Conference PaperPDF Available

Enhancing Charging & Parking Processes of AGV Systems: Progressive Theoretical Considerations

Conference Paper

Enhancing Charging & Parking Processes of AGV Systems: Progressive Theoretical Considerations

Abstract and Figures

This paper presents our work in progress for the development of an efficient charging & parking strategy. Our research aim is to develop a strategy that not only provides an efficient approach to charging AGV batteries, but also reduces traffic density in a highly utilised large-scale AGV system. Alongside the current state-of-the-art solution, three new allocation methods are introduced: Trivial+, Pearl Chain and a method based on the Generalised Assignment Problem (GAP). These four methods vary in their scope, in terms of number of vehicles considered, when calculating a decision for a specific vehicle. Furthermore, two types of availability rules for vehicles are introduced and evaluated. Their combination with the allocation methods lay the foundation for future research. All allocation methods and availability rules are explained in detail and this is followed by a summary of the expected outcomes.
Content may be subject to copyright.
Enhancing Charging & Parking Processes of AGV Systems:
Progressive Theoretical Considerations
Maximilian Selmair
BMW Group
80788 Munich, Germany
Email: maximilian.selmair@bmw.de
Tobias Maurer
Regensburg University
of Applied Sciences
93049 Regensburg, Germany
Email: tobias1.maurer
@st.oth-regensburg.de
Abstract—This paper presents our work in progress for the
development of an efficient charging & parking strategy. Our
research aim is to develop a strategy that not only provides an
efficient approach to charging Automated Guided Vehicle (AGV)
batteries, but also reduces traffic density in a highly utilised
large-scale AGV system. Alongside the current state-of-the-art
solution, three new allocation methods are introduced: Trivial+,
Pearl Chain and a method based on the Generalised Assignment
Problem (GAP). These four methods vary in their scope, in terms
of number of vehicles considered, when calculating a decision for
a specific vehicle. Furthermore, two types of availability rules for
vehicles are introduced and evaluated. Their combination with
the allocation methods lay the foundation for future research. All
allocation methods and availability rules are explained in detail
and this is followed by a summary of the expected outcomes.
KeywordsAGVs; Automated Guided Vehicles; Battery Charg-
ing; Charging Strategy.
I. INTRODUCTION
Due to the increasing demand for individualised goods,
and tighter production schedules, production and the related
logistics have needed to become more and more flexible [1].
It follows that internal logistics have to adapt efficiently to the
increasing dynamics and complexity of production processes.
The visionary concept of Logistics 4.0 is proposed to offer a
wide range of solutions for this [2] [3]. The use of AGVs, in
particular, is gaining attention and is becoming the focal point
of researchers across different industries [4]–[7].
AGVs are ready for use at any time of day, can be installed
with relatively little effort and can move freely in their
application environment, due to the lack of required physical
guiding systems [8]. The increasing environmental flexibility
means that vehicles can be used for a greater range of scenarios
[5]. Taking into consideration the progressing flexibility and
technological advances, Automated Guided Vehicles have the
potential to play an important role in factory logistics [9].
Yet, it is proposed that the organisational complexity increases
alongside with greater flexibility of application [10].
Besides the apparent problem of efficient task processing,
other issues relevant to the resource efficient operation of
the entire system require consideration. More specifically, this
paper focuses on scheduling the charging and parking missions
of the vehicles. Although this issue may seem to be trivial
at first, it becomes increasingly complex for large-scaled and
highly utilised systems. By choosing an optimal strategy, the
vehicle fleet can be reduced to a minimum number of vehicles
and thus fewer resources are necessary. The lower number of
AGVs results in less traffic density, which in turn leads to
fewer traffic jams and less blocked pathways [5]. As such, the
charging strategy is proposed to be an important aspect and
has the potential to improve the total performance of such a
material flow system [11]. According to Ryck et al., this topic
has not been explored in great detail so far [12].
Therefore, this paper considers and evaluates existing meth-
ods discussed in the relevant literature, and proposes concepts
to address the issue of scheduling the charging and parking
missions of the vehicles efficiently.
This paper is structured as follows: in Section II, we outline
the environmental requirements that provide the theoretical
framework of this study and Section III provides an overview
of the relevant literature. The strategies that we have developed
are introduced in Section IV . Finally, Section V and SectionVI
describe the experimental environment and the predicted out-
comes of the projected simulation study. Its results are going
to be presented in a subsequent publication.
II. ENVIRONMENTAL REQUIREMENTS
This section describes the parameters of our study, stipulated
by the constraints of a research project undertaken by the
BMW Group, to which we are contributing. For the automa-
tisation of the the internal material flow, the BMW Group
has developed the so-called Smart Transport Robot (STR)
(see Figure 1), designed to substitute common tucker trains
which currently play a central role in the automotive material
handling processes. The industrial use-case of BMW Group
specifies the following requirements:
1) Every vehicle can carry one load carrier at a time.
2) All transportation requests are issued randomly and
therefore unknown prior to their receipt.
3) Vehicles without a task are required to either recharge
or park.
Fig. 1. BMW’s STR in its natural habitat.
A. Optimisation Objective and Key Performance Indicators
The overall goal of an automotive transport vehicle system
is to complete all tasks on time. This requirement is consid-
ered to be a hard constraint, which is non-negotiable to the
detriment of any other Key Performance Indicators (KPIs).
That is, in order to meet this requirement, other resource
intensive variables are minimised to save resources. An im-
portant resource in an automotive production environment are
the pathways. These are not only utilised by AGVs, but also
by forklifts, tow trains, bicycles and pedestrians. In order to
maximise their availability for purposes other than AGVs, the
subsequent charging and parking strategy was developed.
Additional KPIs applied to evaluate the quality of loading
strategies in terms of their flexibility, efficiency and scalability
will be introduced below. In this context, flexibility is defined
as the availability of as many vehicles as possible at any time,
allowing for an immediate reaction to short notice events, such
as highly-prioritised urgent tasks.
In order to assess the efficiency of a charging strategy, it
is necessary to take into account the non-linear quality of
the charging process. That is, the amount of time required
to recharge a battery to a predetermined level depends on the
initial State of Charge (SOC). This KPI can be evaluated by
accessing the utilisation records of the charging stations.
As scalability is also an important variable of such a system,
the required computational power needs to be considered.
Consequently, a low computing effort is more likely to ensure
that a later expansion of the fleet will not cause performance
issues based on a lack of computational power.
Thus, a total of five KPIs are used to compare and evaluate
the strategies presented in this paper.
B. Source of Energy
The Smart Transport Robot is supplied with energy from
a lithium-ion battery module of BMW’s electric vehicle i3.
Side fact: the i3 is built of eight such modules. Each module
contains twelve battery cells and has 48 V as well as a capacity
of 120 Ah. The time required to fully charge one module by
means of a common method like Constant Current Constant
Voltage (CCCV) is 3.2 h. However, fast-charge methods are
likely to reduce this duration substantially. The charging
current for this method decreases from 38 and 36 A. CCCV
and also other fast-charge methods have in common that they
are are able to charge faster in a relatively small SOC than in
a high one [13]–[15].
Lithium-ion batteries have a number of advantageous prop-
erties relevant to the practical application of AGVs. Especially
the minimal negative memory effects of short charging periods
over a few minutes should be emphasised. The memory effect
refers to a loss of capacity which results from frequent partial
discharges. This feature allows for efficient, short charging
processes and continuous use over 24 hours [16]–[18]. More
advantageous properties are, e. g. the long life-cycle, low levels
of self-discharge, their price and their relatively small size and
weight [16].
C. Task Allocation and Prioritisation
The tasks are allocated by a central operating optimisation
algorithm in this experimental setting. The algorithm ensures
that all tasks are performed on time, whilst minimising the
driving effort of the vehicle fleet measured in meters. For
this purpose, all driving efforts are transferred to a matrix
that constitutes the associated GAP (for more details see
Section IV-A4). The solution to the GAP, referred to as Vogel’s
Approximation Method for non-quadratic Matrices (VAM-nq),
was developed by Selmair et al. [3], and is an extension of the
original Vogel’s Approximation Method (VAM) introduced by
Reinfeld et al. [19]. Unlike exact methods like the Hungarian
Method, proposed by Kuhn [20], and Integer Linear Program-
ming, VAM-nq approximates a solution for the GAP. As a
result, the quality of the solution is sufficient for most cases,
but the algorithm computes solutions substantially faster than
exact methods. For a detailed review, we refer the reader to
Selmair et al. [3].
To ensure that all tasks are performed on time, the system
uses a prioritisation method to schedule all tasks. This method
ensures that the system uses the temporal flexibility of each
task to minimise the driving effort of the AGVs. Extensive
research has been conducted about this prioritisation method
by Selmair et al. [21].
III. REL EVANT LITERATURE
This section presents a description of the state-of-the-art
planning strategies summarised in Table I and explains their
findings, identifies possible weaknesses and defines potential
for consolidation. Based on a review of the relevant literature,
the following key components of an allocation method are
identified:
Timing of the charging process
Choice of a charging station
Duration of the charging process
Table I serves as the basis for a comparison and summary
of planning strategies provided by studies on battery charging.
On the basis of these studies, the findings are evaluated and
examples of specific research are selected.
TABLE I. OVERVIEW OF ALLOCATION METHODS
Timing Selection
Criteria
Duration
Ebben
2001 [22]
Intermediate /
Charge
Range
Heuristics Duration of
Battery
Exchange
Kawakami &
Takata
2012 [23]
Min.
Deterioration
- Duration of
Battery
Exchange
Zou et al.
2018 [24]
Intermediate /
Below
Threshold
- Handling
Time /
Fully Loaded
Kabir & Suzuki
2018a [25]
Below
Threshold
Heuristics Duration of
Battery
Exchange
Kabir & Suzuki
2018b [11]
Below
Threshold
Nearest
Station
Fully
Loaded /
Above
Threshold
Colling et al.
2019 [26]
Scheduled Permanent
Assignment
Fully Loaded
Selmair et al.
2019 [9]
Intermediate /
Below
Threshold
Minimal
Costs
Fully
Loaded /
Displaced
Zhan et al.
2019 [27]
Below
Threshold
Heuristics Fully
Loaded /
Above
Threshold
A. Timing of the Charging Process
The trigger that prompts the charging process is usually
a predefined low battery charge level and it coincides with
a point in time when a vehicle is not performing a task
[9] [11] [22] [24] [25] [27]. In an effort to determine the opti-
mal timing, Kawakami et al., for instance, focus on minimising
battery deterioration [23]. However, as previously established,
this aspect is deemed to be negligible when it comes to
lithium-ion batteries.
Interestingly, Collinget al. distributed vehicles equally, by
using charging cycles and thus focused the attention on the
overall system [26]. However, the mentioned study is carried
out on a system with six AGVs, it remains uncertain whether
this method can be implemented for a larger system.
Charging idle vehicles, even if their battery charge levels
are above the triggering SOC, also referred to as intermediate
charging, is also suggested to be a feasible tool to enhance
a charging strategy [9] [22] [24]. Possibilities for intermediate
charging and the consideration of the SOC in comparison with
other vehicles on the field or in charging stations could lead
to an increase in flexibility, efficiency and scalability of the
system. These considerations should also be taken into account
and tested when developing the charging strategy.
B. Selecting a Charging Station
Heuristics can be particularly helpful in selecting a suitable
charging station. Commonly, this involves selecting charging
stations based on either their distance to an agent or that can be
used with the least overall delay [11] [22] [27]. The total delay
consists of the travel time to a station and the waiting time
required by an AGV to use the charging station. In contrast,
predetermining charging stations without monitoring the dis-
tance to and availability of charging stations, as implemented
by Colling et al., does not appear to be feasible for a flexible
and scalable system [26].
In fact, using heuristics can increase efficiency and flexibil-
ity. Furthermore, if the parameters allowed for the displace-
ment of vehicles currently at a charging station by vehicles
with a higher need for recharging, it is assumed that efficiency
and flexibility can be increased.
C. Duration of the Charging Process
The duration of a charging process can either be de-
fined as the time required to physically replace a battery
[11] [22] [23] or to reach its planned SOC [24] [26]. Kabir et al.
and Zhan et al. suggest that batteries should only be charged
to a level of 90 % or 95 % to shorten the inefficient phase of a
battery charge (see Section II-B) [25] [27]. All the mentioned
studies have in common that only the condition of one AGV
is considered in any decision. Selmair et al. provides a first
method that take into account the condition of more than a
single vehicle [9]. In their paper, two vehicles are compared
with each other by using an objective function that includes
the distance of a vehicle to an occupied station as well as the
SOC of both. The following conclusions can be drawn from
Selmair et al.: the possibility to interrupt a charging process
due to tasks or displacements can increase flexibility. For this
purpose, accurate parameters are necessary in order to achieve
robust processes to attain a high level of flexibility.
IV. DEV EL OP ED ST RATE GI ES
This section describes potential strategies for charging and
parking vehicles, which maintain the availability of the vehicle
fleet on the one hand, and minimise the distance travelled on
the other. Every strategy can be divided into two aspects: the
first is the allocation method itself (Subsection A) and the
second pertains to the availability of a vehicle, referred to as
availability rule, for tasks or charging and parking processes
(Subsection B).
Table II illustrates the two aspects of the strategies in
the form of a morphological box. Every combination of the
left and the right column represents a strategy – with one
exception: for the Trivial allocation method, which is derived
from the current state of the art, only the first availability rule
is applied.
TABLE II. THE TWO COMPONENTS OF A CHARGING & PARKING
STRATEGY
Allocation
Method
Availability Rule
Trivial Hard Constraint
Trivial+ Soft Constraint
Pearl Chain
GAP
A. Allocation Methods
The four allocation methods are presented in the order of
their scope, in terms of numbers of vehicles considered by
each method. Therefore, the first allocation method – namely
Trivial – only takes into account the vehicle that requires either
charging or parking. This allocation method is common in
today’s industrial transport vehicle applications (see Table I).
The Trivial allocation method is followed by three new
allocation methods, which consider other factors than merely
the vehicle that requires a charging or parking station. These
are in namely Trivial+,Parl Chain and GAP. Their scope is
illustrated in Figure 2.
Trivial Trivial+ Pearl Chain GAP
maximum number of considered vehicles
V1 V2 V3
S1 14 29 80
S2 30 23 55
S3 15 44 17
Fig. 2. The four allocation methods and their illustrated scope.
1) Trivial Allocation Method:
The Trivial allocation method combines the methods most
frequently encountered in the literature review. They are
mostly utilised in threshold-rules and nearest-agent-first allo-
cations. Table III shows an overview structured by the defined
key components, followed by their specification.
The triggering moment, in which the charging process is
prompted, is predefined as the moment when the battery falls
below a certain SOC (e. g., below 15 %). As soon as the current
transport tasks are completed, the AGV drives to the nearest
vacant charging station and the charging process continues
until the battery has reached a specific SOC (e. g., above 95 %).
This process can only be disrupted by the Availability Rules
of Vehicles (see Subsection IV-B). If a vehicle is not in the
process of performing a task and all charging stations are
occupied, it is assigned to a parking station. Vehicles in a
TABLE III. SCHEDULING RULES FOR THE TRIVIAL CHARGING &
PARKING ALLOCATION METHOD
Timing Selection
Criteria
Duration
Charging Below
Threshold
Nearest
Station
Above
Threshold /
Call for Task
Parking No Task/
Above
Threshold /
No vacant
Charging
Station
Nearest
Station
Above
Threshold /
Call for Task
TABLE IV. SCHEDULING RULES FOR THE TRIVIAL+ CHARGING &
PARKING ALLOCATION METHOD
Timing Selection
Criteria
Duration
Charging Intermediate /
Below
Threshold
Minimal
Costs
Above
Threshold /
Call for Task/
Displaced
Parking No task /
Above
Threshold /
No vacant
Charging
Station
Nearest
Station
Above
Threshold /
Call for Task
parking station can be pulled out by performing a task or if
their charge level falls below a specific threshold (e. g., below
15 %).
2) Trivial+ Allocation Method:
The Trivial+ allocation method is designed to be more
flexible, with the help of the insights gained from the literature
review. The key difference to the Trivial allocation method is
that intermediate charging are now feasible and vehicles can
be displaced from their charging stations. This requires that
the vehicle which has to be allocated must be compared to a
vehicle within a occupied station. A description of this method
is included in the Table IV and text below.
With Trivial+, vehicles can now be assigned to charging
stations regardless of their SOC. This introduces the concept
of intermediate charging. Vehicles with a very low SOC (e. g.,
less than 15 %) will still be allocated exclusively and therefore
with priority to charging stations.
Additionally, occupied charging stations will also be con-
sidered for allocation if the occupying vehicle has a, e. g.
25 %, higher battery charge level than the vehicle seeking a
charging station. This ensures that charging stations are not
occupied by vehicles that have a significantly higher battery
level than those seeking to recharge. Within the parameters
of Trivial+, the driving distance to the stations remains the
most important criterion for selecting a charging station. The
relevant station with the lowest costs (sum of driving distance
and optional additional costs) will be assigned. For occupied
stations, additional costs are imposed on the vehicle that has
to vacate its current station and travel to a new one. These
extra costs ensure that vehicles in charging stations are not
immediately displaced at will if there is a more efficient
alternative for the seeking vehicle.
The charging process is carried out until the vehicle has
either reached a sufficient charge level (e. g. above 95 %), gets
displaced or receives a task as soon as the charge level has
reached a minimum level (e. g. above 25 %).
The allocation of parking stations is unaffected by these
modifications to the methods and works the same way as
described for the Trivial allocation method.
3) Pearl Chain Allocation Method:
The Pearl Chain allocation method is based on the basic
principle of the Trivial+ allocation method. However, this
allocation method not only compares the battery level of
both vehicles, like the Trivial+, but also examines whether
displacements are efficient considering the total distance in
meters that the vehicles would have to travel. Furthermore, a
distinction is made between two situations:
1) If the seeking vehicle has, for instance, a 40 % lower
battery level than the occupying vehicle, displacement
is permitted without further constraints.
2) If the seeking vehicle’s battery level is, for example,
between 20 % and 40 % lower than that of the occupying
vehicle, displacement is permitted if the total distance
travelled by both vehicles is shorter than the distance to
other charging station options. The travelling effort of
the vehicle that needs to charge for a potential displace
is calculated as follows:
C=DVS+DVD+MD,
where DVSrepresents the travelling effort of the seeking
vehicle to the occupied charging station, whereas DVD
stands for effort involved for the displaced vehicle to
travel to another station as described below. Finally, MD
represents the potential effort required for the displaced
vehicle to maneuver out of its station.
The displaced vehicle is permitted to choose between
the stations listed below this paragraph. This principle
prevents performance issues, by a reasonable restriction
of the solution space. Therefore, the ”pearl chain“ has a
maximum length of three vehicles (see Figure 2).
Vacant charging stations
Vacant parking stations
Occupied charging stations which fulfill the case
described in Situation 1
3) If the seeking vehicle’s battery level is not at least 20 %
lower than that of the occupied vehicle, a displacement
is not permitted.
4) Allocation Method based on the Generalised Assignment
Problem:
The GAP finds its roots in the research field of applied
mathematics [28]. This problem, in its original form, consists
of a number of agents and a number of tasks. Any task can
be assigned to one of the agents. The total costs arising from
a solution may vary, depending on each chosen agent-task
assignment [29]. The optimal solution is defined as the one
that maximises the total profit by minimising the associated
costs.
Our idea for charging & parking vehicles by using the GAP,
is to transfer the demands (in the mathematical context costs)
of vehicles for driving to a station into a GAP. By using this
method, a number of vehicles can be assigned to a number of
stations by minimising the driving effort and meeting defined
constraints at the same time.
These costs can be manipulated to ensure that vehicles are
being assigned, on the basis of their SOC, to the closest
stations according to their, even if this means bypassing some
in order to allow prioritised vehicles access to these. This
mechanism ensures that vehicles of a fleet with a relatively low
SOC will have lower costs for charging stations than others,
and will have greater priority to access charging stations. The
constraints are defined as extra costs added to the driving
distance to a station in meters. These extra costs are proposed
as follows:
1) An occupied charging station is only accessible to vehi-
cles whose battery charge level is 10 or more percentage
points lower than that of the occupying vehicle. The
theoretical allocation of a vehicle that does not fulfill
this requirement is sanctioned with a penalty cost of
10,000 units and should not be accepted by a balanced
system for any reason.
2) An occupied parking station is not accessible for any
seeking vehicles with a battery charge level lower than
the level of the vehicle currently occupying the station.
The theoretical allocation of a vehicle that does not
fulfill this requirement is sanctioned with a penalty cost
of 10,000 units and should not be accepted by a balanced
system for any reason.
3) The theoretical allocation of vehicles, with a battery
level below 30%, to a parking station will be sanctioned
with a value of 1,000 units. This ensures that such
vehicles will only park if no other option is available.
4) If a vehicle is occupying a charging or parking station,
maneuver costs of 7 units are added.
In our opinion, the constraints formulated above will in-
crease the likelihood of the efficient assignment of charging
and parking stations. Constraint 1 allows the assignment to
Fig. 3. Exemplary two-bin principle: the full container is facing the
production line. The empty container has already been collected.
Replenishment will be delivered shortly.
an occupied charging station if the occupying vehicle can be
assigned to any other station and the demand vehicle’s battery
level is at least 10 % below the battery level of the vehicle
currently occupying the charging station. Due to Constraint 4,
maneuver costs are also factored into every effort undertaken
by a vehicle to vacate a station. Simultaneously, Constraint 2
allows vehicles in parking stations to become displaced by
others if it is beneficial to the overall system in terms of
distance travelled. Constraint 3 ensures that vehicles with a
battery level lower than 30% are more likely to be assigned to
charging stations than to parking stations. This was determined
to ensure that the vehicles’ batteries are not depleted entirely
– provided that enough charging stations are in the system.
B. Availability Rules of Vehicles
The defined availability rules for vehicles constitute an
interface between the task allocation and the charging &
parking strategy. These rules regulate the number of vehicles
available for performing tasks within the system. These rules
can been amended by applying two types of constraints: hard
or soft constraints.
A hard constraint cannot be disregarded for any reason. By
their nature, they can be easily explained and understood. Such
a constraint can be formulated as follows:
A vehicle on the field or at a parking station is available
for tasks when its SOC is above, e. g. 25 %
If a vehicle is charging, it will become available for tasks
as soon as its SOC exceeds, e. g. 70 %
Soft constraints can be modelled – for example – by
manipulating costs in some manner in order to change the
system’s behaviour to a desired behaviour. In the presented
case, where the system assigns vehicles to tasks by taking
into account their distance to these and their SOC, such a soft
constraint can be formulated to sanction vehicles with extra
costs if they are in a charging process or only have little battery
left. Such a formulation can be defined as follows:
A vehicle incurs extra costs in addition to the travelling
distance in meters to a new task if it is located at a
charging station. These extra costs can be calculated for
example as:
(1SOC) + 70
These extra costs will be negative for vehicles with a SOC
above 70 %. Therefore, the system will reward these vehicles
when it allocates tasks to them. They are more likely to be
chosen than other vehicles on the field or parking, even if
the distance to the task is the same. On the other hand, the
extra costs will be positive if a charging vehicle’s SOC is
below 70 %. Then, this vehicle will receive penalty costs and
is therefore less likely to be considered in comparison to other
vehicles with the same distance. The lower a vehicle’s SOC,
the higher the penalty costs and the less probable their chance
of receiving a new task.
V. EXP ER IM EN TAL ENVIRONMENT
The model layout, as shown in Figure 4, is designed to
represent a typical automotive flow production system. There
are three types of warehouses designed in this layout:
1) a full goods warehouse,
2) a storage for empties and
3) a mixed storage facility that serves as a full and empties
warehouse.
The full goods warehouses (1 and 2 in Figure 4) are the
starting points of the intracompany supply chain. The empty
dollies are collected to be prepared for reuse in Warehouse 2
and 3. In the center of the layout there are three production
lines that represent a flow production. Along the production
lines are work stations, also called fitpoints (see Figure 3).
Every work station is supplied by a predefined warehouse
at certain intervals. These intervals are determined by the
number of parts stored in a dolly, the frequency at which a
part is required and the cycle-time of the production line. The
replenishment is carried out by using a two-bin system, an
example of which is shown in Figure 3. As soon as an empty
container is collected, a full one is delivered by an AGV from
the warehouse. There is only space for one vehicle at each
station at a time. Thus, if a second vehicle was required to
enter a station that is currently occupied, it has to wait in front
of it and will block the pathway for other vehicles, potentially
causing delays. Parking stations are established between some
fitpoints along the production line (see blue house-symbols in
Figure 4). The charging stations are visualised by green dots
and are placed at the end of each production line.
VI. PREDICTED OUTCOMES
During the development of the previously presented alloca-
tion methods and availability rules, we endeavoured to predict
Fig. 4. Planned simulation environment consisting of three production lanes,
two warehouses and one station for empty dollies.
TABLE V. PREDICTED PROPERTIES OF THE STRATEGIES
Resource-
saving
Flexibility Efficiency Scalability
Trivial − −− − ++
Trivial+ 0 ++ ++ +
Pearl
Chain
+ ++ + +
GAP ++ ++ +
the advantages and disadvantages of the resulting strategies.
For this purpose, these strategies were evaluated according to
the previously defined KPIs presented in Table V. It has to
be highlighted that these outcomes are merely of a predictive
nature. Future simulations will evaluate the strategies and their
characteristics in more detail.
In summary, it is assumed that each strategy has certain
strengths, but these may be disadvantageous for other char-
acteristics like traffic density. The Trivial allocation method
is expected to be easily scalable due to its simplicity, as
this method does not require any significant computational
effort. As such, any additional vehicles do not influence the
individual calculation process. However, for this allocation
method, the vehicles’ current status, such as their position
and SOC, are not compared with each other, which does
not support an efficient and flexible allocation process as
defined beforehand. In the Trivial+ and Pearl Chain allocation
method, vehicles are compared with each other in terms of
their position and SOC. The resulting data are used to decide
whether it is beneficial to initiate a displacement process, in
which lower charged vehicles are prioritised. Furthermore, by
recharging vehicles with a low SOC, the previously described
loading curve, i. e. the lower the battery level, the faster the
charging process, can be taken advantage of. For these reasons,
positive effects in flexibility and efficiency are expected for
both methods. Pearl Chain also compares whether certain
displacements are beneficial when considering the total driving
distances measured in meters. Although this requires greater
computational effort of the entire system, it predicted to be
more resource efficient in terms of available pathways. Due to
the holistic approach of the GAP allocation method, a positive
result is expected for the KPIs resource-saving, flexibility,
efficiency. However, this holistic approach is expected to result
in a statistically significant higher computational effort. This
effort increases exponentially with the size of the system. This
fact suggests that the GAP allocation method is unlikely to be
a scalable method.
In summary, it is suggested that the combination of these
allocation methods and availability rules, as illustrated in
Table V, holds significant potential for any industrial setting
in which AGVs are applied on a larger scale.
VII. CONCLUSION
In this paper, we have published our work in progress per-
taining to the development of an efficient charging & parking
strategy, which will reduce traffic density in a high utilised
large-scale system. Alongside a state-of-the-art solution, three
new methods were introduced: Trivial+,Pearl Chain and GAP.
These methods vary in scope when calculating a decision for a
specific vehicle. While Trivial+ compares two vehicles, Pearl
Chain is able to consider up to four vehicles for a decision and
GAP actually takes all vehicles with a demand for a station
into account. Furthermore, two types of availability rules for
vehicles were proposed. Combining these availability rules
with the various allocation methods, provides several strategies
that could be examined in future research.
To analyse a system’s behaviour and the efficiency of
all strategies, a simulation study is proposed to finalise this
research. Within a simulated industrial production area, each
strategy will be simulated and the resulting decisions scruti-
nised thoroughly, and finally, the entire performance will be
compared to all other strategies.
REFERENCES
[1] M. t. Hompel, “Individualisierung als logistisch-technisches Prinzip
(Individualisation as a logistical-technical principle),” in Internet der
Dinge in der Intralogistik, W. G ¨
unthner and M. t. Hompel, Eds. Berlin,
Heidelberg: Springer Berlin Heidelberg, 2010, pp. 3–7.
[2] K.-I. Voigt, J. M. M ¨
uller, J. W. Veile, W. Becker, and M. Stradtmann,
“Industrie 4.0 – Risiken f¨
ur kleine und mittlere Unternehmen (Industry
4.0 - Risks for small and middle companies),” in Gesch¨
aftsmodelle in der
digitalen Welt, W. Becker, B. Eierle, A. Fliaster, B. Ivens, A. Leischnig,
A. Pflaum, and E. Sucky, Eds. Wiesbaden: Springer, 2019, pp. 517–538.
[3] M. Selmair, A. Swinarew, K.-J. Meier, and Y. Wang, “Solving non-
quadratic matrices in assignment problems with an improved version
of vogel’s approximation method,” in ECMS 2019 Proceedings edited
2019, 2019a, pp. 261–266.
[4] M. t. Hompel and M. Henke, “Logistik 4.0 (Logistic 4.0),” in Industrie
4.0 in Produktion, Automatisierung und Logistik, T. Bauernhansl, M. t.
Hompel, and B. Vogel-Heuser, Eds. Wiesbaden: Springer Fachmedien
Wiesbaden, 2014, pp. 615–624.
[5] P. Pagani, D. Colling, and K. Furmans, “A neural network-based algo-
rithm with genetic training for a combined job and energy management
for AGVs,Logistics Journal: Proceedings, 2018.
[6] A. Leupold, F. Senger, U. Weber, and T. Kappler, “Industrial IoT,
Digitale Welt, vol. 2, no. 3, pp. 78–84, 2018.
[7] H. Sch¨
oning, “Industry 4.0,” it - Information Technology, vol. 60, no. 3,
pp. 121–123, 2018.
[8] H. Maier, Grundlagen der Robotik (Basics of Robotics), 2nd ed., ser.
Lehrbuch Studium, 2019.
[9] M. Selmair, S. Hauers, and L. Gustafsson-Ende, “Scheduling charging
operations of autonomous AGVs in automotive in-house logistics,
Simulation in Production and Logistics 2019, 2019.
[10] V. Plenk and F. Ficker, “Industrie 4.0 (Industry 4.0),” in Digitalisierung:
Segen oder Fluch, D. Wolff and R. G¨
obel, Eds. Berlin, Heidelberg:
Springer, 2018, pp. 29–53.
[11] Q. S. Kabir and Y. Suzuki, “Comparative analysis of different routing
heuristics for the battery management of automated guided vehicles,”
International Journal of Production Research, vol. 57, no. 2, pp. 624–
641, 2018a.
[12] M. De Ryck, M. Versteyhe, and K. Shariatmadar, “Resource manage-
ment in decentralized industrial automated guided vehicle systems,”
Journal of Manufacturing Systems, vol. 54, pp. 204–214, 2020b.
[13] D. Fasthuber, “Integration der Ladeinfrastruktur in das elektrische En-
ergiesystem (integration of a charging infrastructure in an electric energy
system),” e & i Elektrotechnik und Informationstechnik, vol. 6, no. 2, p.
125, 2020.
[14] P. Keil and A. Jossen, “Charging protocols for lithium-ion batteries and
their impact on cycle life—an experimental study with different 18650
high-power cells,Journal of Energy Storage, vol. 6, pp. 125–141, 2016.
[15] W. Shen, T. T. Vo, and A. Kapoor, “Charging algorithms of lithium-
ion batteries: An overview,” in 2012 7th IEEE Conference on Industrial
Electronics and Applications (ICIEA). IEEE, 2012, pp. 1567–1572.
[16] D. U. Sauer, G. Fuchs, B. Lunz, and M. Leuthold, “Technology overview
on electricity storage - overview on the potential and on the deployment
perspectives of electricity storage technologies.
[17] T. Sasaki, Y. Ukyo, and P. Nov´
ak, “Memory effect in a lithium-ion
battery,Nature materials, vol. 12, no. 6, pp. 569–575, 2013.
[18] E. Rahimzei, K. Sann, and M. Vogel, Kompendium: Li-Ionen-Batterien
im BMWi F¨
orderprogramm IKT f ¨
ur Elektromobilit¨
at II: Smart Car –
Smart Grid – Smart Traffic, Grundlagen, Bewertungskriterien, Gesetze
und Normen. VDE Verband der Elektrotechnik: Frankfurt a. M., 2015.
[19] N. V. Reinfeld and W. R. Vogel, Mathematical Programming, Prentice-
Hall, Englewood Cliffs, 1958.
[20] H. W. Kuhn, “The hungarian method for the assignment problem,” Naval
Research Logistics Quarterly, vol. 2, no. 1-2, pp. 83–97, 1955.
[21] M. Selmair, V. Pankratz, and K.-J. Meier, “Efficient task prioritisation
for autonomous transport systems,” in ECMS 2020 Proceedings edited
by Mike Steglich, Christian Mueller, Gaby Neumann, Mathias Walther.
ECMS, 2020, pp. 322–327.
[22] M. Ebben, Logistic control in automated transportation networks. En-
schede: Twente Univ. Press, 2001.
[23] T. Kawakami and S. Takata, “Battery life cycle management for
automatic guided vehicle systems,” in Design for Innovative Value
Towards a Sustainable Society, M. Matsumoto, Y. Umeda, K. Masui,
and S. Fukushige, Eds. Dordrecht: Springer Netherlands, 2012, vol.
171, pp. 403–408.
[24] B. Zou, X. Xu, Y. Gong, and R. de Koster, “Evaluating battery
charging and swapping strategies in a robotic mobile fulfillment system,
European Journal of Operational Research, vol. 267, no. 2, pp. 733–753,
2018.
[25] Q. S. Kabir and Y. Suzuki, “Increasing manufacturing flexibility through
battery management of automated guided vehicles,” Computers & In-
dustrial Engineering, pp. 225–236, 2018b.
[26] D. Colling, J. Oehler, and K. Furmans, Battery Charging Strategies for
AGV Systems. Wissenschaftliche Gesellschaft f ¨
ur Technische Logistik,
2019.
[27] X. Zhan, L. Xu, J. Zhang, and A. Li, “Study on AGVs battery charging
strategy for improving utilization,Procedia CIRP, vol. 81, pp. 558–563,
2019.
[28] H. Kellerer, U. Pferschy, and D. Pisinger, Knapsack problems: With 33
tables, 1st ed. Berlin: Springer Berlin, 2010.
[29] L. ¨
Ozbakir, A. Baykaso˘
glu, and P. Tapkan, “Bees algorithm for general-
ized assignment problem,” Applied Mathematics and Computation, vol.
215, no. 11, pp. 3782–3795, 2010.
... Average SoC of e-AGVs [9] [11] Customer satisfaction Performance from a logistics Perspective. ...
Conference Paper
Full-text available
Substituting fuel-based and manual delivery for the last miles with electric automated guided vehicles (e-AGVs) promises to substantially reduce CO2 and noise emissions, and traffic congestion in urban areas. As they are battery-based vehicles, efficient battery charging which alleviates range and recharging limitations is crucial for the delivery system’s performance. However, when, where, and how many parcels should be transported is uncertain and so is the energy demand of e-AGVs. In this work, we develop three recharging heuristics to ensure the energy supply for e-AGVs in the last-mile delivery. We propose a simulation framework that enables a search for the best combinations of heuristics’ parameters (charging goal, state of charge lower limit, etc.) and a multi-criteria comparison among the charging heuristics regarding six key performance indicators (KPIs). We find recharging processes have a significant impact on the system performance. Only using forced emergency charging performs the worst. Interval charging performs slightly better than opportunity charging and both can be chosen based on different preferences. The parcel volume distribution also affects the system performance. The results from the three heuristics in this paper can be used as references in further work, which should focus on energy demand forecasting to design charging strategies to intervene emergency charging. In this way, the impact of recharging processes on the delivery system can be reduced.
Conference Paper
Full-text available
The efficient distribution of scarce resources has been a challenge in many different fields of research. This paper focuses on the area of operations research, more specifically, Automated Guided Vehicles intended for pick-up and delivery tasks. In time delivery in general and flexibility in particular are important KPIs for such systems. In order to meet in time requirements and maximising flexibility, three prioritisation methods embedded in a task allocation system for autonomous transport vehicles are introduced. A case study within the BMW Group aims to evaluate all three methods by means of simulation. The simulation results have revealed differences between the three methods regarding the quality of their solutions as well as their calculation performance. Here, the Flexible Prioritisation Window was found to be superior.
Article
Full-text available
Zusammenfassung Der stetige Zuwachs von Ladeinfrastruktur für die Elektromobilität stellt die Verteilnetzbetreiber vor neue Herausforderungen. Vor allem im Zusammenspiel mit leistungsstarken Verbrauchern und Erzeugern (Wärmepumpen, Photovoltaik etc.) kann es zu erheblichen Mehrbelastungen im Netz kommen, welche dieses an dessen Betriebsgrenzen führen kann. Aus dieser Motivation heraus soll diese Arbeit das grundlegende Verständnis für die Problematik der Integration von Ladeinfrastruktur in das elektrische Energiesystem erörtern. Dabei wird das prinzipielle Verhalten von Elektrofahrzeugen beim Ladevorgang aus Sicht des Netzes erarbeitet und eine mathematische Modellierung für die Analysen in zeitreihenbasierten Lastflussprogrammen vorgestellt. Zusätzlich werden die beeinflussenden Parameter des Ladevorganges diskutiert. Neben der Standortwahl ist die zu erwartende maximale Ladeleistung der Elektrofahrzeuge das entscheidende Kriterium für die Netzplanung. Eine Aufarbeitung dieser beiden Punkte zeigt, welche Erkenntnisse daraus für die Netzplanung abgeleitet werden können. Abschließend wird ein Ausblick über weitergehende Herausforderungen zur allgemeinen Thematik der Integration von Elektrofahrzeugen im Verteilnetz für die nächsten Jahre gegeben. Einige Handlungsempfehlungen für eine erfolgreiche hochskalierte Integration von Ladeinfrastruktur schließen letztlich die Abhandlung.
Article
Full-text available
The battery charging strategy of automation guided vehicles (AGVs) is playing an important role in improving the utilization of AGVs. This paper proposes two dual-stage battery charging strategies of AGVs, which equip with lithium-ion battery for improving the utilization. In stage 1, two routing decisions based on heuristic algorithm are developed, namely the nearest charging station (NCS) and the minimum delay charging station (MDCS); while in stage 2, the duration of each charging occurrence is reduced considering the charging characteristic of lithium-ion battery. A real case is adopted to illustrate the applicability and effectiveness of the proposed approach.
Conference Paper
Full-text available
Scheduling approaches for the charging of Automated Guided Vehicles (AGVs) are based on three key components: the timing of charging processes, the selection of a charging station and the duration of the charging process. Based on literature research introduced in this paper, two scheduling approaches have been studied: a rigid approach, based on state-of-the-art solutions, captures the optimal case for a single AGV. A flexible approach, particularly focusing on autonomous behaviour of AGVs, aims for an optimum for the whole AGV fleet. Therefore, the concept of auction-based task allocation is transferred. A closed-loop simulation compares both scheduling approaches for the application of automotive in-house logistics. The flexible approach shows a higher scheduling effectiveness, although influenced by the charging station allocation.
Conference Paper
Full-text available
The efficient allocation of tasks to vehicles in a fleet of self-driving vehicles (SDV) becomes challenging for large-scale systems (e. g. more than hundred vehicles). Operations research provides different methods that can be applied to solve such assignment problems. Integer Linear Programming (ILP), the Hungarian Method (HM) or Vogel's Approximation Method (VAM) are frequently used in related literature (Paul 2018; Dinagar and Keerthivasan 2018; Nahar et al. 2018; Ahmed et al. 2016; Koruko˘ glu and Ballı 2011; Balakrishnan 1990). The underlying paper proposes an adapted version of VAM which reaches better solutions for non-quadratic matrices, namely Vogel's Approximation Method for non-quadratic Matrices (VAM-nq). Subsequently, VAM-nq is compared with ILP, HM and VAM by solving matrices of different sizes in computational experiments in order to determine the proximity to the optimal solution and the computation time. The experimental results demonstrated that both VAM and VAM-nq are five to ten times faster in computing results than HM and ILP across all tested matrix sizes. However, we proved that VAM is not able to generate optimal solutions in large quadratic matrices constantly (starting at approx. 15 × 15) or small non-quadratic matrices (starting at approx. 5 × 6). In fact, we show that VAM produces insufficient results especially for non-quadratic matrices. The result deviate further from the optimum if the matrix size increases. Our proposed VAM-nq is able to provide similar results as the original VAM for quadratic matrices, but delivers much better results in non-quadratic instances often reaching an optimum solution. This is especially important for practical use cases since quadratic matrices are rather rare.
Article
This paper proposes an advanced decentralized method where an Automated Guided Vehicle (AGV) can optimally insert charging stations into an already assigned optimal tour of task locations. In today's industrial AGV systems, advanced algorithms and techniques are used to control the whole fleet of AGVs robustly and efficiently. While in academia, much research is conducted towards every aspect of AGV control. However, resource management or battery management is still one aspect which is usually omitted in research. In current industrial AGV systems, AGVs operate until their resource level drops below a certain threshold. Subsequently, they head to a charging station to charge fully. This programmed behaviour may have a negative impact on the manufacturing systems performance. AGVs lose time charging at inconvenient moments while this time loss could be avoided. Using the approach, an AGV can choose independently when it will visit a charging station and how long it will charge there. A general constrained optimization algorithm will be used to solve the problem and the current industrial resource management will be used as a benchmark. We use a simple extension of the Traveling Salesman Problem (TSP) representation to model our approach. The paper follows a decentral approach which is in the interest of the authors. The result of the proposal is a compact and practical method which can be used in today's operative central or decentral controlled AGV systems.
Article
This paper explores how different routing techniques for the battery management of automated guided vehicles (AGVs) can affect the performance of a system. Four heuristics available in the literature were the basis of this study. Simulation models were developed to investigate how the routing of an AGV towards a battery station can affect the productivity of a manufacturing facility. Results show that the best productivity can be achieved when a routing heuristic tries to jointly minimise the total travel distance and waiting time at a battery station. The gain in productivity, when compared with the highest possible gain theoretically achievable, is quite substantial. It was also found that higher frequency of decision-making (i.e. decisions with smaller time interval) about battery swapping helps to increase the productivity of a system.
Article
Purpose of this paper is to explore the possibility of increasing manufacturing capacities in the short run through battery management of automated guided vehicles (AGVs). The paper investigates how the duration of battery charging for AGVs can be varied to increase flexibility of a manufacturing system. The key concept is that a lead acid battery, the most widely used battery type for AGVs, receives most of its charge during the initial phase (time) of charging as opposed to the later phase. Consequently, more productive hours can be obtained from the AGVs by reducing the duration of each charging occurrence (i.e., by recharging the batteries to less than full capacity). In this approach, an AGV needs to be recharged more frequently, but the total productive hours available from the AGVs can increase. Simulation models were developed to investigate the effect of this approach. Results show that productivity of a manufacturing system increases significantly through this approach. This approach can be quite helpful for a firm if AGVs are the bottleneck in its manufacturing plant and the firm needs to have a significant improvement of its productivity in the short-run. Also, this approach does not require a firm to make a large investment to add more AGVs to its facility.
Article
Robotic mobile fulfillment systems (RMFS) have seen many implementations in recent years, due to their high flexibility and low operational cost. Such a system stores goods in movable shelves and uses movable robots to transport the shelves. The robot is battery powered and the battery depletes during operations, which can seriously affect the performance of the system. This study focuses on battery management problem in an RMFS, considering a battery swapping and a battery charging strategy with plug-in or inductive charging. We build a semi-open queueing network (SOQN) to estimate system performance, modeling the battery charging process as a single queue and the battery swapping process as a nested SOQN. We develop a decomposition method to solve the analytical models and validate them through simulation. Our models can be used to optimize battery recovery strategies and compare their cost and throughput time performance. The results show that throughput time performance can be significantly affected by the battery recovery policy, that inductive charging performs best, and that battery swapping outperforms plug-in charging by as large as 4.88%, in terms of retrieval transaction throughput time. However, the annual cost of the RMFS using the battery swapping strategy is generally higher than that of the RMFS using the plug-in charging strategy. In the RMFS that uses the inductive charging strategy, a critical price of a robot can be found, for a lower robot price and a small required retrieval transaction throughput time, inductive charging outperforms both plug-in charging and battery swapping strategies in terms of annual cost. We also find that ignoring the battery recovery will underestimate the number of robots required and the system cost for more than 15%.