See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319855564

Scheduling of truck arrivals, truck departures and shop-floor operation in a

cross-dock platform, based on trucks loading plans

Article in International Journal of Production Economics · September 2017

DOI: 10.1016/j.ijpe.2017.09.008

CITATIONS READS

54 2,864

3 authors, including:

Xavier Delorme Alexandre Dolgui

Mines Saint-Etienne IMT Atlantique
139 PUBLICATIONS 1,658 CITATIONS 933 PUBLICATIONS 23,120 CITATIONS

SEE PROFILE SEE PROFILE

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

International Journal of Production Research View project

All content following this page was uploaded by Alexandre Dolgui on 31 October 2017.

The user has requested enhancement of the downloaded file.

 Scheduling of truck arrivals, truck departures and shop-floor operation in

a cross-dock platform, based on outbound trucks loading plan

Christian Serrano 1, 2, Xavier Delorme 1, Alexandre Dolgui 3

1
Ecole des Mines de Saint-Etienne, LIMOS, CNRS, France

42023 Saint-Étienne, France

christian.serrano@emse.fr

delorme@emse.fr
3
Renault, Technocentre, Paris, France
3
IMT Atlantique, LS2N, CNRS

44300 Nantes, France

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

in the given industrial context.

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.

Warehouses allow companies to make economies in transportation (by making full

truckloads), reduce the effect of demand variability and execute additional operations closer

to final customers (Dolgui and Proth, 2010).

According to Napolitano (2000), warehousing is evolving from acting as a passive storage

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

Services 2011; Ertek 2005).

Cross-dock shop-floor

Inbound Outbound

Unloading - Treatment - Loading

Figure 1. Cross-dock shop-floor generic flows

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

crossdocking distribution network, as well as the associated vehicle routing problems. At

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

latter subjects. We propose a mixed-integer linear programming model to schedule inbound

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

instances. Finally, we address both conclusions and perspectives.

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

Decrivez votre cas d’abord.

Deplacer devant la section 4, puis fusionner la section 2 et 3 en reduisant au maximum la
section 2, la section 2 actuelle toute seule n’a aucun sens)

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.

At strategic level, main parameters are:

 Number of doors: defines the number of unloading and loading docks available at

the crossdock.

 Internal transportation: products can be moved on an automated way (conveyor

belts), manually (operators, fork-lift trucks) or a combination of both (combined).

 Repackaging: yes if some or all product need a repacking operation at the cross-

dock, no otherwise.

At tactical level, a cross-dock can be characterised with the following parameters:

 Product packaging: if products use different type of packages, this parameter is

considered as diversified. Otherwise, the product packaging is considered as

equivalent.

 Service mode: if dock doors perform a unique operation (loading or unloading)

the service mode is considered as exclusive. If dock doors can be used indistinctly

for both operations, service mode is combined.

 Pre-emption: yes if trucks’ loading or unloading operations can be interrupted, no

otherwise.

 Storage capacity: it is related to the space to store the products during its transit

through the crossdock. It can be zero, limited or unlimited.

 Handling capacity: can be associated to a number of packages per time period or a

number of operators/fork-lifts available at the crossdock. In those cases, the

handling capacity is limited. If there is no restriction, it is considered as unlimited.

Finally, operational characteristics are listed below:

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

defined as a dimensional capacity, linked to volume or weight. Finally, if for each

truck there is a list of products to be loaded, capacity is considered as not relevant.

 Product allocation: if only the destination is known for a given product, product

allocation mode is destination. If there is also the information on the exact

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

trucks is defined by a number of product per type, allocation mode is defined as

post-distribution.

 Interchangeability: a product is interchangeable (yes) if it can replace another of

product of the same type in a truck. This applies to the allocation modes of

destination and post-distribution. For the mode pre-distribution interchangeability

is not allowed (no).

 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

departure times are defined per truck.

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)

consider an unlimited handling capacity, which represents a strong simplification regarding

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

high diversity of types of packaging. The post-destination mode is considered in 11 papers

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

Table 1. Articles on crossdock truck scheduling (part 1).

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)

Arrival / ILP, constraint

Fazel Zarandi N Post-D
1/1 Excl. N Y Z U ND NR Per truck Y departure times, programming,
et al. (2014) (ND) (Y)
preemption heuristics

Ladier & Alpan N Destination Storage, arrival /

>1/ >1 Excl. N N U U Equiv. Max. number Per truck Y ILP, heuristics
(2014) (ND) (Y) departure times

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

Table 2. Articles on crossdock truck scheduling (part 2).

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;

Carbone & Martino 2003; Cârstea 2013; Guerrero & Ng 2015).

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

crossdocking network of 10 crossdocking platforms, called ILN (International Logistic

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.

Figure 2. Renault ILN network.

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.

5. Problem Description and Modelling

5.1. Problem definition and assumptions

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

an outbound truck or to the outbound staging zone to be stored.

 From the arrival area, packages needing repack might be transported either to the

workshop to be repacked or to the repack staging zone to be stored.

 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

loaded on an outbound truck or to the outbound staging zone to be stored.

 From the outbound staging zone, packages can be moved to the departure area to be

loaded on an outbound truck or be kept in store.

11
 Arrival Internal Departure
Package Package treatment Package
unloading Moving, repacking, storing loading

Suppliers Customers

Inbound Outbound Outbound

Truck staging Truck
arrival zone departure
Repack
staging
zone Repack
workshop

Package flow
Truck flow

Figure 3. Studied cross-dock network flows and shop-floor operation flows.

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

A C1 1.41 0.24 Direct transfer P1

B C1 1.41 0.24 Direct transfer P1
C C1 0.06 0.03 Grouping P2
D C2 1.41 0.24 Direct transfer P3
E C3 0.74 0.45 Repackaging P4
F C3 1.11 0.53 Direct transfer P5
G C3 0.74 0.45 Repackaging P4

Table 3. Illustration of component aggregation process.

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.

(planning horizon here is a week ? or a day ? please precise)

5.2. Model development and formulation

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

to an unplanned increase/decrease of workload and the possibility to overload the shop-floor

activities when the workload is bigger than its capacity, for example.

13
Our MILP model uses the following notation:

Sets

a є A Package handling at shop-floor (1: direct transfer, 2: grouping, 3: repackaging)

c є C Components.

iєI Type of products.

k є K Destinations (customers).

lєL Sources (inbound trips).

m є M Inbound trucks.

n є N Outbound trucks.

tєT Time periods (hours).

Component parameters

qic,l Quantity of component c delivered on inbound trip l.

qoc,k Quantity of component c to be sent to destination k.

tpc,i 1 if component c belongs to product type i, 0 otherwise.

Product parameters

vi Volume (m3) of product i.

mc Space occupied by component c in a truck (in m3).

ri,a 1 if product i is concerned by treatment a, 0 otherwise.

Inbound parameters

vit Maximal capacity of an inbound truck (in m3).

itlm,l 1 if inbound truck m belongs to source l, 0 otherwise.

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.

ind Number of inbound doors at the crossdock.

wit Fixed treatment time of an incoming truck.

M Big number.

Internal parameters

wun Truck unloading time (per m3).

wst Storage operation time (per m3).

wrea Operation time for treatment a.

wlo Truck loading time (per m3)

sm Maximal capacity of storage staging zones (m3 per time period).

wf Number of workers available at crossdock.

Outbound parameters

vot Maximal capacity of an outbound truck (in m3).

otdn,k 1 if outbound truck n goes to destination k, 0 otherwise. ∑ k otdn,k = 1 ∀n.

oud Number of outbound doors at the crossdock.

wot Fixed treatment time of an incoming truck.

Decision variables

ITm,t 1 if inbound truck m arrives in period t, 0 otherwise.

Fm 1 if inbound truck m arrives in a contracted arrival period, 0 otherwise.

X1c,m Quantity of component c delivered on inbound truck m.

15
X2i,m,t Quantity of product i delivered on inbound truck m in period t.

SRi,t Quantity of product i stored in period t (before repackaging).

SEi,t Quantity of product i stored in period t (before shipping).

Ri,t Quantity of product i at repackaging workshop in period t.

Y1c,n Quantity of component c shipped on outbound truck n.

Y2i,n,t Quantity of product i shipped on outbound truck n in period t.

OTn,t 1 if outbound truck departs in period t, 0 otherwise.

Figure 4 summarizes the proposed framework:

Arrival Internal Departure

Package Package treatment Package
unloading Moving, repacking, storing loading

Outbound
Inbound Outbound
staging
Truck Truck
Zone
arrival departure
Repack SE
IT OT
staging Repack
F Y1, Y2
Zone Workshop
X1, X2
SR R

Figure 4. Decision variables of the proposed model.

The mathematical formulation is presented below:

Minimize cpc * ∑ m (1 - Fm) (1)

Subject to:

Fm ≤ ∑ l,t (iltm,l * ITm,t) * trdl,t ∀m (2)

∑ t ITm,t = 1 ∀m (3)

∑ m ITm,t ≤ ind ∀t (4)

16
X2i,m,t <= M * ITm,t ∀i,m,t (5)

X2i,m,t <= ∑ c (X1c,m * tpc,i) ∀i,m,t (6)

X2i,m,t >= ∑ c (X1c,m * tpc,i) – M * ITm,t ∀i,m,t (7)

∑ m X1c,m * itlm,l = qic,l ∀c,l (8)

∑ i,t (X2i,m,t * mi) <= vit ∀m (9)

Ri,t = (∑ m X2i,m,t * ∑ a’ = [2,3] ri,a’) – SRi,t ∀i,t=1 (10)

Ri,t = (∑ m X2i,m,t * ∑ a’ = [2,3] ri,a’) – SRi,t + SRi,t-1 ∀i,t>1 (11)

∑ t Ri,t = ∑ c,l qic,l * tpc,i * ∑ a’ = [2,3] ri,a’ ∀i (12)

SEi,t = ∑ m X2i,m,t – Ri,t – ∑ n Y2i,n,t ∀i,t=1 (13)

SEi,t = SEi,t-1 + Ri,t-1 + ∑ m X2i,m,t – Ri,t – ∑ n Y2i,n,t ∀i,t>1 (14)

∑ i (SEi,t + SRi,t)* vi ≤ sm ∀t (15)

∑ i Ri,t * vi ≤ rep ∀t (16)

∑ m (ITm,t * wit) + ∑ i [vi * (∑ m (X2i,m,t * wun) + ∑ a (ri,a * Ri,t * wrea))

+ (vi * (SEi,t + SRi,t) * wst + ∑ n (Y2i,n,t * wlo))]

+ ∑ k (OTk,t * wot) ≤ wf * wd ∀t (17)

∑ t OTn,t = 1 ∀n (18)

∑ n OTn,t ≤ oud ∀t (19)

Y2i,n,t <= M * OTn,t ∀i,n,t (20)

Y2i,n,t <= ∑ c (Y1c,k * tpc,i) ∀i,n,t (21)

Y2i,n,t >= ∑ c (Y1c,k * tpc,i) – M * OTn,t ∀i,n,t (22)

∑ n Y1c,n * otdn,k = qoc,k ∀c,k (23)

∑ i,t (Y2i,n,t * mi) <= vot ∀n (24)

X1c,m ; X2i,m,t ∈ N ∀c,i,m,t (25)

17
Y1c,n ; Y2i,n,t ∈ N ∀c,i,n,t (26)

SEi,t ; SRi,t ; Ri,t ; Fm ∈ N ∀i,m,t (27)

ITm,t ; OTn,t ∈ {0,1} ∀m,n,t (28)

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,

equations (25) to (28) are integer constraints.

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

8GB RAM AMD A8-4500M @ 1.9GHz CPU.

6.1. Input data and instances

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

departure activities are shown in Table 5.

Values for Values for

Parameter Description
crossdock # 1 crossdock # 2
vit Inbound trucks capacity 13,2 [m] 13,0 [m]
ind Inbound doors 7 [doors] 10 [doors]
sm Staging zones capacities 125 [m3 / hour] 200 [m3 / hour]
rep Repackaging workshop capacity 20 [m3 / hour] 100 [m3 / hour]
wf Handling capacity 22 to 35 [workers] 60 to 75 [workers]
oud Outbound doors 5 [doors] 7 [doors]
vot Outbound trucks capacity 12 [m] 12 [m]
cpc Penalty cost on inbound trucks arrival time 40 30

Table 4: Characterization of the studied crossdocks.

Parameter Description Values for both crossdocks

wit Truck arrival 17,0 [min]
wun Package unloading 3,4 [min/m3]
wst Package storage 0,7 [min/m3]
wrea Package handling 0 / 18,7 / 53,1 [min/m3]
wlo Package loading 2,6 [min/m3]

19
 wot Truck departure 20,0 [min]
wd Total working time 420-480 [min]

Table 5: Operational times at studied crossdocks.

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,

around 24 trucks are loaded each day, to serve 4 to 6 destinations.

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 6: Instances for the numerical experiments.

6.2. Result analysis

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.

As explained in Section 4, currently there is no scheduling process done at Renault ILN, so

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

Table 7. Numerical experiments’ results.

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

departures at the last period.

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)

Staging areas Departure area

150 6 5
119.1 125.0 124.8 112.9
Number of trucks

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

Time periods (hours) Time periods (hours)

Figure 5. Detailed results for one instance of numerical experiments.

7. Evaluation of resultants and perspectives

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

loading plans, showing how to stack all packages.

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

activity much more complex.

25
 Crossdock # 1 - Arrival Crossdock # 1 - Departure
30 30

Number of outbound trucks

Nuùber of inbound trucks

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

CPLEX Cargo loading software CPLEX Cargo loading software

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

avoid the approximation errors, as explained before.

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

operation scheduling problem at a cross-dock. However, few authors present an

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

some components to adapt their packages to maritime transportation constraints. Therefore, a

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

the MILP model, which will need some minor adaptations.

References

Alpan, G. et al., 2011. Heuristic solutions for transshipment problems in a multiple door cross

docking warehouse. Computers and Industrial Engineering, 61(2), pp.402–408.

Alvarez-Perez, G.A., Gonzalez-Velarde, J.L. & Fowler, J.W., 2008. Crossdocking-Just in

Time scheduling: an alternative solution approach. Journal of the Operational Research

Society, 60(4), pp.554–564.

Bellanger, A., Hanafi, S. & Wilbaut, C., 2013. Three-stage hybrid-flowshop model for cross-

docking. Computers & Operations Research, 40(4), pp.1109–1121.

Berghman, L., Leus, R. & Lopez, P., 2012. The truck scheduling problem at crossdocking

terminals. In Conference on Project Management and Scheduling (PMS 2012). Louvain,

Belgium.

Boloori Arabani, A., Fatemi Ghomi, S.M.T. & Zandieh, M., 2011. Meta-heuristics

implementation for scheduling of trucks in a cross-docking system with temporary

storage. Expert Systems with Applications, 38(3), pp.1964–1979.

Boloori Arabani, A., Zandieh, M. & Ghomi, S.M.T.F., 2011. Multi-objective genetic-based

algorithms for a cross-docking scheduling problem. Applied Soft Computing, 11(8),

pp.4954–4970.

Boysen, N., 2010. Truck scheduling at zero-inventory cross docking terminals. Computers

and Operations Research, 37(1), pp.32–41.

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

research classification and framework. European Journal of Operational Research,

239(3), pp.593–608.

Carbone, V. & Martino, M. De, 2003. The changing role of ports in supply-chain

management: An empirical analysis. Maritime Policy & Management, 30(4), pp.305–320.

Carrera, S. et al., 2008. Negotiation models for logistic plarform planning and scheduling. In

11th International Workshop on Project Management and Scheduling. Istanbul, Turkey,

pp. 43–46.

Cârstea, V., 2013. Delocalization - the automotive industry’s answer to cost reduction.

Romanian Economic Business Review, 8(3.1), pp.180–183.

Chen, F. & Lee, C.Y., 2009. Minimizing the makespan in a two-machine cross-docking flow

shop problem. European Journal of Operational Research, 193(1), pp.59–72.

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 &

supply chain congress. Istanbul, Turkey.

Fanti, M.P., Stecco, G. & Ukovich, W., 2016. Scheduling internal operations in post-

distribution cross docking systems. IEEE Transactions on Automation Science and

Engineering, 13(1), pp.296–312.

Fazel Zarandi, M.H., Khorshidian, H. & Akbarpour Shirazi, M., 2014. A constraint

programming model for the scheduling of JIT cross-docking systems with preemption.

Journal of Intelligent Manufacturing, 27(2), pp.297–313.

Ghobadian, E. et al., 2012. Scheduling trucks in cross docking systems with temporary

storage and dock repeat truck holding pattern using GRASP method. International

Journal of Industrial Engineering Computations, 3(5), pp.777–786.

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

hybrid metaheuristic for truck scheduling in cross-dock problems. Computers and

Industrial Engineering, 91, pp.197–204.

Ladier, A.-L. & Alpan, G., 2014. Crossdock truck scheduling with time windows: earliness,

tardiness and storage policies. Journal of Intelligent Manufacturing, pp.1–15.

Ladier, A.-L. & Alpan, G., 2016. Cross-docking operations: Current research versus industry

practice. Omega, 62, pp.145–162.

Larbi, R. et al., 2011. Scheduling cross docking operations under full, partial and no

information on inbound arrivals. Computers and Operations Research, 38(6), pp.889–

900.

Li, Y., Lim, A. & Rodrigues, B., 2004. Crossdocking—JIT scheduling with time windows.

Journal of the Operational Research Society, 55(12), pp.1342–1351.

Maknoon, M.Y. & Baptiste, P., 2010. Moving freight inside cross docking terminals. In 8th

International Conference on Supply Chain Management and Information Systems. Hong

Kong.

Maknoon, M.Y., Soumis, F. & Baptiste, P., 2017. An integer programming approach to

scheduling the transshipment of products at cross-docks in less-than-truckload industries.

Computers & Operations Research, 82, pp.167–179.

Maknoon, Y. & Laporte, G., 2017. Vehicle routing with cross-dock selection. Computers and

Operations Research, 77, pp.254–266.

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.

Saddle Creek Logistics Services, 2011. 2011 Cross-docking trends report.

Sadykov, R., 2012. Scheduling incoming and outgoing trucks at cross docking terminals to

minimize the storage cost. Annals of Operations Research, 201(1), pp.423–440.

Serrano, C., Delorme, X. & Dolgui, A., 2015. Distribution and operation planning at a cross-

dock platform: A case of study at Renault. In Proceedings of 4th IEEE International

Conference on Advanced Logistics and Transport. Valenciennes, France.

Vahdani, B. & Zandieh, M., 2010. Scheduling trucks in cross-docking systems: Robust meta-

heuristics. Computers & Industrial Engineering, 58(1), pp.12–24.

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

systems with temporary storage. European Journal of Operational Research, 184(1),

pp.377–396.

31

View publication stats