Guide-path design is the first important issue needs to be considered for AGV design and control. The main purpose is to determine and include the connections or guide-path segments in the solution. The number of parallel lanes of a connection also needs to be decided for some specific cases. It is necessary to know the material flow between different departments in order to construct a “from-to” chart for the guide-path design problem [Tuan]. In a network flow model, nodes and arcs are used to represent the pick-up and delivery (P/D) locations and aisle intersections respectively. Arcs are the paths that AGVs can follow when traveling from node to node.
Arcs can also be directed to indicate the traveling direction of vehicles. Furthermore, the cost can also be illustrated by arcs by marking distance of a segment between two end nodes and time cost by a vehicle to travel along the arc. A 0-1 integer optimization model is used to simplify and describe the network-flow model [Tuan]. The main goal of a guide-path design is to find a solution that can minimize the total vehicle travel distance. A non-negligible challenge for a guide-path design is the information shortage. It is difficult to make a precise anticipation about the flow of materials in a warehouse since it can change over time.
The Guide-path system can be generally classified by characteristics shown in Table 1 [Tuan]. The flow topology indicates the the level of complexity of the guide-path network. The most simple guide-path system consists of only one single loop while a tandem configuration is a combination that constructed by several loops. The most complicated guide-path system is the conventional topology. It contains paths, crosses, shortcuts and junctions. A path segment in a network can contain only one lane or multiple parallel lanes. Vehicles can either travel in only one direction (unidirectional) or both directions (bidirectional).
Based on the characteristics of a setting and experiences from the designer, the type of guide-path system can be decided. After that, an appropriate mathematical model is used to obtain the best possible guide-path system. In general, the single-loop systems are more common in cross-dock centers, and tandem configuration is more suitable for manufacturing circumstances where workstations are usually organized into manufacturing cells. The conventional guide-path systems is the most appropriate choice for warehouses and distribution centers.
There are three most popular guide-path systems. The single-loop guide-path system, the tandem guide-path system, and the conventional guide-path system.
In a single-loop design, vehicles can only travel in one loop without any shortcuts or alternative routes. And vehicles in the single-loop system usually travel in unidirectional mode. The bidirectional travel mode is not suitable for this system is because there might be vehicle interferences. In the single-loop system, the first-encounter-first-serve (FEFS) dispatching rule is applied to control vehicles. By following this rule, vehicle should pick up the first load it encounters and transport it to the target location. With the purpose of finding the best single-loop guide-path and to locate the P/D locations along the loop, and optimal procedure to design a single-loop system is proposed by Tanchoco and Sinriech in 1992 [Tanchoco]. The procedure is described as the following [Tanchoco]:
Because the iterative algorithm about solving two 0-1 IP models consumes a lot of time for practical problems, later in 993, Sinriech and Tanchoco (1993), Chen et al. (1999), and Asef-Vaziri et al. (2000) proposed other models solution procedures. A mixed-IP model was presented by Chen to design guide-paths for a single-loop rail(path) system (SLDR), it is a special class of the single-loop system [Chen]. Although the system consists of only one single loop, vehicles in the system can travel on two parallel tracks. Being able to capture the vehicle failure rate in the objective function, the model can produce more reliable results [Chen].
The throughput capacity of the single-loop system is slightly less than the conventional system by about 6% [Tanchoco and Sinriech]. In order to achieve the same throughput with the conventional system, more vehicles are demanded by the single-loop system. For instance, a single-loop system will require 10 vehicles to have the same throughput as the conventional system with 7 vehicles [Tanchoco and Sinriech]. Despite this, there will be vehicle interference if there are multiple vehicles operating at different velocities.
There are two categories under the conventional guide-path system: the unidirectional system and the bidirectional system.
In facilities like warehouses and distribution centers, the unidirectional conventional guide-path systems are more popularly applied. A 0-1 integer programming (IP) model. The goal of this model is to find the guide-path (flow-path) that can minimize the total vehicle loaded travel time [Gaskins and Tanchoco 1987]. There are three main constraints in their model: (1) ensuring that not any node becomes a sink node (a sink node means the location where vehicle terminates its travel); (2) ensuring not any group of nodes becomes a sink; (3) ensuring the shortest path is chosen (optional). Several improved 0-1 integer programming models for the guide-path design problem were proposed based on the model developed by Gaskin and Tanchoco. However, the the guide-path design 0-1 IP model is still too big for practical purpose because it takes long time to find appropriate procedure solutions. To increase the processing speed, two methods were proposed. Goetz and Egbelu focused on the major flows between departments [Goetz] while Sinriech and Tanchoco considered only interaction nodes in their branch-and-bound algorithm [Sinriech].
Solution quality generated by guide-path system can be negatively affected by empty vehicle travel time and time lost by vehicle interactions [Tuan]. To overcome this challenge, J. K. Lim used the Q-learning technique, a process that maximize the numerical reward by learning how to match states with actions, and the results are superior [Lim].
In general, to design a well performed unidirectional guide-path system, it is necessary to consider the P/D stations, time lost caused by empty vehicle travel and interferences. The faster solution approaches are still in urgent demand.
Although the bidirectional guide-path system will have a higher productivity than unidirectional ones, it is not popular applied in systems dealing with material handling. This issue can be solved by using dual unidirectional lanes but it will take more space and cost more. The studies relate to conventional bidirectional guide-path is very limited. But Gaskin proposed a model that can minimize the the travel distance and number of lanes for vehicles [Gaskin]. The bidirectional guide-path system are
The tandem guide-path system contains multiple zones. For each zone, there is only one vehicle to pick-up and deliver loads. Transfer stations are used to communicate between zones. In a tandem-loop configuration, transportations are achieved by non-overlapping single-loop paths (Fig. 2). Vehicle blockings and interference are eliminated in tandem systems [Bozer]. Compared to conventional system, tandem system is easier to control and expand. However, it might require additional transfer buffers which will increase cost and processing time. Tandem system is vulnerable to system failures and the throughput might not as much as the conventional systems.