Professional Documents
Culture Documents
Scheduling of Truck Arrivals, Truck Departures and Shop-Floor Operation in A Cross-Dock Platform
Scheduling of Truck Arrivals, Truck Departures and Shop-Floor Operation in A Cross-Dock Platform
net/publication/319855564
CITATIONS READS
54 2,864
3 authors, including:
Some of the authors of this publication are also working on these related projects:
ASSISTANT - LeArning and robuSt deciSIon SupporT systems for agile mANufacTuring environments, H2020 – ICT-38-2020, Artificial intelligence for manufacturing View
project
All content following this page was uploaded by Alexandre Dolgui on 31 October 2017.
christian.serrano@emse.fr
delorme@emse.fr
3
Renault, Technocentre, Paris, France
3
IMT Atlantique, LS2N, CNRS
alexandre.dolgui@mines-nantes.fr
Abstract
The extended international development of the automotive industry has been supported in a part by crossdocking
platforms set up between distant manufacturing plants and nearby first-tier suppliers. A cross-dock centre is an
intermediate consolidation point in a supply chain linking a set of suppliers with a set of customers. Main
advantages are a reduced lead time, a decrease of stock level and economies in transportation. At shop-floor,
packages from inbound trucks are unloaded, moved across, sorted by destination and loaded onto outbound
trucks. A case of study at Renault is presented in this paper. Two important points are observed for this Renault
crossdock center: (1) repacking operations are performed for some products and (2) due to the high diversity of
packages’ dimensions and stacking rules, scheduling of inbound, repackaging and loading outbound trucks is a
complex problem. We propose a mixed integer linear programming model to schedule inbound trucks’ arrival
times (considering given soft time windows), shop-floor repackaging operations and outbound trucks’ departure
times. Capacitated temporary storage zones and a capacitated repack workshop are considered at crossdock
shop-floor. The model seeks to minimize penalty costs related to inbound trucks’ arrival times and consequently
unbalanced workload of the repack workshop. Implemented in CPLEX and tested on small and medium size
instances provided by Renault, results of numerical experiments show the efficiency and pertinence of the model
1
Keywords: Automotive industry, Supply chain management, Cross-dock scheduling, Mixed integer linear
program.
1. Introduction
On its route toward the final customer, almost every product goes through a warehouse.
truckloads), reduce the effect of demand variability and execute additional operations closer
centre to being a reactive facility distributing goods in the shortest possible time at a
minimum cost. The latter description refers to the crossdocking strategy. Figure 1 shows the
common flows within a cross-dock centre, where packages are unloaded from inbound trucks,
moved across, sorted by destination and finally loaded onto outbound trucks. Usually, there is
neither permanent storage (less than 24h) nor added-value activities (e.g. repacking).
Companies of all kind of industries are setting up cross-dock centres in order to improve
service level, reduce transportation costs and accelerate logistic flows (Saddle Creek Logistics
Cross-dock shop-floor
Inbound Outbound
Crossdocking-related problems have received a lot of academic attention in the last two
decades. Van Belle et al. (2012) present a state of the art on the subject and propose a
2
decisional-level classification of the current literature. Geographical location and shop-floor
layout are the most common aspects treated at strategic level. Tactical decisions are related to
operational level, research work is focused on truck scheduling, dock-door assignment and
shop-floor operation. Main decisions linked to truck scheduling problems are related to the
arrival and departure times and the allocation of products to trucks, respecting shop-floor
constraints such as the available resources and storage capacity. Our paper is focused on the
truck arrivals, shop-floor operations and outbound truck departures at a cross-dock platform
with two temporary capacitated staging zones and a capacitated repack workshop (needed for
some products).
The rest of the paper is organized as follows: in Section 2, we present a framework of this
study to characterise the functioning mode of the studied Renault cross-dock centre. Based on
that, Section 3 presents a literature review on truck scheduling within crossdocking. Next, in
Section 4, the industrial case of study is described in details. Section 5 presents the problem
description and modelling, as well as the numerical experiments conducted based on real-data
2. Crossdock configuration
Je pense qu’il faut deplacer devant la section 2 et 4 la section 4, puis fusionner les sections 2
et 3
3
Before to present the state of the art analysis, we describe the classification proposed in Buijs
et al. (2014) and Ladier & Alpan (2016), we present below the main parameters that define
the functioning mode of a cross-dock centre. They are grouped by decisional level.
Number of doors: defines the number of unloading and loading docks available at
the crossdock.
Repackaging: yes if some or all product need a repacking operation at the cross-
dock, no otherwise.
equivalent.
the service mode is considered as exclusive. If dock doors can be used indistinctly
otherwise.
Storage capacity: it is related to the space to store the products during its transit
4
Truck capacity: it can be defined as the maximum number of products to be loaded
in a truck, which means that all products are packaged identically. It can also be
Product allocation: if only the destination is known for a given product, product
outbound truck in which the product must be loaded, the allocation mode is pre-
distribution. Finally, if the destination is given and the content of the outbound
post-distribution.
product of the same type in a truck. This applies to the allocation modes of
Arrival and departure times: if all trucks are available at the beginning of the
planning horizon these parameters take the value of zero. Otherwise, arrival and
Ce sont les definitions de la literature, où est la description de notre cas chez Renault.
3. Literature review
We analysed a total of 23 articles treating the truck scheduling problem and we characterised
them based on the description presented on Section 2, as shown in Table 1 and Table 2. The
cost drivers and the solution method are also presented. For a comparison purposes, the first
line of each table shows the configuration of our problem treated in this paper. Over the 23
5
articles analysed here, more than half consider mono-doors environments (1 input and output
door). One single article makes the assumption of a combined service mode. Only 3 articles
present industrial cases of study and two of them use real data in experiments. Repackaging is
considered in two papers. Pre-emption is permitted only in three articles. Regarding the
internal capacity, zero storage policy is considered in five papers. This constraint is often
associated with the food industry or automated crossdock centres. Two articles use a limited
capacity and the rest of them (16) deal with an unlimited capacity. Most of the articles (21/23)
our industrial case. Nine articles deal with equivalent packages and four articles with
diversified packages. For the rest (10 papers), this parameter is not clearly specified. In terms
of product allocation, 5 articles consider the destination mode, which is associated with trucks’
capacity related to capacity of trucks in maximum number of products. Only one article deals
with a dimensional capacity, which is a very important assumption in our study, given the
and the rest of them (7) use the pre-destination mode. In most of the articles (16/23), trucks
are available at time zero and only 9 of 23 include constraints on trucks departure times.
Present in 11 papers, the makespan minimization is the most common objective. Storage level
minimisation is included in 8 articles. Some other cost drivers are penalties on arrival (or
departure) times as well as the activity smoothing. In terms of solution methods, the most
common approaches are heuristics (18 papers) and linear programming (12 papers).
6
Industrial Doors Product Arrival Departure
Research Service Handling Product Truck
data inbound/ Repackaging Pre-emption Storage allocation times times Cost drivers Solution method
work mode capacity packaging capacity
(Industry) outbound (interch.) (inbound) (outbound)
Machine
N Pre-D Arrival / departure
Li et al. (2004) >1/ >1 Excl. N N Z L ND NR Per truck Y scheduling,
(ND) (N) times
heuristics, ILP
ILP, complete
Yu & Egbelu N Post-D
1/1 Excl. N N U U ND NR Zero N Makespan enumeration,
(2008) (ND) (Y)
heuristics.
Storage, activity
Carrera et al. Y Pre-D
NA ND N N L L Diver. Dimensional Per truck Y smoothing, arrival / MILP
(2008) (Footwear) (N)
departure times
Dynamic
Boysen et al. N Post-D
1/1 Excl. N N U U Equiv. NR Zero N Makespan programming,
(2008) (ND) (Y)
heuristics
Machine
Alvarez-Perez N Pre-D Arrival / departure
>1/ >1 ND N N U U ND NR Per truck Y scheduling, meta-
et al. (2008) (ND) (N) times
heuristics
Chen & Lee N Pre-D
1/1 Excl. N N Z U ND NR Zero N Makespan Heuristics
(2009) (ND) (N)
Storage, unloading /
Dynamic
N Pre-D unloading times,
Boysen (2010) >1/ >1 Excl. N N Z U Equiv. NR Zero Y programming,
(Food) (N) arrival / departure
heuristics
times
Maknoon &
N Destination Max.
Baptiste >1/ >1 Excl. N N U U Equiv. Per truck Y Storage MILP
(ND) (Y) number
(2010)
Vahdani &
N Post-D
Zandieh 1/1 Excl. N N U U Diver. NR Zero N Makespan Meta-heuristics
(ND) (O)
(2010)
Dynamic
Alpan et al N Destination Max. Storage,
>1/ >1 Excl. N Y U U Equiv. Zero N programming,
(2011) (ND) (Y) number pre-emption
heuristics
Boloori
Arabani, N Post-D Makespan, arrival /
1/1 Excl. Y N Z U ND NR Zero Y Heuristics
Zandieh, et al. (ND) (O) departure times
(2011)
Excl.: exclusive / Comb.: combined / N: no / Y: yes / U: unlimited / L: limited / Z: zero / Diver.: diversified / Equiv.: equivalent / Interch.: interchangeability / D: distribution
7
Industrial Doors Product Arrival Departure
Service Pre- Handling Product Truck
Research work data inbound/ Repackaging Storage allocation times times Cost drivers Solution method
mode emption capacity packaging capacity
(Industry) outbound (interch.) (inbound) (outbound)
Y
This paper >1/ >1 Excl. Y N L L Diver. Destination Dimensional Per truck N Arrival times MILP
(Automotive)
Boloori Makespan,
N Post-D IPL, meta-
Arabani et al. 1/1 Excl. N N U U ND NR Zero N activity
(ND) (Y) heuristics
(2011) smoothing
Polynomial
Larbi et al. N Destination Storage,
1/1 Excl. N Y U U Equiv. Max. number Zero N algorithm,
(2011) (ND) (Y) preemption
heuristics
Ghobadian et N Post-D
1/1 Excl. N N U U Diver. NR Zero N Makespan Meta-heuristics
al. (2012) (ND) (Y)
Machine
Berghman et al. N Pre-D
>1/ >1 Comb. N N U U Equiv. NR Per truck Y Makespan scheduling,
(2012) (ND) (N)
Branch-and-bound
N Post-D Dynamic
Sadykov (2012) 1/1 Excl. N N L U ND NR Zero N Storage
(ND) (Y) programming
Bellanger et al. N Pre-D
>1/ >1 Excl. N N Z U ND NR Zero N Makespan Heuristics
(2013) (ND) (N)
Mohtashami N Post-D
1/1 Excl. N N U U Equiv. NR Zero N Makespan Heuristics
(2015) (ND) (Y)
Fanti et al. Y Post-D
>1/>1 Excl. Y N U U Diver. NR Zero N Makespan ILP, heuristics
(2016) (Textile) (Y)
Keshtzari et al. N Post-D MILP, meta-
1/1 Excl. N N Z U ND NR Zero N Makespan
(2016) (ND) (Y) heuristics
Maknoon et al. N Destination
1/1 Excl. N N U U Equiv. Max. number Zero N Storage ILP, heuristics
(2017) (ND) (Y)
Excl.: exclusive / Comb.: combined / N: no / Y: yes / U: unlimited / L: limited / Z: zero / Diver.: diversified / Equiv.: equivalent / Interch.: interchangeability / D: distribution
8
4. Industrial Framework
Major carmakers have set up industrial facilities outside their home continents, resulting on a
strong growth of overseas flows, especially on the export of individual components. The latter
demands an optimized management of global sourcing and several car manufacturers use
cross-docking centres to consolidate and deliver parts to all over the world (Witt 1998;
This study was done for Renault Nissan automotive group. 11 out of its 20 vehicle assembly
plants of Renault are located outside Western Europe and accounted for 56% of total vehicle
production. In contrast, mechanical parts (e.g. engines, powertrains) and other OEM1 parts
were manufactured mostly in Western Europe. In this context, Renault has set up a worldwide
Network). Figure 2 illustrates the supply chain network of one of these platforms.
An AILN acts as an intermediate point for both information and physical flows. Customer
demands are integrated and serve as the main input for the distribution planning process, in
which the inbound transportation, the internal activity at the ILN and the outbound
transportation are planned. This planning is done once a week for the daily activities. Based
on this, delivery orders are generated and sent to suppliers. In terms of physical flows,
components are unloaded from incoming trucks, then sorted by destination, repacked (for
about 20-30% of the total volume), moved across or stored and finally loaded onto outgoing
containers. There are mainly two types of repackaging activities. In the first one, the package
is opened and the parts are placed on a different package (this is needed to respect maritime
transportation constraints). The second type consists on taking small packages and putting
them in a bigger package (done to optimize container filling ratio). Another feature of Renault
ILN is related to the high diversity (in terms of dimensions and weight) of the types of
1
OEM: Original Equipment Manufacturer.
9
packages treated. Moreover, the staking rules are complex and therefore the truck (and
container) loading activities are a critical point at Renault ILN. Currently, there is no software
to support this operation and the choice of how to load the containers is done exclusively by
the operators.
Despite the complexity of the internal operation at Renault ILN, currently there is only a daily
planning process. Shop-floor activities (hourly) are guided exclusively by operators’ choices,
based on available parts at crossdock. In this context, we propose a model for scheduling
inbound trucks arrival times, internal operations (including storing and repacking) and
outbound trucks departure times. We complete our previous research work in Serrano et al.
(2015), where has been proposed an approach to improve the distribution and operation
planning processes at Renault ILN, using an integer linear programming model to minimize
transportation costs (inbound and outbound) and internal costs (storage and internal resources).
In the previously published model, the main model’s output are the approximations of the
number of inbound (and outbound) trucks to be used, the internal resources needed (workers
and storage) to fulfil shop-floor operation, as well as the component flow within the network.
These results are the main input for the scheduling model considered in this paper. In the next
10
section we present the main assumptions as well as the mathematical formulation of the new
model.
We propose a model to schedule truck arrival, shop-floor operation and truck departure in a
cross-dock centre for a given horizon (e.g. a day). Storage is allowed and a repacking activity
is needed for some products. Two temporary staging capacitated zones are considered: one at
the repack workshop and another one on the departure area. Figure 3 shows the studied cross-
dock configuration. Once an inbound truck arrives at one of the cross-dock doors, all
packages must be unloaded on the arrival area and moved across. From the arrival area,
packages that do not need repack can go either directly to the departure area to be loaded on
From the arrival area, packages needing repack might be transported either to the
From the repack staging zone, packages can go to the workshop to be repacked or be
kept in store.
From the repack workshop, packages can be moved either to the departure area to be
From the outbound staging zone, packages can be moved to the departure area to be
11
Arrival Internal Departure
Package Package treatment Package
unloading Moving, repacking, storing loading
Suppliers Customers
Package flow
Truck flow
The planning horizon represent one working day, divided in time periods of one hour.
Customer demands for components are known and must be respected by the end of the
planning horizon. Each component is produced by a single supplier, and this data is known.
Other information given for each component is: the reference, the dimensions of the
packaging (and thus the volume in cubic meters), the occupied space in a truck (in meters),
the type of handling given at shop-floor (direct transfer, grouping or repackaging) and the
corresponding product type. A given product type includes all the components coming from
the same source (supplier), occupying the same place in a truck, with the same volume (m3)
and with the same handling operation at shop-floor. Components with these four common
characteristics are interchangeable in the definition of the contents of the trucks (same source
and same occupied space), as well in the assessment of workload linked to the shop-floor
handling operation (same m3 and same operation). Table 3 shows several examples of this
process of aggregation.
12
Handling Product
Component Source Volume (m3) Occupied space (m)
operation type
The number of incoming trucks per source and the number of outgoing trucks per destination
are known, as well as their maximum capacity (in meters). The numbers of doors (for arrival
and for departure) are given as input data, as well as all the operational times for handling
activities. Finally, on the inbound part, contracted arrival times are modelled as soft
constraints and a penalty cost is associated if they are not respected. On the outbound part, all
trucks must leave the crossdock by the end of the planning horizon.
A mixed integer linear program (MILP) is proposed to schedule truck arrivals and shop-floor
operations at the cross-dock centre. It seeks to determine inbound trucks arrival time, internal
flows within shop-floor areas and outbound trucks departure time and content. The objective
is to minimize the inbound cost related to a penalty associated to the non-respect of pre-
contracted arrival times. This is because the non-respect of pre-contracted arrival times leads
activities when the workload is bigger than its capacity, for example.
13
Our MILP model uses the following notation:
Sets
c є C Components.
k є K Destinations (customers).
m є M Inbound trucks.
n є N Outbound trucks.
Component parameters
Product parameters
Inbound parameters
14
trdl,t If arrival time t is part of the contracted arrival periods of trucks coming from source l, 0
otherwise.
cpc Penalty cost of planning an inbound truck arrival time outside the contracted arrival
times.
M Big number.
Internal parameters
Outbound parameters
Decision variables
15
X2i,m,t Quantity of product i delivered on inbound truck m in period t.
Outbound
Inbound Outbound
staging
Truck Truck
Zone
arrival departure
Repack SE
IT OT
staging Repack
F Y1, Y2
Zone Workshop
X1, X2
SR R
Subject to:
∑ t ITm,t = 1 ∀m (3)
16
X2i,m,t <= M * ITm,t ∀i,m,t (5)
∑ t OTn,t = 1 ∀n (18)
17
Y1c,n ; Y2i,n,t ∈ N ∀c,i,n,t (26)
The objective function in (1) seeks to minimize the penalty cost associated to inbound trucks
arrival times. Inbound constraints are represented by equations (2) to (9). Equation (2)
describes, for each inbound truck, if the arrival time corresponds (or not) to a contracted
period. Constraint (3) guarantees that all inbound trucks arrive to the crossdock. The respect
of the number of inbound doors is ensured in (4). Equations (5) to (7) guarantee the coherence
between component delivery (X1), inbound truck arrival (IT) and the aggregation of
components per type of product (X2). Equation (8) guarantees the delivery of all components.
Constraint (9) is related to inbound trucks capacity. Internal activity constraints are described
by equations (10) to (17). Constraints (10) and (11) represent the flow between arrival zone,
repackaging staging zone and repackaging workshop. Equation (12) ensures that all
components needing repackaging pass through the workshop. Constraints (13) and (14) seek
to balance component flow and component storage. Equations (15) to (17) represent,
respectively, the staging zones maximal capacity, repackaging workshop maximal capacity
and internal resources maximal capacity (related to number of workers). Constraint (18)
guarantees that all outbound trucks depart from the crossdock. The respect of the number of
outbound doors is ensured in (19). Equations (20) to (22) guarantee the coherence between
component shipping (Y1), outbound truck departure (OT) and the aggregation of components
per type of product (Y2). Equation (23) ensures the respect of customer demands. Finally,
18
6. Numerical Experiments
In this section, we present the data, the instance generation process as well as the results of
numerical experiments. The proposed model was implemented and tested with CPLEX on an
Input parameters were initialized based on real data from two Renault ILN. Table 4 shows the
main parameters related to trucks capacities, dock doors and shop-floor capacities for each
studied crossdock. The operational times related to arrival, unloading, handling, loading and
19
wot Truck departure 20,0 [min]
wd Total working time 420-480 [min]
Table 6 shows the size of the different instances. For the crossdock #1, the number of
components to be treated daily varies from 264 to 604, representing around 60 different types
of products. The number of inbound trucks goes from 14 to 22, coming from around 15
sources (suppliers). 9 destinations are served with around 15 trucks. In terms of treated
volumes, in the highest set the platform # 1 treats 1100 m3 and in the lowest one only 620 m3.
The repacking activity represents around 8% of the total volume. Finally, the total number of
packages varies from 1274 to 3606. The crossdock #2 has a biggest activity than the
crossdock #1: 70% in terms of volume (m3) and 45% in terms of number of packages. The
repackaging activity represents 30% of the total volume. Around 30 inbound trucks are
treated daily, coming from 13 to 27 different sources. Finally, on the outbound segment,
20
Crossdock # 1
Time periods [T] = 7
Components [C] Sources [L] / Destinations [K] Total Repackaging
Number of
Set / Type of Inbound / Outbound volume volume
packages
products [I] trucks [M] trucks [N] (m3) (m3)
1 345 / 72 17 / 22 7 / 16 1974 915.50 47.10
2 264 / 52 11 / 16 6 / 13 2198 805.02 55.02
3 439 / 70 16 / 19 7 / 16 2326 893.90 34.43
4 285 / 56 11 / 14 8 / 11 1789 620.40 76.11
5 603 / 66 11 / 16 8 / 14 3606 799.80 96.13
6 221 / 60 14 / 21 6 / 20 1030 1156.99 16.93
7 277 / 52 14 / 20 7 / 16 1274 1015.33 39.76
8 468 / 66 14 / 20 9 / 16 2393 906.33 62.67
9 282/ 58 15 / 18 8 / 12 1196 697.14 77.67
10 291 / 69 15 / 16 9 / 14 2611 734.72 100.64
Crossdock # 2
Time periods [T] = 8
Components [C] Sources [L] / Destinations [K] Total Repackaging
Number of
Set / Type of Inbound / Outbound volume volume
packages
products [I] trucks [M] trucks [N] (m3) (m3)
1 372 / 57 15 / 31 6 / 26 2146 1607.77 483.95
2 587 / 55 20 / 41 6 / 34 4398 2088.34 471.94
3 434 / 46 16 / 30 4 / 23 1883 1469.16 445.92
4 452 / 48 16 / 30 6 / 24 3146 1485.16 470.47
5 439 / 47 17 / 28 6 / 21 2674 1317.60 483.45
6 350 / 50 13 / 24 6 / 20 1758 1212.22 384.07
7 636 / 54 27 / 42 6 / 33 6541 2081.79 322.28
8 496 / 59 23 / 33 6 / 23 2704 1439.96 380.65
9 460 / 57 13 / 25 6 / 21 2759 1306.43 371.91
10 397 / 58 15 / 24 4 / 19 1732 1218.48 354.26
Table 7 shows some statistics and the results of numerical experiments. Instance size varies
between 18290 and 82038 in terms of number of variables and between 31508 and 99788 for
the number of constrains. As expected, instances from crossdock #2 have a bigger size than
those from crossdock #1. Solutions were found for all instances of crossdock # 1. For four of
21
them (2, 3, 9 and 10) the optimal solution was given by CPLEX in less than five minutes. For
instances 4, 5 and 7 the optimal solution is found in less than eight minutes. The optimal
solution is zero, since no penalty cost is generated. For the rest of instances (1, 6 and 8) the
optimal solution was not reached after thirty minutes (we consider this as the maximum
reasonable time to solve this model, since this should be done in a daily basis). We can
therefore conclude that the model responds to the crossdock #1 configuration. Concerning
instances of crossdock #2, feasible solutions were given only for instances 3 and 6. This
results shows the complexity of the problem when the size of instances grows.
we cannot compare model’s results with an as-in scenario. However, based on our field work,
we consider that the model can bring important improvements and profits at Renault ILN.
Crossdock # 1
Sets
1 2 3 4 5 6 7 8 9 10
Statistics
Variables 34062 19523 34249 18290 33562 27849 24440 35138 22086 24895
Constraints 59289 31508 58215 32149 47116 48204 39403 53366 40275 50060
Objective
40 0 0 0 0 80 0 40 0 0
fonction
CPU time (s) 1800 246 247 412 510 1800 701 1800 179 253
GAP (%) 100 ND ND ND ND 100 ND 100 ND ND
Crossdock # 2
Sets
1 2 3 4 5 6 7 8 9 10
Statistics
Variables 49051 78986 44064 46758 42307 34576 82038 56105 43897 38783
Constraints 64947 87590 52603 56924 53841 47163 99788 80608 56280 55913
Objective
ND ND 90 ND ND 120 ND ND ND ND
fonction
CPU time (s) 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800
GAP (%) ND ND 100 ND ND 100 ND ND ND ND
22
6.3. Detailed results for one instance
Figure 5 illustrates the results for one of the instances of crossdock #1. We note 22 incoming
trucks over the 7 time periods. The number of doors available for unloading is 7. At shop-
floor, the handling workload varies from 14 hours (in period 3) to 23.8 hours (in period 2).
The related maximum capacity is 24 hours (representing the 24 workers). The workload at the
repackaging area is as well showed, with a peak of activity at 14.1 hours (in period 5) and a
zero workload in period 7 (since all packages must be shipped by the end of the planning
horizon). The staging zones are almost saturated in 4 out of 7 time periods, with zero at the
last one. Finally, in the departure area, 17 trucks are loaded and shipped, with a peak of
23
Arrival area Shop-floor handling workload
8 30
23,8 23,5 22,2
Number of trucks
Number of hours
6 5
4 20 16,2 17,6 18,6
3 3 3 14,0
4
2 2
2 10
0 0
1 2 3 4 5 6 7 1 2 3 4 5 6 7
Time periods (hours) Time periods (hours)
Repackaging area
25
20
14.1
15
m3
8.6
10 6.3 6.9 6.2
5.0
5
0.0
0
1,0 2,0 3,0 4,0 5,0 6,0 7,0
Time periods (hours)
5
100 82.7 4 3 3 3
m3
3 2
42.9 2
50 1
0,0 1 0
0 0
1,0 2,0 3,0 4,0 5,0 6,0 7,0 1 2 3 4 5 6 7
Since the 1990s, an operational research team was established at Renault with specialists in
the field, with skills in logistics as well, major field of application of the OR models. One of
the related subjects concerns the constitution of packaging stacks and placement of the stacks
into the trucks. For this purpose, a heuristic algorithm with a limited depth tree exploration
was developed, as well as a tool for visualizing 2D and 3D truck loading plans (Nguyen &
Brenaut (2013) ; Sergeant (2017)). The main input for the tool consists in a list of packages to
24
be loaded, as well as the associated dimensions and weights. Considering complex staking
constraints (issued from field study at Renault ILN and plants), the tool provides the truck
We have used this truck (cargo) loading software to evaluate the results of our linear
programming model. As reminder, variables X1c,m define the quantity of component c loaded
into inbound truck m and variables Y1c,n define the quantity of component c loaded into
outbound truck m. Using this information as input for the truck loading software, Figure 6
shows the comparison between CPLEX and the truck loading software results, in terms of
number of trucks (both inbound and outbound), for the 10 instances of crossdock # 1. On
average, regarding the inbound trucks, the approximation obtained with the optimization tool
is 15% percent above CPLEX results. For the outbound trucks, this values ups to 22%. These
results show the complexity of the truck loading activity at Renault ILN, due to the high
diversity of packages’ dimensions and complex staking rules. The difference between the
inbound and the outbound segment seems logical since inbound trucks are usually coming
from few suppliers, so the packages’ diversity is less important. In contrast, the outbound
trucks are loaded grouping packages from a high number of suppliers, making the loading
25
Crossdock # 1 - Arrival Crossdock # 1 - Departure
30 30
25 25
20 20
15 15
10 10
5 5
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Sets Sets
Figure 6. Assessment of CPLEX results in trucks contents, using truck loading software.
Based on this work, Figure 7 shows our research perspective to integrate the MILP model we
propose in this paper and the cargo loading software available at Renault. The first step will
consist of generating the inbound and outbound trucks loading plan, based on the input data
listed for the MILP model. This information will be included as input for the MILP model,
which will be modified so that the information of truck contents becomes a constraint (instead
of a decision variable). This will assure the respect on the number of trucks to be used and
Figure 7. Assessment of CPLEX results in trucks contents, using cargo loading software.
26
8. Conclusions
For several years now, crossdocking has been receiving a considerable attention in both
academic and industrial fields. Several optimization models have been proposed to solve the
implementation of their work and therefore a gap between industrial practices and current
research is observed (Ladier & Alpan 2016). In this paper, we present a case study at Renault
crossdock platforms (called ILN), focused on the scheduling problem of truck arrivals, shop-
floor operations and truck departures and we propose a mixed integer linear program (MILP)
to solve it. The model seeks to define the truck contents (inbound and outbound), the arrival
and departure times and the component flow between the different areas of the shop-floor. We
consider that the inbound trucks have a pre-contracted arrival schedule and a penalty cost is
generated is it not respect. Therefore, the objective function aims to minimize the total penalty
cost. The model was implemented and tested in CPLEX, using real-data instances from two
Renault ILN. The model provides optimal solutions for most of the instances of crossdock # 1,
but none for the crossdock # 2 (bigger instances). A deeper analysis on model’s performance
is needed, which results could lead to a development of heuristics methods to solve bigger
instances.
Two following particularities taken from the case study that have not been deepened in
current literature are included in our approach. The first is a repacking activity, needed for
repackaging is modelled within the shop-floor areas. The second characteristic concerns the
high diversity of types of packages. The truck loading activity is impacted by this fact, since
the combination and the complexity of stacking is very large. To deal with this, the MILP
model includes a parameter linked to each product that defines the occupied space in a truck.
Moreover, simulations were done with a truck/cargo loading software available at Renault and
27
results showed that the approximation on the number of trucks as well as its content definition
were not totally accurate. This validated the fact that the package diversity will impact the
scheduling at crossdocks and eventually the associated costs. As a research and industrial
perspective, we propose an alternative approach combining cargo loading software results and
References
Alpan, G. et al., 2011. Heuristic solutions for transshipment problems in a multiple door cross
Bellanger, A., Hanafi, S. & Wilbaut, C., 2013. Three-stage hybrid-flowshop model for cross-
Berghman, L., Leus, R. & Lopez, P., 2012. The truck scheduling problem at crossdocking
Belgium.
Boloori Arabani, A., Fatemi Ghomi, S.M.T. & Zandieh, M., 2011. Meta-heuristics
Boloori Arabani, A., Zandieh, M. & Ghomi, S.M.T.F., 2011. Multi-objective genetic-based
pp.4954–4970.
Boysen, N., 2010. Truck scheduling at zero-inventory cross docking terminals. Computers
Boysen, N., Fliedner, M. & Scholl, A., 2008. Scheduling inbound and outbound trucks at
28
cross docking terminals. OR Spectrum, 32(1), pp.135–161.
Buijs, P., Vis, I.F.A. & Carlo, H.J., 2014. Synchronization in cross-docking networks: A
239(3), pp.593–608.
Carbone, V. & Martino, M. De, 2003. The changing role of ports in supply-chain
Carrera, S. et al., 2008. Negotiation models for logistic plarform planning and scheduling. In
pp. 43–46.
Cârstea, V., 2013. Delocalization - the automotive industry’s answer to cost reduction.
Chen, F. & Lee, C.Y., 2009. Minimizing the makespan in a two-machine cross-docking flow
Dolgui, A., Proth, J.-M., 2010. Supply chain engineering: Useful methods and techniques,
Springer.
Ertek, G., 2005. A Tutorial On Crossdocking. In Proceedings of 3rd international logistics &
Fanti, M.P., Stecco, G. & Ukovich, W., 2016. Scheduling internal operations in post-
Fazel Zarandi, M.H., Khorshidian, H. & Akbarpour Shirazi, M., 2014. A constraint
programming model for the scheduling of JIT cross-docking systems with preemption.
Ghobadian, E. et al., 2012. Scheduling trucks in cross docking systems with temporary
storage and dock repeat truck holding pattern using GRASP method. International
29
Guerrero, D. & Ng, A.K.Y., 2015. The evolving roles of ports in global supply chains: some
evidences from french car manufacturers. In IAME 2015 Conference. Kuala Lumpur,
Malasia.
Keshtzari, M., Naderi, B. & Mehdizadeh, E., 2016. An improved mathematical model and a
Ladier, A.-L. & Alpan, G., 2014. Crossdock truck scheduling with time windows: earliness,
Ladier, A.-L. & Alpan, G., 2016. Cross-docking operations: Current research versus industry
Larbi, R. et al., 2011. Scheduling cross docking operations under full, partial and no
900.
Li, Y., Lim, A. & Rodrigues, B., 2004. Crossdocking—JIT scheduling with time windows.
Maknoon, M.Y. & Baptiste, P., 2010. Moving freight inside cross docking terminals. In 8th
Kong.
Maknoon, M.Y., Soumis, F. & Baptiste, P., 2017. An integer programming approach to
Maknoon, Y. & Laporte, G., 2017. Vehicle routing with cross-dock selection. Computers and
Mohtashami, A., 2015. Scheduling trucks in cross docking systems with temporary storage
and repetitive pattern for shipping trucks. Applied Soft Computing, 36, pp.468–486.
Napolitano, M., 2000. Making the Move to Cross Docking: A Practical Guide to Planning,
30
Designing, and Implementing a Cross Dock Operation, WERC.
Sadykov, R., 2012. Scheduling incoming and outgoing trucks at cross docking terminals to
Serrano, C., Delorme, X. & Dolgui, A., 2015. Distribution and operation planning at a cross-
Vahdani, B. & Zandieh, M., 2010. Scheduling trucks in cross-docking systems: Robust meta-
Van Belle, J., Valckenaers, P. & Cattrysse, D., 2012. Cross-docking: State of the art. Omega,
40(6), pp.827–846.
Witt, C.E., 1998. Crossdocking: concepts demand choice. Material Handling, 53(7), pp.44–49.
Yu, W. & Egbelu, P.J., 2008. Scheduling of inbound and outbound trucks in cross docking
pp.377–396.
31