Automated Guided Vehicle Systems Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Offering versatilities by automated equipments in modern manufacturing systems has brought easiness in laboring along with complexity and complicacy in control and management of manufacturing system. So familiarization to various related issues in setting and implying automated systems seems necessary. This study surveys most involving momentous modules; both software and hardware structures in design, planning, management and control level for automated guided vehicles as popular handling equipments in manufacturing floor. Current paper purposes key definitions related to automated vehicles system. Some of expressions are presented in tables to simplify categorization of interpretations and applications. In addition to explanations, important models, methods and protocols of scheduling, routing, navigating, dispatching and positioning in literature along long term view are discussed distinctly in specified sections which cover more aspects of alluded studies, while most of these researches could be placed in different parts. Eventually we evaluate categories of our paper to determine main attentions in recent studies.


Automated Guided Vehicles (AGV); Requirement and Design; Control and Management


The manufacturing processes cause obligatory moves and flows of raw materials as well as work-in-processes materials and finished goods between different depended departments of an organization. These movements need equipments and absolutely management through them which called material handling system. According to Tompkins [1] who provided great essence in definitions, scopes, principles and cost discussions in material handling, there are three different views to material handlings as follow:

Conventional; handling between two locations within the same manufacturing system,

Contemporary; integrated handling plan in a factory or warehouse,

Progressive; totally material handling of all activities from suppliers to customers

This paper advocates the second category. Since material handling is a significant part of total production cost, therefore designing an efficient system of transportation in a plant can lead to considerable diminishing in manufacturing expenditures. Hence it is important to know main features involving in such systems.

In order to refine a handling system that is emphasized in [1] the first focus should be on the materials, second on the movements and third on the methods. Therefore the most considerable themes in material handling are resources to handle material. These resources that are responsible for duties of movements, storages and retrieving goods during processes could be humans or machines.

Previously the creativity was helpful for simplifying transportation system even using gravity as natural force for movements. Recently some renovation activities in manufacturing structure like scheduling, arrangements, management information, utilizing automatically had compiled great changes in logistic systems. These changes had affected on material handling system and brought new generation of automatic handling that are adaptable with automatic processing, complicated intelligent management and control systems. Majorly could be said that the roles of humans has became lighter and inversely the machines' duties have been highlighted. Therefore material handling equipments are weighted possessions to handle rather than in the past which could be mostly categorized into four main classes:


Material transporter,


Industrial vehicles,

Monorail, hoists and cranes

Storage and retrieval trucks and racks

Bulk material handling,

Automated identification and communication equipments

Here, we deal with related publications about industrial vehicles in material transporters category that could contain Automated Guided Vehicles (AGV), Automated Storage and Retrieval Machines (ASRM) and Automated Electrified Monorail (AEM). Industrial vehicles dependably provide appropriate material handling at the production and distribution systems with the goal of yielding long term mutual rewards. Among these vehicles, AGVs provide reliable horizontal transportation when space is at a premium and flexibility is critical for improving productivity, speed and accuracy. This is the reason for their special place in different industrial areas and numerous theoretical and practical researches in this field. In other words, the characteristics of these studies on this special handling equipments are in order to applying different supply chain experiences to help companies achieve their corporate objectives with the lowest possible risk, highest possible Return On Investment (ROI), improving shareholder value and qualified customer service .

For the sake of simplicity and concentration we limit our paper to Automated Guided Vehicle System (AGVS) among varied automated equipments in handling systems. An AGVS consists of several AGVs operating concurrently and management system to control them. For many years, discussions about AGVS have been published by many researchers in this field. AGVSs revealed for the first time in 1954 in the USA [2], but emergence of Integrated Circuit (IC) technology and microprocessors changed computer control of AGVSs in the late 70s by wiping out a great deal of hardware. The evolution of AGVs has been compared to that of robots. So far it has been justified, judging from the number of worldwide installations, but future outwardly belongs to a more advanced type of AGVSs.

There are few review papers on AGV systems. However, they ponder on only limited parts of AGVS related problems, like [3][4] that are not up to date. The current paper attempts to fill the gap by broaden overview of existing literature, including the most recent contributions in design and control of AGVS.

There are many factors could be included in an AGVS like adaptability, flexibility, load capacity, space, speed, path and etc. that all are grouped in design and management classes. Management is needed to support some plans in designed system, the major ones are: vehicle scheduling, vehicle routing, vehicle positioning and deadlock prevention. These issues are belonged to different levels of the decision making processes, like strategic level (guide path design), technical level (scheduling) and operational level (vehicle routing). Whereas each subject stands alone, there are some relations between different issues as we will discuss in this paper. (See Fig.1)


The remaining sections of this paper are as follow; in the next section we view the related researches in AGVS design like needed characteristics, guide paths and navigations' strategies. Subsequently the required elements to control and management which contains routing and scheduling, dispatching plans divided into real time and preplanned policies, location problems and decision makings encounter to deadlocks and collisions are denoted in section there. Finally we summarize addressed discussions in this paper to show our objectives for this study.

Design of AGVS

Design phase of AGVS consists of vehicles' capacity sizes and numbers (As characteristics of AGVs), path figure, intentions of navigation and some other included parameters depended experiments about processes which we move forwards with them in this section.Together with knowing requirements and setting a system, in order to measure the performance of the designed system, various criteria can be used. According to the literature some of the objectives of a designed AGV system are:

Maximize throughput of the system (i.e. number of loads handled per time unit),

Minimize time required to complete all jobs (i.e. make span),

Minimize vehicle travel times (empty or/and loaded),

Evenly distribute workload over AGVs,

Minimize total costs of movement,

Minimize time job is handled after its due time (i.e. tardiness),

Minimize maximum or average throughput times of AGVs to travel to the destination of new jobs,

Minimize expected waiting times of loads.

Many decision variables arise in design problems. The impact of decisions on mutual interactions and performance might be difficult to predict. It might be hard to decide on one thing without considering other decision variables. To simultaneously address some of these design problems, simulation models (See [8]) or combinations of simulation and analytical models [10] might be used for design and control of AGVS. Van der Heijden et al. [12] built a simulation model for an automated underground transportation system around an airport which proposed terminals, needed AGVs, routing, scheduling and traffic situations. Aiello et al. [13] studied the problem of determining the location of departments in a facility and sequentially designing the material handling system. The authors formulate a genetic algorithm which determines the location of departments, location of pickup and delivery points and direction of guide paths, such that material handling costs are minimized. It has been shown that this algorithm outperforms the classical approach of first determining location of departments and thereafter designing the material handling system.

Concluding, a few authors simultaneously addressed various design problems of AGV systems by using simulation. Some analytical models have been developed which can be applied for AGV systems with a relatively small number of AGVs and workstations. However, layout problems and control problems, which are highly interrelated, are often separated. By doing this, attention should still be paid to the overall performance of the system. A well developed layout with a bad control rule might result in a decrease of the system performance. Whereas performance of operations has became more and more important, especially from an economic point of view. This mainly applies to non-value added activities such as transportation and transhipment. As a result, the performance of an AGV system operating in distribution, transhipment and transportation areas should be high.

Therefore, more attempts should be made to simultaneously address multiple design problems either by applying analytical methods or simulation. Moreover, the mutual relationship with the design of other material handling systems, such as storage systems, should be considered. The direction of this section is in this manner to described main related subjects separately in designing. So we illustrate some needed attributes, guide path and navigation system distinctively in following sub-sections.

AGVs' Characteristics

There are various problems identifying with traits of AGVs that we point substantial features among them. The attributes of AGVs influences on the scheduling and travel time of routing. There are some studies in literature focused on these parameters. Johnson [14] studied the impact of empty vehicle traffic on flow path design and vehicle requirements. The author shows that the usage of empty trip information significantly improves the performance of the systems design. Koo and Jang [15] presented stochastic travel time models for AGVs to determine loaded and empty vehicle travel times. First time number of needed AGVs has discussed by Maxwell [16] and after that has followed by Maxwell and Muckstadt [17]. Number of AGVs is affected by the volume of loads in the system and the kind of load hauling capability of AGVs. So the type of used AGVs in carrying is effective in system design and control management too.

Assumptions of most papers are under only one type vehicle. Benefits of this supposition enable us to divide network to smaller areas that have same AGVs' conditions. Bilge and Tanchoco [18] surveyed most important keys identifying with multi load AGVs. The authors demonstrated the benefits of using this kind of AGVs in a transportation system and by using a simulation approach exhibited that these vehicles decrease the sensitivity of system related to guide path in bottlenecks. Therewith multi loads AGVs are very adaptable vehicles, especially when breakdowns happened and there are different loads to transfer. Totally there are two approaches in routing of multi load AGVs in a system. 1) Fix path; pick up loads which are mentioned in the specific plan toward a close destination. 2) Variable Path routing that AGV would change current path in order to pick up some new requests. Variable path is more flexible but needs some modifications. Table1 shows main advantages and disadvantages of unit load and multi load AGVs.

Table 1

Could be cited that problems which are connected to unit load AGVs are like Multiple Travelling Salesman Problems (M-TSP) and which ones related to multi load AGVs problems are like Dial-A-Ride Problems (DARP).

Guide Path

The guide path type usually is set up based on the experiments of manufacturing attributes. An expert system can be useful to support the guide path system selection process. After specifying an appropriate type of guide path system, a mathematical model could be used for a proper scheduling and routing plan by designer. The control system can be used in different path topologies; either bidirectional or unidirectional, dual track or single track flow paths. We subsume summary of guide paths in Table2 that mentions these different path systems in detail.

Table 2

As noted in Table2 there are there classes in topology of layout that number of lanes and direction would be mentioned for them. Fig.2 displays these configurations in layout obviously.

Fig. 2

As Le-Anh and Koster [19] described; guide path design is an important member in AGVS family. It is one of the very first problems to be considered. Most published works on the guide path design problems assume that facility layout and locations of pickup/delivery (P/D) stations are given and fixed. The main problem is to decide the connections or guide path segments to be included in the solution. In some cases, the number of parallel lanes of a connection is to be decided as well.

Wallace [20] used artificial intelligence to build up an AGV controller for large complex guide paths. Wijesoma et al. [21] suggested another intelligent control system for outdoor AGVs. In their study fuzzy logic control methods are used to control an automated golf car. Shirazi et al. [22] provided a zero-one multiple objective nonlinear programming problems in Flexible Cellular Manufacturing System (FCMS) under six sigma quality control compliment to determine suitable machines for Tandem Automated Guided Vehicles (TAGV). Beside the attribution of real time, the paper by Asef-Vaziri and Goetschalckx [23] is a good reason for using dual track guide path. They compared a dual track and a segmented single track in the same topology and decided that a dual track loop is more economical in additional space rather than single track.

The Navigation of AGVs

The mean of navigation is rooted from the way to control and guide vehicle movements; could be manually driven as well as automatically. There are different trend in autonomous navigation problems. Soft computing techniques such as evolutionary algorithms, fuzzy logic and neural network are developed. One of the evolutionary algorithms applied to the navigation problem is developed by Lin et al. [24]. They supposed network optimization heuristic in order to overcoming AGVs' capacity constraint, constraints among sequences of processes and deadlock situations. Then Genetic Algorithm (GA) was applied to minimize makespan and number of AGVs. Suitable routes were constructed between locations. While it might no information on a factory was given or a navigation route is predetermined. Evolutionary approaches are very useful for modifying routes in navigation problems. If AGVs straggled from the supposed route by some accident or lags, it would become troublesome for AGVs to move in the factory, because of some obstacles in the factory, AGVs must avoid them and correct the route again.

The need to cope with uncertainty and imprecision resulted from natural environment is the reason for using fuzzy logic as evident in the realm of autonomous navigation for robots. Saffiotti [25] surveyed some literature to specify design, coordinates, using data and integration in excitation level in fuzzy skeleton. Pradhan et al. [26] developed different fuzzy membership function in Fuzzy Logic Controller (FLC) to navigate multiple mobile robots. They presumed ultrasonic and infrared sensors for measuring the distances and detecting the target. A conceptual Petri Net Model (PNM) was implied to recognize robots near each other. Finally Gaussian membership introduced as the best membership function of FLC in navigation of multiple mobile robots. It is worthwhile to note when the uncertainty inherited from stochastic process, instead of fuzzy logic and large amount of its tools, the probability distributions should be used. Pratihar et al. [27] utilized a hybrid of a GA and fuzzy approach in this manner that genetic algorithm plays a tuner role for scaling factors along with the FLC to find near optimal plan in navigating robots. They demonstrated the merit of their rule base algorithm compared with other knowledge base FLC approaches.

In the neural network frame Pomerleau [28] discussed on different feature of Autonomous Land Vehicle in a Neural Network (ALVINN) with using back dissemination of training algorithm, which could be accounted as an integrated reference in this area.

Although these soft computing methods are used to apply for general robots but the methods applied for improvement in robots' motions could be applicable for AGVs too.

Control and Managment of AGVS

The construction of a good control system is a principle progress in which materials handling practice can affect saving time and money particularly in continuous manufacturing, where all operations are integrated by management plans. It provides safety working condition and fewer rejects along with reduce the storage of materials and give higher productivity.

One of the main purposes of a control procedure is to satisfy demands for transportation by maximizing the speed and minimizing occurrence of conflicts among AGVs. Hence, at least the following activities need to be performed by a system controller:

• Dispatching of loads to AGVs,

• Route selection,

• Scheduling of AGVs,

• Dispatching of AGVs to parking locations.

Generally control system of AGVS is divided into two strategies; offline (Pre-planned) and online (Real time) control policy. To use offline control for AGVS, knowing all transportation demands characteristics like, transportation origin, destination, release time and transportation time is necessary. Offline control system needs transportation requests to be rightly predictable and information to be accurate. Then all control activities on dispatching, routing and scheduling can be made in advance in this system. Otherwise, due to the unpredictable nature of the transportation processes, control systems with real time decision making ability are required. Table3 provides descriptions and benefits of these two main divisions and sub-divisions in control system. As mentioned offline policy could be managed simultaneously or based on the priorities in the queue.

Table 3

Despite the fact that pre-scheduling policy has more fortune to be collision free, but online systems are more flexible and adaptable for judgmental situations such as dynamic interactions, randomized processing time, overlapping between requests and breakdowns. These online control systems can be used peripherally and centrally. Berman and Edan [29] suggested a complete control methodology for decentralised autonomous control of dynamic manufacturing AGVS. They believe this strategy for material handling brings high robustness and flexibility. A system management navigation and load transfer beside a hierarchical fuzzy control accompanied by a priori path optimization and a direct determination was supposed in their paper. Watanabe et al. [30] applied Sparse Distributed Memory (SDM) of neural network for the recognition and acquisition in the navigation of AGVs and then used Q-learning (Q-L) to navigate AGVs. Q-L could refine routes in the guiding AGVs problem based on stochastic dynamic programming. They avoid collisions by mutual understanding of behaviours between vehicles.

If a single control system at the same time controls all AGVs in the system, it will be considered as centralized. Centralized control system supports all data about AGVs in the system; the position, loading, unloading, remained capacity and etc. So in order to service cells, dispatching AGVs could be based on the priority of AGVs or the priority of workstations. Mantel and Landeweerd [31] considered the kind of track layout and number of AGVs for both job shop and flow shop in a centralized control system. They mentioned that in a centralized system all transportation tasks are considered concurrently and some policies like First Encountered First Served (FEFS) rule is useful for decentralized systems like a tandem or single loop configuration.

Sometimes a combination of traditional and progressive methods could perform better; Maza and Castagna [32] used a mixture of a pre-planned and a real time method in order to construct conflict free routs. They showed the efficiency of their work on the framework of Arena® simulation software. According to Bozer and Srinivasan [33] the control schemes used in practice are getting gradually more intricate and more expensive to be developed, keeping principles of control systems has changed to smart systems which observe all vehicles and even system status continuously. Since any system wish to improve its job more and more with minimum expenditure, a real time planning can adjust system to its timetable. Have a real time consultant background for a pre-planned control is adviced to counter to critical situations is a wise scheme. So it is the cause that dynamic and real time routing for AGVS has attracted lot of attenuations from researchers and counted as popular subjects in Flexible Manufacturing System (FMS). An example of these studies is done by Shah et al. [34] which provided a software model for an integrated control system in dynamic vehicle guiding according to conflict free routing and scheduling based on Material Requirements Planning (MRP). For another study we can point Mes et al. [35] that considered an intelligent agent based vehicle routing by using a Vickrey auction algorithm for incoming requests. Agents are used to manage traffic by allowing AGVs to access to different sections of the guide path. The authors proved that their solution is more preferable rather than best tradition hierarchical heuristic methods. Ebben et al. [36] study the problem of real time one-way traffic control in underground automated transportation systems with large numbers of AGVs.

In other view, there are two main classifications of control which could be agent based (Manual) or automatic. Manual control is used as on board control which would be useful where there are limited P/D points and distances between each pair of P/D point is long. This kind of control is not suitable for complex systems because AGVs should be assigned to jobs independently and it would be a problematic try to bring an optimum solution with chancy assignment. In automatic dispatching, the central controller like Warehouse Management System (WMS) or Machine Operating Software (MOS) receives and prioritizes tasks and would determine proper services with the objective function of minimization travelling distances/times or number of AGVs in the system. There are large numbers of decisions in a real time control system caused from dynamic nature of servicing after servicing. In such a system it is better to use unit load AGVs to make it less conditional rule and inflate shorter queue to support stations. Totally we could define usual and essential features in manual and automatic control system as Table4.

Table 4

Anyway there are different management guidelines to control a transportation system but these policies are not as an extremity target for a control system. Also some problems are introduced in inner manufacturing transportation field which are interpreted as control and management problems .Asef-Vaziri and Goetschalckx [23] supposed that there are three main problems involved in AGVS: Block layout design problem, guide path track layout problem and Pick up and drop off stations' locations problem. However, most of the literature just study one or two of these problems simultaneously. In fact, expanding research activities for the joint optimisation of dispatching, routing and scheduling problems might be advisable to avoid deadlock conditions in large scale AGVS. It could be said that here are three main marks engaged in control system for AGVS as below:

Scheduling and routing: means Route determination and assignment of vehicles to route with minimum cost,

Dispatching: means task assignments strategy for vehicles,

Positioning: means point assignment for stopping, charging, parking, loading and downloading.

In the next subsections, transportations' characteristics like routing and scheduling, dispatching, idle positioning decisions and deadlock resolution will be discussed in more detail.

Vehicle Routing and Scheduling

Vehicle scheduling problem makes decisions about when, where and how a vehicle should act to perform tasks, for instance the routes it should take. The scheduling problem can be solved offline, if all tasks are known in advance. Though practically, precise information about jobs (tasks) is usually known at a very late instant to be scheduled. This makes offline scheduling almost impossible. Therefore to obtain more survey on vehicles, the online scheduling system seems necessary to control vehicles.

There are different criteria for routing and scheduling like; performance, queue length and handling cost. One of the most used objectives of AGVs routing is to minimize loading times and distances. To solve this problem, algorithms have to be developed. The simplest solution for the routing and scheduling problem is no movement and as result no cost that is difficulty practicable and acceptable for a complete manufacturing process so there is need to find proper travel cost. In this case the cost of routing is depended on the units of products and distances between applicant locations.

In this regard two categories of algorithms can be distinguished: static and dynamic algorithms. With static algorithms the route from node to node is defined in advance and is always used if a load has to be transported between. In dynamic routing, real time information plays main role for control scheme to make touting decision and, as a result, various routes between can be chosen. Static routing problems in AGV systems are related to vehicle routing problems (VRP) studied in transportation literature. References [37] provided an overview of literature in this area. A more recent paper observing this problem is from Kelly and Xu [40]. A set partitioning based heuristic has been proposed in their study. In a systematic way, sections of routes are joined to gain high quality solutions.

Dynamic vehicle routing problems are also studied in transportation literature. Complex demands for service involving in a real time way, have to be satisfied by vehicles. Psaraftis [41] points out the differences between static and dynamic vehicle routing. Gendreau et al. [42] propose a parallel tabu search method for real time vehicle routing and dispatching. According to Savelsbergh and Sol [43]; applying a branch and price algorithm can solve the dynamic routing of independent vehicles. Gans and Van Ryzin [44] represented the problem as a classical set covering model. In their research performance is measured in terms of congestion. As Taghaboni-Dutta and Tanchoco [45] mentioned dynamic route planning can be done in two ways: complete route planning and incremental route planning. Employing complete route planning the whole route from beginning to the final destination is determined at the same time. With incremental route planning, the route is planned segment for segment until the vehicle reaches its final destination. The disadvantage of complete route planning is that a route may become invalid during operation due to unexpected events. However, with applying incremental route planning, the optimality of the complete route is disregarded[46].

Meersmans and Wagelmans [47] proposed a model that deals with the integrated scheduling of all types of material handling equipments at an automated container terminal. In their study, a beam search heuristic was presented for large scale problems. It has been shown that with the heuristic methods, solutions close to optimality can be achieved in reasonable computation times. Both the integrated approach of [[47], [48]] did not take into account that there is a limited space for AGVs waiting to be loaded/ unloaded at the cranes. Ebben et al. [49] presented a generic approach to model integrated scheduling problems with a finite horizon such that capacity, parking space and release time constraints are met. The main decision within the model concerns when and how to transport all jobs from their origins to their destinations. The output of their model includes an assignment of orders to vehicles, a route for each vehicle and an assignment of vehicles to parking areas. This study is a mixture of scheduling and dispatching problems.

To conclude, in the context of manufacturing areas, to solve the vehicles routing and scheduling problems, static and dynamic algorithms have been developed. Different techniques like network models, queuing networks, simulation and intelligent routing techniques are used to optimize AGVs route network. The routing of AGVs through distribution, transhipment and transportation systems is hardly studied. From the literature discussed above, we deduce that scheduling and routing issues could cover each other partly. However the combination of scheduling and routing aspects however forms a challenging problem, more researches on integration of scheduling and routing problems are needed. Furthermore, more attentions should be paid to the simultaneous scheduling of different types of material handling equipment in large AGV systems.

Vehicle Dispatching

Dispatching refers to a rule used to select a vehicle to perform a transportation demand [46]. Strongly dispatching is depended on three factors:

Layout of the system and appropriate guide path,

Fleet size,

Movement strategy

Dispatching, routing and scheduling decisions can be made parallel or separately. Taghaboni and Tanchoco [50] developed an intelligent control in their study to dispatch, route and schedule a fleet of free ranging AGVs in a compatible way, based on real time information.

For dispatching, a specific work list for each AGV accomplish under different policies like:

NWF: Nearest- Work center- First,

STTF: Shortest- Travel- Time- First,

FRFS: First- Ready- First- Served,

FEFS: First- Encountered- First- Served

Since the dispatching strategy is depended on the layout and guide path, it is noticeable to define a summary of different kind stations to complete dispatching regarding to literatures. We attribute pick up and drop off point as P/D points and notice that one-to-one, one-to many, many-to one and many-to-many are identified as 1-1, 1-M, M-1and M-M respectively. M-M is more common among other dispatching methods in servicing between P/D points. More description about dispatching strategy between P/D points and suitable guide path is described in four classes:

1-1: Balanced distributed requisitions across layout with specific route for each AGV,

1-M: One depot is responsible to feed several work centers,

M-1: Outgoing of different work centers are collected toward one depot,

M-M: Several work centers have some things to send for each other

Furthermore dispatching of AGVs is drastically influenced by the statuses of AGVs for P/D tasks that are classified in Table5. This table assumes the kind of an AGV which could be busy or idle based on the capacity and assigned tasks would determine dispatching customs between nodes (i.e. A busy vehicle does not launch next task until accomplish last task). It is clear that an empty AGV could be assigned to any task and after this assignment it will not be empty but before that was.

Table 5

We examined dispatching problems from different points of view. Firstly, a work enter request for an available idle AGV to load ready good and secondly, a most idle vehicle would be assigned to a new task. Consequently, the dispatching problem is divided into two categories of online and offline dispatching rules.

In offline control systems all data on transportation requests are available at the start of the transportation process. As a result, vehicles can be assigned to loads in an optimal way by formulating the dispatching problem as an assignment problem.

Usually a simple heuristic used in online control systems is the FEFS rule, which dispatches a free AGV to the load that requested transport at the earliest time. According to Egbelu and Tanchoco [51]; the following heuristic rules can be applied in decentralized control systems for work centre initiated dispatching:

Random vehicle rule: Pick-up task is randomly assigned to any available vehicle regardless of the location of the vehicle and the load,

Nearest vehicle rule: The vehicle at the shortest distance of the load is assigned to the load,

Farthest vehicle rule: The vehicle at the greatest distance of the load is assigned to the new transportation request,

Longest idle vehicle rule: The vehicle that has remained idle for the longest time among all idle vehicles is dispatched to the load,

Least utilized vehicle rule: The vehicle with the minimum mean utilisation is assigned to the new job.

Nearest vehicle rule could be elucidated regarding to the kind of AGVs either empty or busy and consecutively unit load or multi load. Table6 determines these characteristics to dispatch nearest AGV.

Table 6

Yamashita [52] proposed two variations on FEFS policies. These policies dispatch empty vehicles such that variations in waiting times of the various pickup and delivery points are minimized. Hodgson et al. [53] model an AGV system as a Markov decision process. Dispatching policies resulting from the semi-Markov decision model have been described and have been tested for single load and dual load vehicles. Kim et al. [54] proposed an AGV dispatching procedure based on the objective to balance work loading the system. A dispatching rule with self adaptive capability has been introduced in Kim and Hwang [55]. A knowledge based system for selecting an AGV and selecting a work centre from a set of work centres requesting transport simultaneously is presented in [56]. Yim and Linn [57] test the performance of various heuristics with PNM. Bozer andYen [58] presented dispatching rules which take advantage of information available in centrally controlled systems. Tan and Tang [59] used fuzzy logic to dispatch AGVs in a manufacturing environment. Concepts of the Taguchi method have been used to determine an optimal set of weights for various decision criteria. Aytug et al. [60] inspected reciprocal actions between dispatching rules and deadlock avoidance policies. Jeong and Randhawa [61] developed a multi attribute dispatching rule which considers three criteria simultaneously. A neural network approach has been used to assign weights to each of the criteria based on the current status of the system. These weights are continuously changing in the multi attribute rule. It has been shown that this rule performs better than single attribute rules and a multi attribute rule with fixed weights, from the viewpoint of various performance measures such as block time of an AGV. Hall et al. [62] considered a manufacturing system with AGVs travelling in unidirectional loops. The authors analysed three widely used dispatching policies.

Summarising, in the area of manufacturing numerous researches has been executed on the dispatching of vehicles to jobs. The solution methods vary from simple dispatching heuristic rules, Markov decision processes, to fuzzy logic and neural network approaches. Whereas it can be concluded that these new concepts hardly exceed traditional heuristic rules and their modifications in performance. The dispatching of AGVs in distribution, transhipment and transportation systems can be exploring in future research. Although, for real life systems with large numbers of AGVs more research for advanced heuristics or optimal approaches is required. In this context the following aspects are of utmost importance: low computation times for large systems with high workloads, interface with other vehicles to avoid congestion, deadlocks and delays, infinite or rolling planning horizons, and a small gap between optimal solutions and solutions of heuristics. In the new areas of application, the interface with the other types of equipments such as storage and loading/unloading equipment that is very important. Hardly any attention has been paid in literature to the simultaneous dispatching of multiple types of material handling equipment. In integral dispatching of vehicles and other types of equipment in large AGV systems, side constraints are very important. One could think of capacity restrictions at the transfer buffer from storage equipment to AGVs or priority of one type of equipment above another one.

Vehicle Positioning

The pattern of material flows in a plant definitely affect the material handling costs and is planned in advance, but there are some places should be located such as P/D points, parking, charging places or some stopping spots as dwell points in the transportation process which are usable for AGVs. The problems related to these points are classified in the layout and location problems.

Vehicle idleness is unavoidable in automated guided vehicle systems. Rather than forcing vehicles to return to the vehicle depot, it is better to park vehicles at locations (vehicle home locations or dwell points) that are closer to load-release locations than the vehicle depot. According to Egbelu [63] the following rules are mostly used for positioning idle vehicles:

Central zone positioning rule,

Circulatory loop positioning rule,

Point of release positioning rule.

Reference [63] is one of the first in literature to study the problem of positioning an idle AGV in a loop layout. A static positioning strategy is proposed in [64], [65]]. In such a strategy, the location of a parking area is not changed. This version of the problem with one vehicle can be solved in polynomial time, by modelling it as a discrete time stationary Markov chain. Lee and Ventura [66] developed a dynamic programming algorithm to determine dwell points for AGVs in unidirectional and bidirectional loop layouts while minimizing mean weighted response times. They show that this algorithm can also be used in a tandem loop layout with multiple vehicles.

Briefly, analytical approaches, such as dynamic programming and Markov chains, have been used to position idle AGVs in various layouts in manufacturing areas. For a fixed number of AGVs the problem can be solved to optimality in polynomial time. Within large AGV systems, AGVs need to travel large distances before they reach a parking location. For these types of systems it is important that an AGV can be disturbed in its route to a parking area and can be rerouted to a new job. New efficient algorithms should be developed to deal with these dynamic aspects.

Deadlock and Collisions Prevention

One of the results from decentralized real time planning is deadlock, which is one of the realistic subject to handle large logistics systems with highly dynamic matters and unknown future events. A deadlock occurs when there are no handling equipments to proceed to transportation processes. Encountering online requirements need to decompose the entire transportation control system into various components for different types of reaction.

There are infrequent meet to this kind of AGVS problem in traffic scheduling or dispatching strategy. Although [[67],[68]] determined deadlock and traffic circumstance in their papers but Lehmann et al. [69] is the study that objectively determined deadlocks. They provided two different methods for deadlocks; the first one was based on a matrix of the terminal system and the second one was established upon tracking the requests. Consequently three different methods were proposed to adapt the handling operations or to decide about alternative resources so that conflicts are refined. They determined quay cranes and AGVs involving handling and demonstrated their studies in a simulation framework.

Rely on the guide path and control system; deadlock resolution problems are important in AGVS. Although this issue does not take apart in tandem guide paths but usually happens in other systems, particularly conventional guide path systems. These unsightly matters could be under control through paths balancing, traffic management and zone planning.

In addition to as far as we discussed, there are many components, questions and absolutely proposition for every kind of AGVS that are argued for many years. Each impasse problem draws inward some solutions exactly or locally, but the measurement of successful explanation is depended on harmony of system's output to which is expected. So implementing and applying an AGVS must be followed exactly, otherwise the system may never get implemented effectively.


In This paper related literature to automated guided vehicle systems have been reviewed. Simply classified different kind of definition, capabilities, advantages and disadvantages for these vehicles in different practical areas have been defined. Requirements and sub-requirements in the system design and handling management to set up an AGVS in flexible manufacturing system are mentioned. Because each necessary feature in FMS supplies others and vice versa that could be proved by surveying literature in this field, therefore, it is hard to determine which parts of these complex handling systems are more significant and studied rather than other involving varied parameters and variables. It can be seen in Table7 that most studies mentioned here are overlapped with other studies in different sections of design and management. Except the references of sec.2.3, other references could be found in different categories of Table7. Although there are attractive topics that gather more attentions but regards to the kind of processes it would be changed. For example as Fig.3 displays there are increasing interest for dispatching strategies after routing and scheduling of AGVS in discussed papers that are scattered from 1981 to 2010 in different areas. While management and control polices attracted more researchers, guide path draws most coverage between other subsections in design field. Mainly and usually recent researches in AGVS supposed and covered different aspects of this system which had made it more complicated. Therefore in current collection we integrate most relevant or perhaps common subjects for design, routing, scheduling, dispatching and positioning of automated transportation via intelligent AGVs to clarify how a suitable system could be developed. However further studies may be required to improve its growing.


Fig. 3

Conflict of Interest

The authors declare that they have no conflict of interest.