A dynamic hybrid Berth Allocation Problem with

routing constraints in bulk ports

Hamza Bouzekri, Gülgün Alpan, Vincent Giard

To cite this version:

Hamza Bouzekri, Gülgün Alpan, Vincent Giard. A dynamic hybrid Berth Allocation Problem with
routing constraints in bulk ports. APMS 2020, Aug 2020, Novi Sad, Serbia. pp.250-258, �10.1007/978-
3-030-57993-7_29�. �hal-02921374�

HAL Id: hal-02921374

https://hal.science/hal-02921374
Submitted on 19 Apr 2022

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 A Dynamic Hybrid Berth Allocation Problem with
Routing Constraints in Bulk Ports

Hamza Bouzekri1,2, Gülgün Alpan1,2, and Vincent Giard1,3

1 EMINES - School of Industrial Management, Mohammed VI Polytechnic University,
43150 Ben Guerir, Morocco
hamza.bouzekri@emines.um6p.ma
2 Univ. Grenoble Alpes, Grenoble INP, CNRS, G-SCOP, 38000 Grenoble, France

gulgun.alpan@grenoble-inp.fr
3 Université Paris-Dauphine, PSL Research University, 75016 Paris, France

vincent.giard@dauphine.psl.eu

Abstract. The Berth Allocation Problem (BAP) is considered as one of the

most important operational problems in the seaside area of ports. It refers to the
problem of assigning a set of vessels to a given berth layout within a given time
horizon. In this paper, we study the dynamic and hybrid case of the BAP in the
context of bulk ports with multiple quays, different water depths, and heteroge-
neous loading equipment, considering routing constraints (routes between stor-
age hangars and berths). This study is motivated by the operations of OCP
Group, a world leader in the phosphate industry, at the bulk port of Jorf Lasfar
in Morocco, recognized as the largest ore port in Africa. The objective of the
problem is to enhance the coordination between the berthing and yard activities,
besides maximizing the difference between the despatch money and the demur-
rage charges of all berthed vessels. We propose an integer linear programming
model formulated with predicates, which ensures maximum flexibility in the
implementation of the model. Finally, the proposed model is tested and
validated through numerical experiments based on instances inspired by real
bulk port data. The results show that the model can be used to solve to optimali-
ty instances with up to 40 vessels within reasonable computational time.

Keywords: Berth Allocation Problem, Conveyor System, Bulk Ports.

1 Introduction

Although containerization has played a significant role in developing the port sector
and maritime transport, bulk cargoes are still the essential and enduring trades that
support the dynamism of maritime shipping. It has to be noted that bulk port opera-
tions are very different from container port operations. Indeed, in bulk ports, it is nec-
essary to consider the cargo type and to model the interaction between the storage
locations of goods on the yard and the berthing locations of vessels. Hence, establish-
ing a set of feasible routes between berths and storage locations to guarantee that
goods are shipped on schedule when making berth allocation decisions, is critical.
2

Our analysis considers the bulk port of Jorf Lasfar where a complex conveyor sys-
tem, composed of different routes that share one or more conveyor belts, is used to
transport goods from storage hangars to berths. In addition, we consider the draft
restrictions on vessels that limit the feasible berthing positions to only those berths
having a water depth higher than their draft. To solve this problem, we propose an
integer linear programming model. The spatiotemporal constraints of the problem are
formulated as disjunctive constraints, thanks to the use of spatiotemporal binary vari-
ables. Moreover, all the conditions of the problem are expressed as predicates, which
ensures maximum flexibility in the implementation of the model and significantly
improves its computational performance. Indeed, it is no longer necessary to intro-
duce the conditions of the problem as constraints in the model, and the space search
of solutions becomes smaller.
The rest of the paper is organized as follows: Section 2 provides a literature re-
view. In Sections 3 and 4, the problem and the mathematical formulation are intro-
duced. The results of the numerical experiments are reported in Section 5. Finally,
conclusions and future research directions are addressed in Section 6.

2 Literature Review

The BAP in bulk ports has received little attention in Operations Research literature
compared to container ports. In this section, we present a brief review of past research
on the BAP in the context of bulk ports. There is a multitude of BAP formulations
depending on the spatial and temporal constraints involved in the problem. The spatial
attribute concerns the berth layout (discrete, continuous or hybrid) and the draft re-
strictions, while the temporal one includes the arrival process and the handling time of
vessels. Umang et al. [1] studied the dynamic hybrid BAP taking into account the
cargo type and the draft of each vessel. Ernst et al. [2] solved the continuous BAP
with tidal constraints that limit the departure of fully loaded vessels from the terminal.
In contrast, Barros et al. [3] solved the discrete BAP considering homogeneous berths
with tide and stock level constraints, prioritizing vessels related to the most critical
mineral stock level.
Since the problems of berth allocation and yard management are interrelated, some
authors have integrated the BAP with the Yard Assignment Problem. Indeed,
Robenek et al. [4] extended the dynamic hybrid BAP to account for the assignment of
yard locations, with the assumptions that each vessel has only one single cargo type.
To solve this integrated problem, the authors proposed an exact solution algorithm
based on a branch and price framework and a metaheuristic approach based on criti-
cal-shaking neighborhood. Unsal and Oguz [5] proposed a MILP model for an inte-
grated problem that consists of three operations: berth allocation, reclaimer (a large
machine used to recover bulk material from a stockpile) scheduling and stockyard
allocation, considering tide constraints. In the same logic of integrating problems,
Pratap et al. [6] developed a decision support system to solve the integrated problem
of berth and ship unloader allocation. Menezes et al. [7] integrated production plan-
ning and scheduling problems with a First In, First Out (FIFO) policy for berthing
vessels. This integrated problem defines the amount and destination of each input or
 3

output order between reception, stockyards and piers, establishing a set of feasible
routes between these three subsystems, to guarantee that goods are stored and shipped
on schedule and to minimize operational costs.
In our paper, we solve the dynamic and hybrid BAP under routing constraints, con-
sidering the type of cargo and the capacity limits of the equipment. The storage loca-
tions of goods are provided as input parameters to the model. To reduce the gap be-
tween the abstract representation of the studied problem and its applicability in real
situations, we consider many aspects such as draft restrictions, the heterogeneity of
equipment, Charter Party clauses and multiple cargo types on the same vessel.

3 Problem Description

We consider a bulk port with multiple quays and heterogeneous loading equipment
linked to storage hangars by a conveyor system. This latter is composed of different
routes that share one or more conveyor belts (see Fig. 1a). Each quay has as hybrid
layout where large vessels may occupy more than one berth, however, small vessels
cannot share a berth (see Fig. 1b). Each berth is characterized by a length, a fixed
loading equipment and a water depth. All the berths of a quay can have the same wa-
ter depth, or the water depth increases seaward by berths. We assume dynamic vessel
arrivals (i.e. Fixed arrival times are given for the vessels; hence, vessels cannot berth
before the expected arrival time). Each vessel is characterized by a length, a draft, a
maximum waiting time in the harbor and a number of cargo types with different
amounts to be loaded in it. These amounts of cargo types can be expressed as batches.
Each batch has an availability date and is stored in a hangar. It has to be noted that the
batches to be loaded in a single vessel can be stored in different hangars. Handling
times of vessels depend on their berthing position due to the productivity of the load-
ing equipment at the berth. We assume that two (or more) batches cannot be loaded at
the same time, but they can be loaded in any order, with no downtime. We also con-
sider technical constraints of vessels that prohibit their berthing at some berths or
oblige them to berth at a specific berth. Finally, we consider Charter Party clauses by
defining for each vessel the laytime (i.e. contractual handling time), the despatch
money (i.e. the bonus payment offered by the shipowner to the charterer if the vessel
completes loading before the laytime has expired), and the demurrage charges (i.e. the
fees paid by the charterer to the shipowner for exceeding the laytime). These contrac-
tual clauses are more detailed in Bouzekri et al. [8].

Fig. 1a. Port conveyor system. Fig. 1b. Hybrid berth layout.
4

4 Model Formulation
4.1 Notation
Table 1. Indexes, sets, input parameters and variable decisions.
Index Description
v Index of vessels V  1,..., V .
b Index of berths B  1,..., B .
t Index of time periods T  1,..., T .
p Index of pairs of berths that share a berth and cannot be used simultaneously
P  1,..., P (e.g. in Fig. 1b, the pair of berths 3 and 5 share the berth 3, so
they cannot be used at the same time).
iv Index of batches to be loaded in vessel v I v  1,..., I v  .
r Index of routes R  1,..., R . Each route links a storage hangar to a berth.
g Index of groups of routes that share at least one conveyor belt of the conveyor
system to transport batches G  1,..., G . Hence, the routes in a given group
cannot be used all at once.
Parameter Description
Lb Length of berth b.
Wb Minimum water depth of berth b.

Eb
p Boolean parameter that equals 1 if berth b belongs to the pair p of berths that
share a berth, 0 otherwise.
Av Expected time of arrival of vessel v.
Mv Maximum waiting time in the harbor of vessel v.
λv Length of vessel v.
Dv Draft of vessel v when it is fully loaded.
Nvb Boolean parameter that equals 1 if vessel v can berth at berth b, 0 otherwise.
Jv Contractual handling time of vessel v.
δv Contractual finishing time of vessel v: δv  Av  J v  1, v V .
αv Contractual demurrage by hour of vessel v.
βv Contractual despatch by hour of vessel v.

θ iv
vb
Loading time of batch iv in vessel v when this latter is berthed at berth b.
Ovb Loading time of vessel v, which equals the sum of loading times of all the
batches loaded in this vessel: O vb   θ ivbv ,  v V ,  b  B .
i I v v

i
K vv Date of availability of batch iv to be loaded in vessel v.

H vv
i
Storage hangar of batch iv to be loaded in vessel v.
Qr Index of the berth linked to route r.
Sr Index of the storage hangar linked to route r.
 5

g
Fr Boolean parameter that equals 1 if route r belongs to group g of routes that share
at least one conveyor belt of the conveyor system, 0 otherwise.
Ug Maximum number of routes that can be used simultaneously in group g of
routes.
Variable Description
xvbt 1 if vessel v starts berthing at berth b in time period t, 0 otherwise.
iv
yvbtr 1 if batch iv starts to be loaded in vessel v at berth b in time period t using route
r, 0 otherwise.
uv Integer, delay of vessel v.
wv Integer, advance of vessel v.

4.2 Mathematical Model

The existence of the decision variable xvbt is subject to four conditions:

1. Vessel v must be able to berth at berth b: Nvb  1 .

2. The length of vessel v must not exceed the length of berth b: λ v  Lb .
3. The draft of vessel v must not exceed the water depth of berth b: Dv  Wb .
4. Vessel v can berth only after its expected time of arrival without exceeding its
maximum waiting time in the harbor: Av  t  Av  Mv .
iv
The existence of the decision variable yvbtr is subject to seven conditions:

1. Conditions 1, 2 and 3 of the existence of the decision variable xvbt .

4. Batch iv can be loaded in vessel v between the expected time of arrival of this ves-
sel and its finishing time as it reaches its maximum waiting time in the harbor, mi-
nus the loading time of this batch: A v  t  A v  M v  O vb  θ ivb . v

5. Batch iv can be loaded in vessel v only after its date of availability: t  K ivv .
6. The route used to load the batch iv in vessel v must be linked to the berth b of this
vessel: Q r  b .
7. The route used to load the batch iv in vessel v must be linked to the storage hangar
of this batch: S r  H ivv .

We define the intermediate variables v and viv , which give for each vessel v, re-
iv
spectively, the berthing position in both decision variables xvbt and yvbtr .
v   bB Nvb 1 λv  Lb  Dv  Wb  tT Av  t  A v  Mv
b  xvbt , v V

vi   bB
v
N vb 1 λ v  Lb  D v  Wb  tT A v  t  A v  M v  Ovb  θvb
i
v  t  K iv  r R Q r  b  Sr  H ivv
b  yvbtr
iv
,
v

 v V ,  iv  I v
Similarly, we set for each vessel v, the berthing and finishing time  v and  v by re-
placing b  xvbt in v , respectively, by t  xvbt and  t  O vb  1  xvbt . Likewise, we
6

define for each batch iv to be loaded in vessel v, the loading start and finishing time
 viv
and  viv by replacing b  yvbtr
iv
in viv , respectively, by t  yvbtr
iv
and
t  θ iv
vb 1  y iv
vbtr .

The model for the BAP with routing constraints can be formulated as follows:

M ax  v V β v  w v  α v  uv  (1)
s.t.  bB N vb 1 λv  Lb  Dv  Wb  tT Av  t  A v  M v
xvbt  1, v V (2)

 bB N vb 1 λ v  Lb  D v  Wb  tT A v  t  A v  M v  O vb  θvb

i
v  t  K iv
v
 r R Q r  b  Sr  H ivv
iv
yvbtr  1,
 v V ,  iv  I v (3)
 v   ,  v V , iv  I v
iv
v
(4)
   v , v V , iv  I v
iv
v (5)
   v , v V , iv  I v
iv
v (6)
iv
t   Av  M v  Ovb  θ vb
   
t  t
r  1,
iv
iv  I v bB Nvb 1 λ v  Lb  Dv  Wb i
t  A v t  K ivv  t   θ vb
v 1 t rR Q r  b  Sr  H ivv
yvbt

t  T ,v V (7)
i
t  t t   A v  M v  Ovb  θvb
    
v

vV iv  I v bB N vb 1 λ v  Lb  D v  Wb t  A v t  K viv  t   θvb

i
v 1 t rR Frg 1 Q r  b  Sr  Hivv
y iv
vbt r  Ug ,

t  T ,g G (8)

 
t  t t  A v  M v
vV N vb 1 λ v  Lb  Dv  Wb t  A v t   Ovb 1 t
xvbt   1,t  T ,b B (9)

  
t  t t  A v  M v
vV bB E bp 1 Nvb 1 λ v  Lb  Dv  Wb t  Av t   Ovb 1 t
xvbt   1,t  T ,p P (10)

uv   v  δv , v V (11)
wv  δv   v , v V (12)
uv  wv   v  δv , v V (13)
uv , wv  0, v V (14)
Objective function (1) maximizes the difference between the despatch money and
the demurrage charges of each vessel v. Equation (2) ensures that each vessel v starts
berthing at a unique berth b and in a unique time period t. Equation (3) ensures that
each batch i v starts its loading in vessel v at a unique berth b, in a unique time period
t, and is transported in a unique route r. Equation (4) ensures that berth b in both deci-
i
sion variables xvtb and yvtbr
v
is the same. Equation (5) ensures that the loading of each
batch i v can only begin once vessel v has been berthed. Equation (6) ensures that
each vessel v can only leave the port when all batches have been loaded. Equation (7)
ensures that two (or more) batches cannot be loaded at the same time in each vessel v.
Equation (8) avoids simultaneous use of routes that share at least one conveyor belt of
the conveyor system. Equation (9) avoids the overlapping of vessels in each berth b.
Equation (10) ensures that only one berth can be used from each pair of berths that
 7

share a berth since the berth layout of each quay is hybrid. Equations (11)-(14) deter-
mine the delay and the advance of each vessel.

5 Numerical Experiments
The experiments were conducted using a computer with a core Intel® Xeon® CPU
E3-1240 v5 @ 3.50 GHz - 64 Go RAM, running a 64-bit version of the commercial
solver Xpress-IVE 1.24.24. The method used for solving the problem is the primal
simplex algorithm. The detailed characteristics of test instances and results can be
found at Mendeley in Bouzekri et al. [9].

5.1 Input Data

Test instances were generated based on a sample of data obtained from OCP group.
This latter operates six quays in the port of Jorf Lasfar to import raw materials (sulfur
and ammonia) and export raw materials and products (phosphate rock, phosphoric
acid and fertilizers). We focus on the first two quays that are dedicated to the export
of fertilizers and partitioned into five berths each (1, 2, 3, 4, and 5  3  4 ). Each
berth has a minimum water depth and a fixed quay crane with a specific productivity.
The produced fertilizers (around 50 different types) are stored in 9 hangars. All the
hangars are linked to all the berths by a conveyor system composed of 90 routes. The
data sample received provides information about all the vessels that were berthed
during the year 2019. We consider 3 sets of 5 instances each for V  20, 30, 40 ,
generated from the data sample, for a planning horizon of 20 days (480 hours).

5.2 Computational Results

The output of the model refers to the scheduling of vessels and batches. These deci-
sions can be illustrated in a same Gantt chart (see Bouzekri et al. [9]). For each set of
5 instances of a given size, the table shows the number of instances solved, the num-
ber of instances solved to optimality, the average computation time in seconds, and
the average and maximum gap in percentages. For each instance, the computation
time was limited to 1 hour and the gap was provided by the solver as
100   ub  lb  / ub , where ub is the best upper bound obtained within the time limit,
and lb is the value of the objective function corresponding to the best integer solution
achieved. Overall, from the results, we can observe that the computation time increas-
es with the number of vessels and the solver can solve to optimality most of the cases.
Table 2. Computational results.
V Solved Optimum Avg. time Avg. gap Max. gap
20 5 5 5.9 0 0
30 5 5 203.6 0 0
40 5 3 1734.2 2.4 7.4
8

6 Conclusions and Future Research

In this paper, we study the Berth Allocation Problem with routing constraints in bulk
ports. Our study is motivated by the port of Jorf Lasfar, but it is also valid for any
bulk port. A new integer linear programming model is proposed to solve this problem.
The formulation proposed herein is flexible thanks to the use of predicates and it can
be used to solve real cases in bulk ports. Computational experiments show that our
model is able to solve the problem instances of realistic size (up to 40 vessels, 10
berths, 9 storage hangars, and 90 routes) in a reasonable computation time.
Further improvements are intended to be made such as considering tide constraints
and extending our model to integrate storage locations decisions under the restrictions
that forbid two or more cargo types to be stored in adjacent yard locations to avoid
intermixing. Also, a heuristic could be developed to obtain faster results.

References
1. Umang, N., Bierlaire, M., Vacca, I.: Exact and heuristic methods to solve the berth alloca-
tion problem in bulk ports. Transportation Research Part E: Logistics and Transportation
Review. 54, 14–31 (2013). https://doi.org/10.1016/j.tre.2013.03.003
2. Ernst, A.T., Oğuz, C., Singh, G., Taherkhani, G.: Mathematical models for the berth allo-
cation problem in dry bulk terminals. J Sched. 20, 459–473 (2017).
https://doi.org/10.1007/s10951-017-0510-8
3. Barros, V.H., Costa, T.S., Oliveira, A.C.M., Lorena, L.A.N.: Model and heuristic for berth
allocation in tidal bulk ports with stock level constraints. Computers & Industrial Engi-
neering. 60, 606–613 (2011). https://doi.org/10.1016/j.cie.2010.12.018
4. Robenek, T., Umang, N., Bierlaire, M., Ropke, S.: A branch-and-price algorithm to solve
the integrated berth allocation and yard assignment problem in bulk ports. European Jour-
nal of Operational Research. 235, 399–411 (2014).
https://doi.org/10.1016/j.ejor.2013.08.015
5. Unsal, O., Oguz, C.: An exact algorithm for integrated planning of operations in dry bulk
terminals. Transportation Research Part E: Logistics and Transportation Review. 126,
103–121 (2019). https://doi.org/10.1016/j.tre.2019.03.018
6. Pratap, S., Nayak, A., Kumar, A., Cheikhrouhou, N., Tiwari, M.K.: An integrated decision
support system for berth and ship unloader allocation in bulk material handling port. Com-
puters & Industrial Engineering. 106, 386–399 (2017).
https://doi.org/10.1016/j.cie.2016.12.009
7. Menezes, G.C., Mateus, G.R., Ravetti, M.G.: A branch and price algorithm to solve the in-
tegrated production planning and scheduling in bulk ports. European Journal of Operation-
al Research. 258, 926–937 (2017). https://doi.org/10.1016/j.ejor.2016.08.073
8. Bouzekri, H., Alpan, G., Giard, V.: Integrated Laycan and Berth Allocation Problem. In:
2019 International Conference on Industrial Engineering and Systems Management
(IESM), pp. 1–6. Shanghai, China (2019).
https://doi.org/10.1109/IESM45758.2019.8948110
9. Bouzekri, H., Alpan, G., Giard, V.: A Dynamic Hybrid Berth Allocation Problem with
Routing Constraints in Bulk Ports: Data sets and results. Mendeley Data, v2 (2020).
http://dx.doi.org/10.17632/bzmgbjttv3.2