5.8 Conflict Resolution
5.8.2 Zone Control Construction
The used zone control approach for avoiding vehicle collisions consists of the declaration of dedicated zones in the potroom and a set of elementary traffic rules. We define the following zones:
• northern back aisle of each segment (indicated green in Figure 5.10); • northern center aisle of each segment (indicated blue);
• southern center aisle of each segment (indicated red); • southern back aisle of each segment (indicated white); • cross aisle in-between segments (indicated yellow); • cross aisle in-between potrooms (indicated black).
5.8. Conflict Resolution 89
FIGURE 5.10: Zone partitioning. Zones in each segment are: northern back aisle (green), northern center aisle (blue), southern center aisle (red), southern back aisle (white). Additionally, a zone is dedicated to the paths in-between segments (yellow)
and in-between potrooms (black).
Independent of the type of transport, we restrict the number of AGAPTVs to be at most one per zone. As soon as a vehicle wants to enter the next zone, it is checked at the checkpoint, which are the intersection points of the zones, whether that zone is already occupied. When this is the case, priority is given first to vehicles according to the designated priority rule. Based on defined zones and intersection checkpoints at the zones, we decide to consider stochastic priority rules that give priority to a vehicle at random. Each time both vehicles reach the intersecting checkpoint, priority is given based on an uniform probability distribution function with a variable as shown in Table 5.4. By default the probability distribution functions grant the priority to a vehicle on 50% of the time. However, in the case of a blocked segment aisle, the vehicle in that aisle always gets priority, which means that the conflicting cross aisle vehicle should conduct the avoidance maneuver.
TABLE5.4: Zone partitioning priority rules probabilities.
Interference with
Segment aisle Blocked segment aisle* Cross aisle Potroom aisle Check- point Segment aisle - - ZSC Cross aisle ZCS ZCBS - ZCP Potroom aisle - - ZP C - *:
Blocked due to either metal tapping or anode changing (crane).
Appendix F illustrates how a collision avoidance maneuver is carried out when an AGAPTV in the cross aisle detects a possible collision at the checkpoint of a segment aisle and the AGAPTV in the segment aisle grants priority. In such a maneuver, the vehicle positioned in the cross aisle maneuvers away from the intersection such that the other vehicle can continue its route. Similar procedures are used in other conflicting cases.
Notice that this approach supports in eliminating deadlocks. In particular system states in which livelocks, non-local deadlocks and local deadlocks happen will benefit from the stochastic element that allows vehicles to (temporary) escape from blockades. Despite that deadlocks cannot be avoided entirely with this strategy, e.g., because the system could end up in a state in which the AGAPTVs
are continuously conducting avoidance maneuvers, we argue that by considering a set of stochastic variables, the waiting times will be acceptable in many cases and the number of avoidance maneu- vers limited. In future research, one could investigate the impact of more sophisticated zone-control methods like adjustable priority rules or dynamic zones.
5.9
Conclusion
The AGV system design discussed in this chapter, outlines how strategic, tactic, and operational decisions are translated to functionalities of the AGV system framework. To summarize the fundings with respect to the decision framework of Le-Anh and Koster (2006) we conclude:
• Requirements and data: the AGV system incorporates facility characteristics and constraints such as blockades. Since there is no uniformly used approach for generating anode pallet de- mand (i.e., every smelter is unique), the system must be able to include a variety of demand patterns.
• Guide-path design: a conventional flow with multiple lanes that are bidirectional. • Estimating the number of vehicles: Formula 5.1.
• Vehicle scheduling: vehicle initiated dispatching (five rules) and work center-initiated dispatch- ing (seven rules).
• Vehicle positioning: distributed-positioning rule where the designation of dwellpoints and allo- cation rules is case specific. Parking location can only be designated in outer parts of the potline where there is sufficient space. Idling vehicles are send to these places by a low priority task. • Battery management: automatic and opportunity charging. The automatic variant includes
charging the battery at the moment an automatic charging threshold is reached and charges either until a fixed levelbplatis reached or until a certain timebδt is passed, while in the oppor-
tunity charging method the battery is recharged once there is sufficient time in-between jobs. An opportunity charge is only scheduled once the opportunity arises that at leastθαtime units can be used to increase the battery at least byα%.
• Vehicle routing: reactive and proactive.
• Deadlock resolution: conflicts are detected by a forward sensing technique in addition to a zone control method. The guide-path network is divided into fixed zones which may be occupied by at most one vehicle at the same time. Priority rules are used for granting an AGAPTV priority above the other. In this zone partitioning rule, a stochastic variable expresses which vehicle is granted priority.
91
Chapter 6
Model Design: Scenario Evaluation
Model
This chapter provides the evaluation model of the AGV implementation with corresponding MAS control strategy. Recall that the MAS functions as a planning and control system for the AGV system (see Figure 6.1). These two models are integrated into an evaluation model of which the conceptual model and the model content are presented in this chapter. Section 6.1 presents the conceptual model of integrated MAS and AGV system. Section 6.2 describes the model content. Section 6.3 concludes this chapter.
FIGURE6.1: Methodology outline from building the models to model validation.
6.1
Conceptual Model
To design a sound and valid simulation model, a conceptual model is presented first. The conceptual model describes an abstraction of the ’real world’ and forms the blueprint of an implementable model. A formal definition of a conceptual model is given by Robinson (2008a): a non-software specific description of the computer simulation model, describing the objectives, inputs, outputs, content, assumption, and simplifications of the model.
We follow a structured approach to design a valid simulation model. We decide to select the framework for conceptual modeling as reported in Robinson (2008b). Basically, this framework con- sists of the activities:
• Understanding the problem situation (Subsection 6.1.1)
• Determining the modelling and general project objectives (Subsection 6.1.2) • Identifying the model outputs (responses) (Subsection 6.1.3)
• Identifying the model inputs (experimental factors) (Subsection 6.1.4
• Determining the model content (scope and level of detail), identifying any assumptions and simplifications (Section 6.2).
Additionally, to enhance the verifiability and validity of our model, we regularly interact with the managers and discuss the assumptions and simplifications with the management. Below we subsequently discuss the aforementioned framework activities.