schedule \(^{\left( a \right) }X^{\prime }\) that meets all the constraints from the given set \(\hbox {C}\) is the solution sought for the problem (11). It can be obtained using standard constraint programming driven software platforms such as ILOG, OzMozart, ECL\(^{\mathrm{i}}\)PSE (Sitek and Wikarek 2015).
5 Computational experiments
Consider GMTN from Fig. 1a where four different products \(W_1 W_4 \), are manufactured along distinguished routes followed by multimodal processes \(mP_1 \)–\(mP_4 \). The GMTN modeled in terms of SCCP (see Fig. 1b) reflects itself in the structure \(\textit{SC}\) (1).
Dispatching rules allotted to shared resources in \(^{\left( a \right) }{\overline{\textit{SC}}}\) from Fig. 6c
Resource  Dispatching rules  Resource  Dispatching rules 

\({ }^{\left( a \right) } R_1 /{ }^{\left( a \right) } R_{17} \)  \({ }^{\left( a \right) } \sigma _{1/17} =\left( {{ }^{\left( a \right) } P_4^1 ,{ }^{\left( a \right) } P_1^1 } \right) \)  \({ }^{\left( a \right) } R_4 \)  \({ }^{\left( a \right) } \sigma _4 =\left( {{ }^{\left( a \right) } P_1^1 ,{ }^{\left( a \right) } P_4^1 ,{ }^{\left( a \right) } P_3^1 ,{ }^{\left( a \right) } P_2^1 } \right) \) 
\({ }^{\left( a \right) } R_2 \)  \({ }^{\left( a \right) } \sigma _2 =\left( {{ }^{\left( a \right) } P_2^1 ,{ }^{\left( a \right) } P_1^1 } \right) \)  \({ }^{\left( a \right) } R_8 \)  \({ }^{\left( a \right) } \sigma _8 =\left( {{ }^{\left( a \right) } P_2^1 ,{ }^{\left( a \right) } P_1^1 } \right) \) 
\({ }^{\left( a \right) } R_3 /{ }^{\left( a \right) } R_{16} \)  \({ }^{\left( a \right) } \sigma _{3/16} =\left( {{ }^{\left( a \right) } P_2^1 ,{ }^{\left( a \right) } P_3^1 ,{ }^{\left( a \right) } P_1^1 } \right) \) 
The obtained dispatching rules guarantee cyclic steady state execution of processes completing the convoluted representation \(^{\left( a \right) }{\overline{\textit{SC}}}\).
It should be noted, that obtained solution for operation times, assumed to be unique, is also valid for other values from \({\mathbb {N}}\). Since the state evolution observed in SCCP depends entirely on the occurrence of discrete events over time, the SCCP as well as \(^{\left( a \right) }{\overline{\textit{SC}}} \) can be treated in terms of discrete event systems (DES) (CuninghameGreen and Butkovič 1995; Floudas and Lin 2005). Consequently, assuming an expected productivity measured by production takts \({ }^{\left( a \right) } mTc_1 \) and \({ }^{\left( a \right) } mTc_2 \) of multimodal processes \({ }^{\left( a \right) } mP_1 \), \({ }^{\left( a \right) } mP_2 \) (see Table 2) as well as expected operation times, (see Table 1) the fourth stage of search methodology sketched in Fig. 4, seeks the response to the question below:
Does there exist a plan of production flow (specified by \({ }^{\left( a \right) } T\), \(X_{\textit{SC}} )\), guaranteeing the following requirements hold:
\(S\ge 0.7\) (level of operations’ availability is not less than \(S_{TH} = 0.7\)),
\(E\ge 0.7,\) (level of job shop productivity, i.e. minimal level value among the all multimodal processes are not less than \(E_{TH} =\)0.7)?.
The response to this question depends on the parameters \({}^{\left( a \right) }X^{{\prime }}\)and \({}^{\left( a \right) }T^{\prime }\) specifying \({}^{\left( a \right) }{\overline{\textit{SC}}} \) (from Fig. 6b), i.e., on the solution to the problem \(\textit{PS}_{\textit{REX}_{i}}^{{\prime }} \) (11). This problem has been implemented and then computed in the constraint programming environment OzMozart system (version 1.3.2, Dual Core 2.67, GHz, 2.0, GB RAM). According to Fig. 5 the search process took into account the space \({\mathbb {L}}\) of admissible solutions (schedules) containing over \(4\times 10^{5}\) instances. The final solution was obtained in less than one second. The space \({\mathbb {L}}\) does not contain any admissible production flow plan following the assumed degrees S and E, see Fig. 7.
Therefore, the following request with lower values of \(S_{TH} \) and \(E_{TH} \) has been considered:
In turn, the obtained schedule \({}^{\left( a \right) }X^{{\prime }}\) enables one to determine the final schedule \(X_{\textit{SC}} \) (8)—due to the fifth stage of the methodology shown in Fig. 4. The Gantt chart shown in Fig. 11 provides a graphical illustration of the process \(mP_2 \) (distinguished in Fig. 10) execution guaranteeing workers availability S on the level 0.66 and productivity E on the level 0.6 (production takt 24 u.t.). Summing up the space \({\mathbb {L}}\) contains an admissible production flow plan following assumed degrees S and E, see Fig. 8. The obtained schedule \(X_{\textit{SC}} \), (following \(S=0.66\) and \(E=0.6)\) is shown in Fig. 11.
\(^{\left( {{\varvec{a}}} \right) }{{{\varvec{S}}}{{\varvec{C}}}}\) from Fig. 6b  SCCP from Fig. 1b  

Flow rate (streams/unit time)  0,083(3)  0,16(6)  
job shop productivity \({{\varvec{E}}}\)  0,6  0,6 
\({ }^{\left( {{\varvec{a}}} \right) } {{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{1}}}} \)  \({ }^{\left( {{\varvec{a}}} \right) } {{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{2}}}} \)  \({{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{1}}}} \)  \({{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{2}}}} \)  \({{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{3}}}} \)  \({{{\varvec{m}}}{{\varvec{P}}}}_{{{\varvec{4}}}} \)  

Flow rate (streams/unit time)  0,041(6)  0,041(6)  0,041(6)  0,041(6)  0,041(6)  0,041(6) 
Flow time \({{{\varvec{FT}}}}_{{{\varvec{i}}}} \) (unit time)  22  23  70  70  71  71 
Takt time \({{\varvec{mTc}}}_{{{\varvec{i}}}} \) (unit time)  24  24  24  24  24  24 
Resolution time of the cyclic steady state reachability problem for different scale SCCPs
SCCP  Computation time (s)  

Number of SETN  Number of local processes  Number of resources  
1  4  17  <1 
2  8  32  <1 
3  12  47  11 
4  16  62  60 
5  20  75  180 
6  24  91  600 
7  28  104  1800 
8\(^{*}\)  32  118  5400 
The flow rate equal to 0,083(3) is calculated due to the observation that 2 workpieces can be completed per 24 u.t. within the ETN modeled by \(^{\left( a \right) }{} \textit{SC}\). This value is twice as small as the 0,16(6) observed in the SCCP from Fig. 1b where 4 workpieces can be completed in the 24 u.t. period. In turn, the flow time calculated due to the following formulae \(FT_i =lz\cdot { }^{\left( a \right) } mTc_i {\Delta }T_i \) depends on a number lz of SETNs passed by multimodal processes and an amount of u.t. determining the faster completion of process performance \({\Delta }T_i <{ }^{\left( a \right) } mTc_i \). Therefore, in case of SCCP from Fig. 1b the flow times for \(mP_1 \) and \(mP_2 \) are equal to \(FT_1 =FT_2 =3\cdot 242=70\) and in case of \(mP_3 \) and \(mP_4 \) are equal to \(FT_3 =FT_4 =3\cdot 241=71\).
To summarize let us point out that computational experiments have been conducted for SCCP composed of 8 SETNs, i.e.\(^{\left( a \right) }{} \textit{SC}\), \(^{\left( b \right) }{} \textit{SC}\), ..., \(^{\left( h \right) }{} \textit{SC}\)—see Fig. 10, or in other words consisting 32 local and 4 multimodal processes as well as 118 resources. Due to proposed research methodology the set of priority dispatching rules \(\varTheta \) guaranteeing reachability of SCCP cyclic steady state have been obtained in less than 1 s. In order to evaluate benefits following from this GMTN decomposition driven approach many additional experiments have been conducted. Some results are collected in Table 5. It should be noted that in the case of SCCP from Fig. 10 (number of SETN \(=\) 8) the first solution, i.e. \(\varTheta \), providing cyclic steady state has been obtained after 5400 s, while in the case of SETN from Fig. 6b, the first solution, i.e. \({ }^{\left( a \right) } \varTheta \), has been obtained in time less than 1 s.
6 Conclusions
The proposed declarative approach, aimed at AGV fleet scheduling stated in terms of constraint satisfaction problem representation, provides a unified framework for performance evaluation of local as well as multimodal processes supported by these. The considered problem is reduced to searching for a multimodal processes schedule that guarantee the requirements of workers’ availability S and productivity E hold while taking into account a set of priority dispatching rules synchronizing local processes execution.
The proposed approach leads to solutions based on sufficient conditions that allow the network designer to compose ETNs in such a way as to obtain the final AGV scheduling system with required quantitative and qualitative behavior features. This provides a step towards a method to replace the exhaustive search for the admissible control by an iterative structural feature transportation system design guaranteeing its required behavior.
In our model setup times are included in the processing times, i.e. similar as in most of classical machine scheduling research where setup times are considered either as negligible or as parts of the processing times of jobs. Of course, since interest about scheduling problems with setup times has grown significantly during the last four decades, see: (Yang and Liao 1999), this issue will be taken into account in our further work as well.
Future research will be aimed at grid communication networks, while devoted to scalability extensions of the proposed declarativemodeldriven approach. Since the regularity of grid transportation networks provide a modelling framework enabling significant simplification of the evaluation of the overall network performance the ordered fuzzy numbers concept (Kosiński et al. 2013) and hybrid techniques (Sitek and Wikarek 2015) joining MLP and CP may dramatically decrease computation time requirements. Coupling both the gridlike structure based methodology of transportation networks and a new problem modelling concepts may result in a new generation of DSS providing online solutions to reallifesized communication transportation problems.
References
 Abara, J. (1989). Applying integer linear programming to the fleet assignment problem. Interfaces, 19, 4–20.CrossRefGoogle Scholar
 Azi, N., Gendreau, M., & Potvin, J.Y. (2012). A dynamic vehicle routing problem with multiple delivery routes. Annals of Operations Research, 199(1), 103–112.CrossRefGoogle Scholar
 Balakrishnan, J., & Chun, Hung C. (2007). Multiperiod planning and uncertainty issues in cellular manufacturing: A review and future directions. The European Journal of Operational Research, 177, 281–309.CrossRefGoogle Scholar
 Baruwa, O. T., & Piera, M. A. (2016). A coloured Petri netbased hybrid heuristic search approach to simultaneous scheduling of machines and automated guided vehicles. International Journal of Production Research, 54(16), 4773–4792.CrossRefGoogle Scholar
 Bocewicz, G., & Banaszak, Z. (2013). Declarative approach to cyclic steady state space refinement: periodic process scheduling. The International Journal of Advanced Manufacturing Technology, 67(1–4), 137–155.CrossRefGoogle Scholar
 Bocewicz, G., Banaszak, Z., & Nielsen, I. (2015). Multimodal processesprototyping subject to fuzzy operation time constraints. In 15th IFAC Symposium on Information Control Problems in Manufacturing—INCOM2015, 48(3), 2103–2108.Google Scholar
 Bocewicz, G., Nielsen, P., Banaszak, Z., & Dang, V. Q. (2012). Cyclic steady state refinement: Multimodal processes perspective. Advances in Production Management Systems, Series: IFIP Advances in Information and Communication, 384, 18–27.Google Scholar
 Bocewicz, G., Nielsen, I., & Banaszak, Z. (2014a). Automated guided vehicles fleet matchup scheduling with production flow constraints. Engineering Applications of Artificial Intelligence, 30, 49–62.CrossRefGoogle Scholar
 Bocewicz, G., Nielsen, I., & Banaszak, Z. (2014b). Iterative multimodal processes scheduling. Annual Reviews in Control, 38(1), 113–122.CrossRefGoogle Scholar
 Bocewicz, G., Nielsen, I., & Banaszak, Z. (2016). Production flows scheduling subject to fuzzy processing time constraints. International Journal of Computer Integrated Manufacturing, published online. doi: 10.1080/0951192X.2016.1145739.
 Bocewicz, G., Muszyński, W., & Banaszak, Z. (2015b). Models of multimodal networks and transport processes. Bulletin of the Polish Academy of Sciences, Technical Sciences, 63(3), 635–650.CrossRefGoogle Scholar
 Bzdyra, K., Bocewicz, G., & Banaszak, Z. (2015). Mass customized projects portfolio scheduling–imprecise operations time approach. Theory and Practice of Industrial and Production Engineering, Series: Applied Mechanics and Materials, 791, 70–80.Google Scholar
 Carić, T., & Gold, H. (2008). Vehicle routing problem. Vienna: ITech.Google Scholar
 CuninghameGreen, R. A., & Butkovič, P. (1995). Discreteevent dynamic systems: The strictly convex case. Annals of Operations Research, 57(1), 45–63.CrossRefGoogle Scholar
 Floudas, C. A., & Lin, X. (2005). Mixed integer linear programming in process scheduling: modeling, algorithms and applications. Annals of Operations Research, 139(1), 131–162.CrossRefGoogle Scholar
 Groenemeyer, S. (July 2012). Novel approaches to cyclic jobshop problems with transportation. Ph.D. thesis University of Nottingham, Nottingham, U.K., 237.Google Scholar
 Hall, N. G., Sriskandarajah, Ch., & Ganesharajah, T. (2001). Operational decisions in AGVserved flowshop loops: Fleet sizing and decomposition. Annals of Operations Research, 107, 189–209.CrossRefGoogle Scholar
 Haghani, A., & Oh, S. C. (1996). Formulation and solution of a multi commodity, multimodal network flow model for disaster relief operations. Transportation Research Part A: Policy and Practice, 30, 231–250.Google Scholar
 Kao, Y., & Chen, M. (2013). Solving the CVRP problem using a hybrid PSO approach. In K. Madani, A. Dourado, A. Rosa, & J. Filipe (Eds.), Computational Intelligence (pp. 59–68). New York: Springer.CrossRefGoogle Scholar
 Kilby, P., & Urli, T. (2016). Fleet design optimisation from historical data using constraint programming and large neighbourhood search. Constraints, 21(1), 2–21.CrossRefGoogle Scholar
 Kosiński, W., Prokopowicz, P., & Rosa, A. (2013). Defuzzification functionals of ordered fuzzy numbers. IEEE Transactions on Fuzzy Systems, 21(6), 1163–1169.CrossRefGoogle Scholar
 Lu, S. P., Kong, X. T. R., Luo, H., Wong, W. R. C, & Luo, K. B. (2013). Dynamic scheduling of AGVS for tobacco automatic warehouse: A case study. In: Proceedings of international conference on computers and industrial engineering CIE 2013 (pp. 725–733).Google Scholar
 Levner, E., Kats, V., Alcaide, D., Pablo, L., & Cheng, T. C. E. (2010). Complexity of cyclic scheduling problems: A stateoftheart survey. Computers & Industrial Engineering, 59(2), 352–361.CrossRefGoogle Scholar
 Maneengam, A., Laotaweesub, W., Udomsakdigool, A., & Sripathomswat, K. (2012). Applying dynamic programming for solving the multimodal transport problem: A case study of Thai multimodal transport operator. In Proceedings of the Asia Pacific industrial engineering and Management Systems Conference, Bangkok, Thailand (pp. 1319–1325).Google Scholar
 Moudani, W., & MoraCamino, F. (2000). A dynamic approach for aircraft assignment and maintenance scheduling by airlines. Journal of Air Transport Management, 6, 233–237.CrossRefGoogle Scholar
 Nielsen, I., Wójcik, R., Bocewicz, G., & Banaszak, Z. (2016). Multimodal processes optimization subject to fuzzy operation time constraints: Declarative modeling approach. Frontiers of Information Technology & Electronic Engineering, 17(4), 338–347.CrossRefGoogle Scholar
 Perera, G., & Ratnayake, V. (2015). Qualitative analysis on layout flexibility of machine layout configurations. Journal of Business and Management, 17(2.Ver. III), 2319–7668.Google Scholar
 Polak, M., Majdzik, P., Banaszak, Z., & Wójcik, R. (2004). The performance evaluation tool for automated prototyping of concurrent cyclic processes. Fundamenta Informaticae, 60(1–4), 269–289.Google Scholar
 Qiu, L., Hsu, W.J., Huang, S.Y., & Han, W. (2002). Scheduling and routing algorithms for AGVs: A survey. International Journal of Production Research, 40(3), 745–760.CrossRefGoogle Scholar
 Reed, M., Yiannakou, A., & Evering, R. (2014). An ant colony algorithm for the multicompartment vehicle routing problem. Applied Soft Computing, 15(2014), 169–176.CrossRefGoogle Scholar
 Silva, A. L. (2010) Critical analysis of layout concepts: Functional layout, cell layout, product layout, modular layout, fractal layout, small factory layout. In Proceedings of international conference challenges and maturity of production engineering: Competitiveness of enterprises, working conditions, environment. Brazil (pp. 2–13).Google Scholar
 Sitek, P., & Wikarek, J. (2015). A hybrid framework for the modelling and optimisation of decision problems in sustainable supply chain management. International Journal of Production Research, 53(21), 1–18.CrossRefGoogle Scholar
 Ullrich, Ch A. (2013). Integrated machine scheduling and vehicle routing with time windows. European Journal of Operational Research, 227, 152–165.CrossRefGoogle Scholar
 Von Kampmeyer, T. (2006). Cyclic scheduling problems. Ph.D. dissertation, Mathematik/Informatik, Universität Osnabrück.Google Scholar
 Yang, W. H., & Liao, Ch J. (1999). Survey of scheduling research involving set up times. International Journal of Systems Science, 30(1999), 143–155.CrossRefGoogle Scholar
 Zheng, Y., Xiao, Y., & Seo, Y. (2014). A tabu search algorithm for simultaneous machine/AGV scheduling problem. International Journal of Production Research, 52(19), 5748–5763.CrossRefGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.