A Negotiation Protocol to Improve Planning Coordination in Transport-driven Supply Chains

7/25/2019 A Negotiation Protocol to Improve Planning Coordination in Transport-driven Supply Chains

http://slidepdf.com/reader/full/a-negotiation-protocol-to-improve-planning-coordination-in-transport-driven 1/14

Journal of Manufacturing Systems 38 (2016) 13–26

Contents lists available at ScienceDirect

Journal of Manufacturing Systems

j ournal homepage : www.elsevier .com/ locate / jmansys

Technical Paper

A negotiation protocol to improve planning coordination intransport-driven supply chains

Zhen-Zhen Jia a,b, Jean-Christophe Deschampsa,b, Rémy Dupas a,b,∗

a Univ. Bordeaux, IMS, UMR 5218, F-33405 Talence, Franceb CNRS, IMS, UMR 5218, F-33405 Talence, France

a r t i c l e i n f o

Article history:

Received 16 January 2015Received in revised form

14 September 2015

Accepted 18 October 2015

Keywords:

Production

Distribution

Planning

Coordination

Negotiation

a b s t r a c t

This paper addresses the coordination problem of activities between manufacturers and transport opera-

tors (third party logistics) in the context of tactical planning. This critical problem is encountered in many

supply chains. Collaborative solutions, such as the Collaborative Planning, Forecasting and Replenishment

(CPRF) model, are not fully automatized and remain poorly suited for enhancing the relation between

manufacturers and transport operators. Furthermore, centralized planning isnot suitable in keeping con-

fidential the objectives of each partner of the same supply chain. Therefore, this work aims to develop a

decentralized planning approach based on a negotiation protocol.

Our approach tries to reach a “win–win” planning solution and to give some decisional flexibility to

transport operators. This protocol is founded on an incentive mechanism that can be used by transport

operators to progressively persuade manufacturers to accept a pickup plan. This study is focused on the

case of one manufacturer and one transport operator. The key determinants of the coordination protocol

and a set of planning models based on linear programming are presented here, followed by the design

of the experiments used to identify the factors affecting the overall performance of each partner. The

results demonstrate that it is possible to obtain plans that satisfy the manufacturer (i.e., the client of the

transport operator) while increasing profit for the transport operator. This is in favor of the application

of these principles to the coordination of multiple transport operators.

© 2015 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Third party logistics providers (3PL) are firms in charge of exe-

cuting a more or less significant part of logistics activities. Using

their services generally provides means for companies to subcon-

tract storage and transport activities to third parties. Nevertheless,

it raisesthe issue of howthe relationshipbetweenthird parties and

distribution activities could be improved when they are performed

by independent industrial partners, who usually aim to keep con-

fidential their own data and knowledge. The synchronization of

distributed operations primarilyoccursthroughaggregatedtacticalinformation sharing,thus giving themaster planningfunction great

importance for insuring an effective coordination of supply-chain

partners.

The present work focuses on the collaborative relationship

between manufacturersand third parties providing transportactiv-

∗ Corresponding author at: Univ. Bordeaux, IMS, UMR 5218, F-33405 Talence,

France. Tel.: +33 553774057.

E-mail address: [email protected] (R. Dupas).

ities, also called transport operators. This relation has two main

singularities in comparison with those that link production facil-

ities in a supply network. First, the transport operator’s profit

margins are much lower than the revenue of manufacturers (i.e.,

clients1 of the transport operator) generated by product sales.

Transport operators also have difficulties forecasting activities

because their various clients (i.e., transport orders) require multiple

and different transport services. In such cases, the transportation

activities of 3PLs are planned by manufacturers, based on the use

of specific tools such as DRP (distribution requirement planning),

when they intend to create a long-term climate of confidence withtheir clients. If these tools provide useful services for companies in

facilitating information and material-flow control, from consumer

demand to raw material supply, their implementation requires

information sharing. However, these tools are rarely implemented

by 3PLs with lowtransport capacity(i.e., 3PLs that owna small fleet

of vehicles: less than 5). Their weak level of computerization and

1 The following notations are adopted: ‘Client ’ refers to a customer of transport

services,and ‘customer ’ refers to a final customer.

http://dx.doi.org/10.1016/j.jmsy.2015.10.003

0278-6125/© 2015 TheSociety of Manufacturing Engineers. Publishedby Elsevier Ltd. All rightsreserved.

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.jmsy.2015.10.003

http://www.sciencedirect.com/science/journal/02786125

http://www.elsevier.com/locate/jmansys

mailto:[email protected]



mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jmsy.2015.10.003&domain=pdf

http://www.elsevier.com/locate/jmansys

http://www.sciencedirect.com/science/journal/02786125




14 Z.-Z. Jia et al. / Journal of Manufacturing Systems 38 (2016) 13–26

the lack of finance to accessto theElectronicDataInterchange (EDI)

usually reducethe usefulness of the DRP. Therefore, thedifficulty is

the coordination of transport operations and the balance between

transport resources and needs.

The main objective of this work consists of developing an

approach to coordinate transportation planning with production

planning models. More precisely, this research aims to study the

problem of production and transportation in the 3PL environment

under a decentralized coordination mode [20].

This paper is organized as follows. Section 2 proposes a litera-

ture review, and Section 3 presents the problem studied. Section 4

proposes the description of mathematical models and the negotia-

tionprotocol. Section 5 gives the numerical results for performance

evaluation. Section 6 summarizes conclusions and future research

directions.

2. Literature review

Collaborative planning in supply chains has drawn strong inter-

est for many years [1,27]. First, we present an overall view of

collaborative supply-chainplanning approaches. Then,we focus on

the relation between distribution and production.

2.1. Supply chain collaborative planning

Although an exhaustive literature survey of this field is beyond

the scope ofthispaper, a classification ofthe mainparadigmsfor the

planning coordination of partners is presented. The collaborative

approaches are broadly composed of two main groups, presented

below:

- Centralized approaches are based on a full model of partners that

supports the decision making for all supply-chain participants

[4]. They rely on the hypothesis of complete information shar-

ing. These models are then solved using either exact approaches,

based on mathematical programming, such as decomposition

approaches [5], or approximated approaches, such as heuristicsor metaheuristics. Also included in this group are hierarchical

planning methods, which aim to address the centralized prob-

lem throughits decompositioninto a hierarchy of interdependent

sub-problems. These centralized approaches are often difficult to

usein practice,primarily because companiesdo notwant to share

their confidential data.

- Decentralized or distributed approaches consider fully indepen-

dent partners. A comprehensive classification of decentralized

coordination methods in supply-chain planning can be found in

Taghipour and Frayret [29]. These approaches can take various

forms, such as information exchange, request for actions or more

advanced cooperation. For instance, supply contracts that link

customers with suppliers currently represent an important influ-

ence on the production and delivery of final products. Amraniet al. [3] showed that supply commitments, such as frozen hori-

zon (i.e., ordered quantities are considered fixed during this time

interval andcannotbe modified between two planningdecisions)

or flexibility rate (i.e., customers can change the ordered quan-

tities within a certain limit outside of the frozen horizon), as

stipulated in this contract, can be a powerful way to manage and

plan the product flow in a supply chain.

- Among moreadvanced cooperation forms, negotiation is a central

paradigm whose definition varies with authors. It can be defined

as an exchange between two or more partners with a view to

obtain an agreement[16]. Automatednegotiationapproaches can

be inexhaustively classified into three main following categories:

o Heuristic approaches: Partners iteratively adjust their local ini-

tial plan according to the capabilities of other partners. One of

the first approaches was proposed by Dudek [10], who devel-

oped a negotiation-based scheme. It combines mathematical

programmingfor theoptimal planningof each party so that the

two parties’ orders/supply plans can be synchronized for plan-

ning in the supply chain. Taghipour and Frayret [30] proposed

an extension of this model to address the dynamic changes

in the supply-chain environment that affect planning. In the

same lineage, Albrecht and Stadtler [2] f ormulated a theoreti-

cal scheme for coordinating decentralized parties that intends

to encompass all functionalities of supply chains. Ben Yahia

et al. [6] proposed a negotiation mechanism for collaborative

planning within a supply chain that is based on fuzzy rules.

Their approach is limited to cooperation between manufactur-

ers, considering onlyproduction planningwithout distribution,

supplier or retailers. These approaches are a practical and easy

way to implement negotiations between partners, though they

are not mathematicallyproven; for instance, theirconvergence

toward an agreement is not guaranteed.

o Game theory-based approaches: The best decision made by a

given partner in a supply chain is found takinginto account the

possible decisions of others. One of the first studies to apply

coordination and negotiation inside a supply chain was pro-

posed by Cachon and Netessine [8], who mentioned that two

main types of games-cooperative and non-cooperative (i.e., acompetitive game)-can be used. Game theory provides very

powerful strategies. However, their implementation to solve

a practical problem, such as planning coordination, remains a

delicate topic due to their reliance on the hypothesis of perfect

rationality.

o Multi-agent system-based approaches: Developed in artificial

intelligence problem solving, this paradigm has been inten-

sively applied to supply-chain collaboration. It is particularly

suited to automated negotiation due to the implementation

of decision mechanisms such as auctions or biding. Hernán-

dez et al. [17] proposed a negotiation-based mechanism that

is supported by a multi-agent system and focuses on the

collaboration of demand, production and replenishment plan-

ning, combined with the use of standard planning methods,such as the material requirement system (MRP) method.

Fischer et al. [15] proposed a methodology and a multi-agent

tool for the simulation of the transportation domain. Their

negotiation-based decentralized planning approach is applied

to the scheduling of the transportation orders among an agent

society consisting of shipping companies and their trucks. The

multi-agent paradigm is a central and powerful paradigm. Its

application for collaborative planning is limited only by the

methodology used to build the model and the decision mecha-

nisms integrated in the agents.

This previous classification has a practical interestto give a sim-

plified view of the domain. However, it must be noted that many

approaches are developed at the cross between each category. Forinstance, the multi-agent paradigm can also be used to implement

some game-theory principles.

2.2. Production and distribution planning

Reviews [12,13,23] haveindicated thatmost studies focus on the

formulationof an integratedproduction- and distribution-planning

model. Barbarosoglu and Ozgur [5] developed a mixed-integer

linear programming model solved by Lagrangian and heuristic

relaxation techniques to transform the problem into a hierarchi-

cal two-stage model: one for production planning and another

for transportation planning. Dhaenens-Flipo and Finke [9] devel-

oped a mixed-integer linear programming-based planning model

in a multi-firm, multi-product and multi-period environment in



Z.-Z. Jia et al. / Journal of Manufacturing Systems 38 (2016) 13–26 15

Fig. 1. Problem context.

which the supply chain is modeled as a flow network. Park [24]

proposed an integrated transport and production planning model

that uses mixed-integer linear programming in a multi-plant,

multi-retailer, multi-productand multi-period environment.Selim

et al. [25] proposed a fuzzy multi-objective linear programming

model that incorporates uncertainty of the individual decision

makers in charge of manufacturing plants or distribution centers.

Song et al. [26] studied a problem of a third party logistics (3PL)

provider that coordinates shipments between suppliers and cus-

tomers through a consolidation center in a distribution network.

The problem is formulated as a nonlinear optimization problem

and solved with a Lagrangian method. Bonfill et al. [7] proposed

a framework to address the interrelated production and trans-port scheduling problems that aims to support the coordination

of production and transport activities to manage the inventory

profiles and material flows between sites. Two approaches are

compared in this study: an integrated model and a solving strat-

egy using sequentially production and scheduling models. Jhaa

and Shankerb [19] studied the coupling of an inventory prob-

lem with a vehicle-routing problem with transportation cost in

a single-vendor multi-buyer supply chain. They proposed an iter-

ative approach for solving the integrated problem to optimality.

Zamarripa et al. [32] proposed an integrated multi-product and

multi-echelon tactical planning model for the coordination of part-

nersinsidea supply chain. The linear programmingmodel proposed

encompasses the production–distribution relation inside a supply

chain of chemicalproducts. Theseauthorsalso comparedtheir inte-grated model with a competitive game theory-based approach,

which enabled them to find the best scenario among several

alternatives [31].

As far as full decentralized approaches are concerned, Jung

et al. [21,22] proposed a negotiation process that aimed to find

a contract for a distributor and a manufacturer in a distributor-

driven supply chain. Nevertheless, the negotiation principle used,

which is based on the opportunity given to the manufacturer to

report shortages, offers little flexibility because it does not take

into account prices, auctions or the availability of extra resources.

Taghipour and Frayret [28] proposed a decentralized coordination

mechanism that is based on explicit negotiation using mathemat-

ical programming and involves two enterprises within the supply

chains.

Indeed, the decentralized-based cooperation of independent

production and distribution partners focused on mid-term tactical

planning has received limited attention.

In this paper, we propose a decentralized approach for distribu-

tion and production collaborationbased on negotiation. A heuristic

decentralized approach is chosen due to the previously mentioned

advantages.

This approach is founded on the negotiation-based collabo-

rative planning process, which was initially proposed by Dudek

[10] and refined by Dudek and Stadtler [11]. Its principle con-

sists of exchanging only non-confidential data between partners

and searching for new compromise solutions through an itera-

tive improvement process. In this process, each partner uses aso-called “preferred plan” as a target plan representing its own

interest. This process also includes the possibility for customers to

claim compensation associated with a compromise proposal. Note

that the process proposed by Dudek et al. is dedicated to the rela-

tion between suppliers and customers. Our negotiation approach,

detailed below, extends their collaborative planning process by

considering explicitly the transport operator as a collaborative

partner.

3. Problem definition

This work falls within the general objective of studying the

complex relationships that link manufacturers with their multiple

transport operators.Our contribution focuses on the study of the relationship

between one manufacturer (i.e., a client of transport services) and

one transport operator (Fig. 1). The manufacturer makes different

products to satisfy the demands of various customers and has a

limited production capacity and a limited finished-product stor-

age capacity. The transport operator manages a fleet of trucks that

have to pick up products from manufacturers and deliver them to

customers before returning to their initial location. Each partner

is in charge of planning its own activities and trying to maximize

its profit while taking into account the limitations of others, which

emerge from the plans exchanged during the negotiation.

Inthis context, we aimto propose a negotiationprotocolto align

the activities of two cooperating partners (manufacturerand trans-

port operator) with the aim to give more flexibility to the transport




operator. This requires that the planningactivities of both partners

be simulated and that the protocol be proposed.

The current study is indeed supported by a numerical simula-

tion, basedon a linear programmingapproach, andan experimental

approach, based on the design of experiments (DOE), to proceed in

a structured way and to reduce the number of experiments. Thus,

we aim to extract from this analysis the main dominant factors

that affect the performance of both partners before extending the

analysis to more complex situations.

Our study is based on the following hypothesis regarding the

three partners:

Customers:

– Customers accept that the quantities of delivered products can

have small deviations from the initial ordered quantities (late

or early deliveries). Nonetheless, they negotiate penalty clauses

when drafting the frame agreement between them and manu-

facturers.

Manufacturer :

– The manufacturer produces different products to satisfy the

demands of various customers. The demands of products from

all of these customers are known over a given non-rolling plan-

ning horizon. The manufacturer knows the transportation prices

and the delivery lead time required to serve the customers.

– Inventory levels of rawmaterials are considered infinite because

we focus only on the interaction between the manufacturer and

the transport operator. The replenishment decisions of materials

are not taken into account in this study.

– If the manufacturer cannot supply the right product quantity

at the right time, as requested by customers, financial penalties

arise from these deviations, as defined in the frame agreement,

and reduce the partner’s profit. These deviations are due to

limited capacity constraints in production or transport and leadto advanced delivery to customers or delayed transportation.

Transport operator :

– This partner has a limited operational capacity but has recourse

to subcontracting when customer demand requires a number of

trucks over its own capacity. The principle of a standard cost,

whose value is independent of the use of subcontracting, is

retained to cover all transport costs. Recourse to the outsourcing

and negotiation of transport arrangements is the sole responsi-

bility of the transport operator.

– To try to maximize its profit, the transport operator has the

opportunity to propose a pickup and transport plan with smalldeviations fromthe delivery planrequestedby the manufacturer.

Any deviation causes the transport operator to pay penalties to

the manufacturers affected by this change; those penalty clauses

are also stipulated in a specific agreement negotiated between

transport operators and manufacturers. These penalties, which

are called planning change penalties, differ from the financial

penalties previously defined.

– Theprovidedtransport service is addressedonly in a globalpoint

of view; the warehouse storage problems that arise in distribu-

tion activities are not studied. The transportservice is considered

a whole activity, including all service times related to the move-

mentof freight, i.e.,the dispatching and consolidation of material

flows, storage,handling andmoving. Thisactivity is characterized

by a determined delivery lead time.

4. Negotiation frameworkandmodeling

We consider a manufacturer and a 3PL that agreed to negotiate

to finda profit-maximizingsupply-chainplanning solutionwithout

sharing any confidential information. The negotiation is based on

the main following characteristics:

– The partnership relation already exists. The producer and trans-

port operator signed a global contract to define the framework

of collaboration, which details all necessary information, such as

price, global quantity for a long period, each partner’s respon-

sibility and penalties. The demand quantities for more detailed

time period (e.g., delivery plan, pickup plan) are not specified

in the contract but are negotiated during the collaborative pro-

cess. Therefore, negotiation takes place within the limits of this

contract.

– The negotiation between partners—manufacturer and logistic

service provider—is not supported by a DRP but is based on the

transmission of distribution plans so that the 3PL can have a

forecast over several days of the load induced by the clients of

transport services (i.e., manufacturer). Moreover, this negotia-

tion aims to be a “win–win” relationship for the two parties:

The required solution must aid the 3PL in maximizing its own

profit without significantlydecreasing the service rateof the finalcustomers.

The presentation of the negotiation framework is composed of

three sections. First, the overall description of the negotiation pro-

tocol is provided. The last three sections, respectively, focus on

the planning models used inside the negotiation protocol, the key

determinants and the flow-control logic of this protocol.

4.1. Overall description of the negotiation protocol

The negotiation protocol can be described as follows. The man-

ufacturer is the first to plan its production under limited-capacity

constraints and to attempt to generate a delivery plan according to

the customer’s demands, which is sent to the 3PL (Fig. 2). Throughthe generation of two different plans (i.e., the best profit and the

best service plans), the transport operator evaluates whether its

own profit shouldbe increasedby proposinga pick-up plan distinct

from the delivery plan requested by the manufacturer. According

its own interest, the transport operator has to payplanningchange

penalties in cases of late or early deliveries and canalso offer finan-

cial compensation to convince the manufacturer to accept the new

plan.

When a pickup plan received by the manufacturer completely

satisfies the delivery constraints of the production–planning pro-

cess, a converged solution is reached. Otherwise, the manufacturer

rejects thepickup plan proposed by the transportoperator, consid-

ering that its own profit is too low or not all production constraints

can be respected according to the received plan. The manufacturerrefuses to modify its initial delivery plan, so the transport operator

needs to make a new pickup plan, called the relaxed pickup plan, by

relaxing some economic constraints.

Itmay happen thatthe transportoperator cannot relaxanymore,

and no proposed pickup plan has ever been considered accept-

able by the manufacturer. In this case, the latter must adapt his

production to the constraints expressed by the transport operator

to generate a relaxed delivery plan. Consequently, a new round of

negotiation begins, based on the same process, until a compromise

solution is found. The transport operator progressively reduces its

economic standard in terms of profit, intending to send an accept-

able pickup plan to the manufacturer.

Note that our approach is based on the definition of an original

negotiation protocol between the manufacturer and the transport




operator.The main feature of this protocol is the degree of freedom

given to the transport operator. The models presented below are

standardbut aredevelopedto simulate thedecisionmakingprocess

of each partner and the global negotiation protocol.

4.2. Planning models

The protocol definition is based on two sets of three linear pro-

grammingmodelsthat characterizethe planningprocessof thetwopartners. These two sets are successively presented below.

4.2.1. Models simulating the planning activities of the

manufacturer

The planning activities of the manufacturer are carried out by

the following models:

– The “best production profit” (BPP) model describes the planning

process that leads to an initial production and delivery plan and

maximizes the manufacturer’s profit.

– The “Production Profit Evaluation” (PPE) model estimates the

admissibility of the pickup plan sent by the 3PL and calculates

the expected profit in cases of acceptance.– The “Relaxed Production Profit” (RPP) model proposes to adapt

the production plan to pick up constraints imposed by the 3PL

to converge toward a consensual planning solution that limits

decreases in the manufacturer’s profit.

Below, we introduce the notations used before formulating the

mathematical models.Sets Parameters

T Set of periods DT i Transportation lead time to customer j

P Set of products DP p Production lead time forproduct p

J Sets of customers SP p,i Selling price of product p to customer j

CS p Unitary inventorycost of product p per period

Indices CP p Unitary production cost of product p

t Index of planning period CR p, j Unitarylate supply cost of product p per period

(financial penalty)

p Index of products CE p, j Unitary early supply cost of product p per period

(financial penalty)

j Index of customers TP j Transportation price/ton to customer j

vu p Quantity of resource required to produce a product p

Decision variables v p Weight or volumeof product p

b p, j,t Late supplied quantity of product p for customer j at period t d pjt Demand of product p from customer j at period t

e p, j,t Early supplied quantity of product p for customer j at period t Emax p, j Upper bound forallowed early supplied quantity

f p,t Production quantity of product p launched in production at period t Pcapt Production capacity at period t

i p,t Inventory level of product p at the end ofperiod t Icapt Inventory capacity at period t

l p, j,t Delivery quantity of product p to be launched in transportation to

customer j at period t

MinP Relaxation lower profit bound of manufacturer

bP p,j,t

Max(0, qq p, j,t − l p, j,t ) M Very large integer

eP p,j,t

Max(0, l p, j,t − qq p, j,t ) qq p, j,t Pickup quantity of product p that the transport

operator decides to transport to customer j at period t

b controlP p,j,t

Binary variable equal to1 if bP p,j,t > 0, otherwise 0

e controlP p,j,t

Binary variable equal to1 if eP p,j,t > 0, otherwise 0

Best Production Profitmodel (BPP)

The BPP model formalizes the manufacturer decisions related

to production, inventory and delivery. The objective function (A.1)

maximizes the profit resultingfrom the revenue from selling prod-

ucts (SP), the production cost (CP), the inventory cost (CS) and the

financial penalties for late and early deliveries (CR and CE). Con-

straint (A.2) is the inventory balance equation. Constraint (A.3)

expresses the difference between the required and the supplied

quantities, taking into account the transportation lead time and

possible late and early deliveries. Constraints (A.4) and (A.5) guar-

antee production loads with respect to production and inventory

capacities. Constraint (A.6) limits the early supplied quantities

to the customer, which provides the producer certain flexibil-

ity to supply products. Constraint (A.7) indicates that delivery

quantities cannot exceed customer demand. Constraint (A.8) is a

non-negativity constraint.

Max

p

t

j

(SP p,j · l p,j,t ) − p

t

(CP p · f p,t )

− p

t

(CS p · i p,t ) − p

t

j

(CR p,j · b p,j,t + CE p,j · e p,j,t )

−

p

t

j

(TP j · v p · l p,j,t )

(A.1)

s.t.

i p,t = i p,t −1 + f p,t −DP p − j

l p,j,t ∀ p ∈ P, ∀t ∈ T. (A.2)

l p,j,t + b p,j,t − e p,j,t

= d p,j,t +DT j + b p,j,t −1 − e p,j,t −1 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.3)

p

u p ·

DP p =1

f p,t − +1

≤ Pcapt ∀t ∈ T. (A.4)

p

(v p · i p,t ) ≤ Icapt ∀t ∈ T. (A.5)

e p,j,t ≤ Emax p,j ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.6)t

l p,j,t ≤t

d p,j,t +DT j ∀ p ∈ P, ∀ j ∈ J. (A.7)

i p,t , b p,j,t , e p,j,t , f p,t , l p,j,t ≥ 0 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.8)

Production Profit Evaluationmodel (PPE)

The PPE model is a variant of the BPP model. The parameters

and decision variables of EPP are identical to those of the BPP

model. Constraints (A.2)–(A.8) f rom the BPP model are included




Fig. 2. Description of the negotiation protocol.

in this model. Note that in contrast to the previous model, parame-

ter qq p, j,t in this equation represents the pickup decisions made by

the transport operator and proposed to the producer.

Constraints(A.1)–(A.8)

l p,j,t = qq p,j,t ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.9)

Relaxed Production Profitmodel (RPP)

The RPP model formalizes manufacturer decisions, aiming torelax certainconstraintsto finda less constrainedplanning solution

adapted to the transport operator requirements.

This modelhastwo main inputparameters: a pickupplan(qq p, j,t )

and the customer demand plan (d pjt ). This pickup plan is received

from the transport operator and corresponds to a supplementary

target: The model aims to satisfy this plan as much as possible this

plan. This model also accepts a decrease in the manufacturer profit

(i.e., relaxed profit) up to a minimal bound labeled MinP . The out-

put variables of the RPP model are the production plan ( f p,t ), the

inventory plan (i p,t ) and the relaxed delivery plan (l p, j,t ).

The objectivefunction (A.10) minimizesthe deviationquantities

between the pickup plan received from the transport operator and

the relaxed output delivery plan; it also minimizes the inventory

levels. Constraints (A.2)–(A.8) f rom the BPP model are included in

this model. Constraint (A.11) f ormalizes the deviation between the

relaxed delivery plan and the pickup plan received from the trans-

port operator. If one of the deviation variables (i.e., bP p,j,t

or eP p,j,t

)

is positive, constraint (A.12) or constraint (A.13) expresses that the

correspondingcontrol variable (i.e.,M.b controlP p,j,t

or e controlP p,j,t

)

must be equal to 1. Constraint (A.14) ensures that only one devi-

ation variable can be positive at the same time. Constraint (A.15)

limits the relaxation of the profit up to a minimum value; note that

P profit corresponds to the profit value of Eq. (A.1). Constraint (A.16)

is the non-negativity constraint.

Min

p

j

t

bP p,j,t

+ eP p,j,t

+

p

t

i p,t

(A.10)

s.t.

Constraints(2)–(8)

l p,j,t = qq p,j,t − bP p,j,t

+ eP p,j,t

∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.11)

bP p,j,t

≤M.b controlP p,j,t

∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.12)

eP p,j,t ≤M.e controlP p,j,t ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.13)

b controlP p,j,t

+ e controlP p,j,t

≤ 1 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.14)




P profit ≥MinP ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.15)

bP p,j,t , eP p,j,t

≥ 0 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (A.16)

Note that financial penalties are explicitly considered costs in

the objective function of production models BPP and PPE. These

penalties are paid by the manufacturer every time the deliveries

requested by the customers cannot be strictly respected.

4.2.2. Models simulating the planning activities of the transport

operator

The transport operator (3PL) planning decision-making is also

described by three models:

– The “Best Transportation Profit” (BTP) model defines an optimal

plan that maximizes the transport operator’s profit even though

some early and late deliveries must be made.

– The “Best Transportation Service” (BTS) model calculates a

plan that matches the initial delivery plan requested by the

manufacturer as much as possible while taking into account

transportation capacity limits.

– The “Relaxed Transportation Profit” (RTP) model intends to pro-

pose a new pickup plan, by accepting a decrease in profit up to a

certain value (i.e., bound).

Specific notations are required to characterize the variables and

parameters used in the models related to the transport operator:

Decision variables Parameters

q p, j,t Pickup quantity of product p to be launched in

transportation from the manufacturer at time

period t to customer j

R Number of trucks initially available: initial transportation capacity of transport

operator

m j,t The number of trucks launching transportation

from manufacturer at time period t to

customer j

capextra Load capacity of an external truck

m extra j,t Thenumber of external trucks launching

transportation from the manufacturer at time


F C ext ra j Destination related cost per external truck

tb p, j,t Late pickup quantities of product p at time


M extra j,t Limit of thenumber of external trucks at period t to customer j

te p, j,t Early pickup quantities of product p at time


D j Transportation lead time of a round trip from depot though customer j back to

depot

bT p,j,t

Max(q∗ p,j,t

− q p,j,t ,0) cap Load capacity of a truck

eT p,j,t

Max(q p,j,t − q∗

p,j,t ,0) q∗

p,j,t Best service pickup plan, quantities of product p in time period t delivered to

customer j

b controlT p,j,t

Binary variable equal to 1 if bT p,j,t > 0,

otherwise 0

MinT Lower bound of transport operator’s profit

e controlT p,j,t

Binary variable equal to 1 if eT p,j,t > 0,

otherwise 0

FC j Destination related transportation cost to customer j pertruck

EC p, j Unitary early pickup cost of product p per period – (planning changepenalty)

BC p, j Unitary late pickup cost of product p perperiod – (planning change penalty)

Best Transportation Profitmodel (BTP)

The BTP model formalizes the pickup decisions of the transport

operator. The objective function (B.1) maximizes the profit result-ing from the transportation revenue (TP), the distance-related cost

(FC), the product-related cost (VC), the planning change penalties

due to late and early pickup (BC and EC) and the external resource

cost (FC extra j). The optimization model tries to avoid late pick-

ups and minimize the planning change penalty costs. Constraint

(B.2) evaluates the difference between the delivery plan sent by the

producer and the pickup plan proposed by the transport operator,

and it fixes late and early pickup quantities. Constraints (B.3)–(B.5)

correspond to the satisfaction of the capacity constraints. These

capacities concern the following three limits: (i) the global capac-

ity of transport, including extra resources-constraint (B.3), (ii) the

capacity of operator-owned vehicles-constraint (B.4), and (iii) the

capacityof external transportation resources-constraint (B.5). Con-

straint (B.6) ensures that the accumulated pickup quantities of all

products over the planning horizon for each customer not exceed

the corresponding accumulated delivery quantities of this cus-

tomer. Constraint (B.7) is the non-negativity constraint.

Max

p

t

j

TP p,j · v p · q p,j,t − j

t

FC j · m j,t

− p

t

j

VC p · q p,j,t − p

t

j

BC p,j · tb p,j,t + EC p,j · te p,j,t

−

j

t

FC extra j +mextra j

(B.1)

q p,j,t − te p,j,t + tb p,j,t = l p,j,t

− te p,j,t −1 + tb p,j,t −1 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (B.2)

p

v p · q p,j,t ≤ m j,t · cap+m extra j,t · capextra ∀t ∈ T, ∀ j ∈ J.

(B.3)

j

D j

i=1

m j,t −i+1 ≤ R ∀t ∈ T. (B.4)

m extra j,t ≤M extra j,t ∀t ∈ T, ∀ j ∈ J. (B.5) p

t

q p,j,t ≤ p

t

l p,j,t ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (B.6)

q p,j,t , tb p,j,t , te p,j,t ,m j,t ,m extra j,t ≥ 0 ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J.

(B.7)

Best Transportation Service model (BTS)

TheBTS model isa variantof BTPmodel;the two modelsare very

similar, except for their objective function. Constraints (B.2)–(B.7)




from the BTP model are included in this model. The BTS objective

functionminimizes the quantities of late deliveriesand the number

of extra vehicles.

Min p

j

t

(tb p,j,t + te p,j,t ) +t

j

(m extra j,t ) (B.8)

s.t. (B.2)–(B.7)

Relaxed Transportation Profit model (RTP)

The RTP model is used to relax certain constraints of the BPT

model to adapt to the producer’s demand. This model is strongly

analogous to the RPP model. It has two main input parameters: the

best service pickupplan(q∗ p,j,t

),as definedby thetransportoperator

during the first step of negotiation, and the delivery plan (l p, j,t ) sent

by the manufacturer. The best service pickup plan corresponds to

a supplementary target to the delivery plan. The output variables

of the RTP model are the transportation resource utilization plan

(m j,t , m extra j,t ) and the relaxed pickup plan (q p, j,t ).

The objective function (B.9) minimizes the deviation between

the relaxed output pickup plans and the best service pickup plan.

The RTP model includes Eqs. (B.2)–(B.7) f rom the BPT model. Con-

straint (B.10) expresses the deviation between the relaxed pickup

plan and the pickup plan obtained from the BST model. Constraints

(B.11) and (B.12) prevent deviations eT p,j,t

and bT p,j,t

from being

positive simultaneously. Constraint (B.13) limits the relaxation of

the profit up to a minimum value; note that T profit corresponds

to the profit value of Eq. (B.1) by the transport operator if the

relaxed transportation plan is accepted. Constraint (B.14) is the

non-negativity constraint.

Min p

j

t

(bT p,j,t

+ eT p,j,t

) (B.9)

s.t.

Equations (B.2)–(B.7)

q p,j,t = q∗P,j,t

− bT p,j,t

− eT p,j,t

∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (B.10)

bT p,j,t

≤ M · b controlT p,j,t ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (B.11)

eT p,j,t

≤ M · e controlT p,j,t ∀ p ∈ P, ∀t ∈ T, ∀ j ∈ J. (B.12)

T profit ≥ MinT (B.13)

bT p,j,t , eT p,j,t

≥ 0 ∀ p ∈ P,∀t ∈ T,∀ j ∈ J. (B.14)

Note that planning change penalties are explicitly considered as

costs in the objective function of the BTP transportation model.

The transportoperatorpays thesepenalties to the manufacturer for

the shift from the requested planning of the product pickup. These

penalties increase the profit of the manufacturer. Nonetheless, this

income is not explicitly expressed in the mathematicalmodel of the

manufacturerbut is includedin thefinancial flowof our negotiation

protocol.

To simplifythe descriptionof thenegotiationprotocol presented

in the next section, the following notations are introduced:

– GModel name specifies the profit resulting from the objective func-

tion of the model labeled Model name.

– GT ∈ {GBTP , GBTS , GRTP } (resp. GP ∈ {GBPP , GPPE , GRPP }) denotes the

profit corresponding to the pickup plans proposed by the trans-

port operator (resp. denotes the profit corresponding to the

production and delivery plans calculated by the manufacturer).

4.3. Key determinants of the negotiation protocol

No negotiation is truly possible if no decisional flexibility exists

between the two involved partners. The objective of negotiation

is to use this flexibility and information exchange to achieve a

win–win negotiation as much as possible. The negotiation proto-

col proposed in this work uses the fundamental notions presented

below.

4.3.1. Negotiation space

A negotiation space is a range of possible values on a one-

dimensional axis representing the profit for each partner.

Considering the transport operator, the negotiation space is

defined by the upper and lower bounds, denoted by GT and G- T .

The upper bound represents the maximum profit resulting from

the pickup plan calculated by the BTP model. The lower bound cor-

responds to the profit resultingfrom the pickup plan elaborated by

the BTS model, which tries to serve manufacturer demand as best

as possible.

A similar principle is used to define the profit range considered

acceptable by the manufacturer. Expected profit value GP corre-

sponds to profits GBPP or GRPP that the manufacturer has estimated,

respectively, through the calculation of initial delivery plans (“Best

Production Profit” model) or the calculation of a relaxed delivery

plan (“Relaxed Production Profit” model). The choice of using GBPP

or GRPP depends on which phase the negotiation protocol is in:

After receiving the pickup plan sent as a first counter-proposal by

the transport operator, the manufacturer has to compare the esti-

mated gain related to this plan with profit GBPP to estimate the loss

of earnings; profit GRPP is introduced later in the negotiation pro-

tocol, when the manufacturer has to adapt his production to the

constraints expressed by the transport operator. The manufacturer

thengenerates a relaxed delivery plan, which becomes the reference

in terms of expected profit.

This expected profit value is used only to estimate the lower

profit bound, above which the manufacturer’s profit is considered

as acceptable. This lower bound is defined relative to the following

expression:

G- P = GP · RG

with RG being the percentage reductionof the expected profit value.

4.3.2. Compensation

Compensation is an incentive mechanism used by the transport

operator to persuade the manufacturer to accept the pickup plans

he proposes. Indeed, the transport operator may propose a pickup

plan as closeas possible tothe manufacturer deliveryplan ordecide

to maximize its profit even if the resulting plan does not timely

respect the delivery quantities. The profit gap, denotedG, is then

defined as the difference in the profits corresponding to each of

these plans:

G = GBTP − GBTS

The transport operator can decide to share a part of this profit

gap with the manufacturer to motivate its acceptance of a pickup

plan. The sharing principle is controlled by a specific parameter

called the compensation percentage (CPP in), which defines the per-

centage of the profit gap that the transport operator intents to pay

to the manufacturer.

4.3.3. Plan acceptance

The plan acceptance criterion is a general mechanism used to

decide whether a plan is acceptable for a given partner. This prin-

ciple is used in Fink [14] f or supply-chain coordination by means

of automatednegotiations.This criterionis defined as the “natural”




Fig. 3. Relaxation degree of manufacturer and transport operator.

behavior of onepartner to decidewhethera newsolution proposed

is acceptable from its corresponding partner. Jain and Deshmukh

[18] also use a similar concept with a fuzzy preference acceptance

criterion for parameters of the negotiation between supply-chain

members.To judge whether a received pickup plan is acceptable, the

manufacturer uses the percentage RG. Concerning the transport

operator, plan acceptance is based on a parameter labeled ACRin,

which is used to decide which plan should be sent to the manufac-

turer: the one from the BTP model or the one from the BTS model.

ACRin is defined as a ratio that is always greater than or equal to 1.

Ratio GBTP /GBTS is calculated and compared with the value of ACRin

to decide whichplan will be sent in response to the manufacturer’s

delivery request.

4.3.4. Relaxation degree

When, after one or several negotiation iterations, one partner

cannot persuade the other one to precisely respond to its requests,

it must relax some financial constraints to propose a new feasibledelivery plan or pickup plan. Without considering this principle,

the negotiation can go on infinitely if each partner maintains its

position in term of expected profits.

Considering the transport operator, the relaxation degree is

considered when the manufacturer insists on its delivery plan.

The relaxation degree is calculated by the following formula:

RDT = (GBTP − GBTS )/MNN in. Parameter MNN in represents the max-

imum number of iterations during which the transport operator

can relax its profit, based on the proposition of the same delivery

plan. The use of the relaxation degree to progressively increase the

value domain of the transport operator’s profit promotes conver-

gence toward a plan, whichensures that a consensual solution will

be reached in a win–win relation.At each step of negotiation trans-

port, the operator reduces its expected profit by the value of the

relaxation degree (Fig. 3).

This principle, applied to the manufacturer planning process, is

nearly identical and is used only in case of executing the “Relaxed

Production” model;when a delivery plan cannot be satisfied by any

pickup plan sent by the transport operator, the relaxation degree

is calculated with the following formula: RDP = pp relax ∗GBPP .

Parameter pp relax represents the percentage of the production

profit that the manufacturer decides to relax after any iteration.

4.4. Control elements of the negotiation

As shown in Fig. 2, the proposed negotiation protocol involves

a distributed decision-making in which the manufacturer and

the transport operator coordinate their own planning activities.

Therefore, the control elements of this figure must be defined to

interact withplanning models to ensure cooperation flexibilityand

convergence through the overall decision process.

The first control element is on the transport operator side (see

CE1, Fig.2). Afterreceiving the delivery plan, the transportoperatormakes two plans. Ratio GBTP /GBTS is calculated and compared with

plan acceptancecriterion ACRin. If conditionGBTP /GBTS ≥ ACRin issat-

isfied, the transport operator proposes the plan issue from the BTP

model and agrees to pay planning change penalties when manufac-

turer requests cannot be completely respected. To encourage the

manufacturer to accept its proposition, the transport operator then

proposes compensation equal to CPP in * (GBTP − GBTS ). If the condi-

tion is not satisfied,the plan resultingfrom the BTS model is sent to

the manufacturer, possible planning change penalties are paid, and

no compensation is proposed.

The second control element (see CE2, Fig. 2), on the manufac-

turer side guarantees that the negotiation process can finish, even

if no compromise solution can be found. Once the manufacturer

receives a pickup plan (resulting from BTS, BTP or RTP models), the

first task is to evaluate with the PPE model whether this pickup

plan is feasible for the manufacturer:

– If the PPE model cannot provide a feasible plan, i.e., a production

plan consistent with the transportation constraints, the manu-

facturer should check further whether this plan equals to the

best service pickup plan. When the received plan equals to a best

service pickup plan, because there is only one transport opera-

tor, the manufacturer has nochoice butto adapt tothis constraint

and executes “Relaxed Production Profit” planning (RPP) to find

a counter delivery plan. Otherwise, the manufacturer insists on

sending a delivery plan computed with the BPP model.

– If the planresulting from the PPE model is feasible, and if expres-

sion (GPEE + compensation+ planning changepenalties) ≥ GP with

GP = {GBPP , GRPP } is satisfied, the manufacturer will agree on this

pickup plan, and the negotiation process stops with success.

Otherwise, the manufacturer will refuse the transport operator

proposal.

It must be noted that during the negotiation, if the RTP model

has been run to its maximum number of times, defined by MNN in,

the negotiation process stops with failure.

5. Simulation experiments and results

The data exchanges representing the dynamic aspects of infor-

mation flows between supply-chain partners are simulated on

a software platform coupling the six planning models with the




Table 1

Economic responses definition.

Name Label Profit

Difference in profit of manufacturer (percentage) DPP out ((GEPP +compensation+ planning change penalties)−GBPP )/GBPP

Difference in profit of transport operator (percentage) DPT out ((GBST or GBPT or GRT )− compensation−GBST )/GBST

Table 2

Factors levels.

Name Label Definition Levels

L1 L2 L3

Capacity ratio R CAP in Aggregated c ustomer requirements/Aggregated manufacturer capacity 1.01 0.96 0.75

Extra capacity ratio R EXT in Extra transportation capacity cost/Transportation cost 2 5 100

Inventory ratio R INV in Manufacturer inventory cost/Early supply cost 0.5 1 2

Selling price ratio R SELLin Selling price to customer 1/Selling price to customer 2 1 1.5 5

Late ratio R LA T in Late supply cost/Late pickup cost 0.5 1 2

Early ratio R EARin Early supply cost/Early pickup cost 0.5 1 2

Acceptance criterion ACRin Acceptance criterion 1.01 1.03 1.04

Compensation percentage CPP in Compensation percentage 30% 50% 70%

Maximum number of negotiation MNN in Maximum number of negotiations 5 10 20

Total number of trucks TNT in Total number of trucks (TNT in equals to R) 115 110 105

control elements. This platform is implemented under an Xpress-

MP2 solver coupled with Excel andVBA3 programming language totransfer and synchronize information between the manufacturer

and the transport operator.

The objective of this experimentationis to investigate the nego-

tiation process as a whole, to identify a list of main parameters

that affect its performance, from the most to the least influen-

tial and to provide conclusions about coordinating production and

transportation activities based on the negotiation principle.

The experimental study presented in this section is based on

the design of experiments (DOE), which is fully integrated in the

platform under Excel files representing the arrays and the results

of experiments. Our case study comprises one manufacturer and

one transport operator, which are in relation with two customers.

5.1. Design of experiments

TheDOE study startswith the determination of theobjectives of

the experiment (responses) and the input factors. Then, the exper-

iment table is chosen, followed by the analysis step of the results.

5.1.1. Responses definition

Based on the mathematical models and negotiation protocol

presentedin the previous section, two economicresponses, labeled

DPP out (i.e., difference in profit of manufacturer), and DPT out (i.e.,

difference in profit of transport operator), are considered. These

responses consist of comparing the initial (i.e., before negotiation)

and the final (i.e., after negotiation) profit values of the manufac-

turer (DPP out ) and the transport operator (DPT out ). Regarding, forinstance, response DPP out , the initial value profit equals GBPP , and

the final value equals (GEPP + compensation + planning change penal-

ties). The responsesdefinitionsare writtenin percentages,as shown

in Table 1.

5.1.2. Factors definitions and levels

The factors of the design of experiments consist of various

parameters related to the planning models of the manufacturer,

the transportoperatorand the negotiation protocol. In theseexper-

2 Dash optimization software.3

Microsoft Visual Basic for Applications language.

iments, the planning horizon comprises 22 time buckets that

correspond to approximately one working month. Table 2 containsthe definition of the ten input factors, each one having three possi-

ble values, labeled L1, L2 and L3. The factors list is divided into two

parts.

The upper part of Table 2 contains the following ratio factors:

– The capacity ratio (R CAP in) defines the balance between the

aggregated customer demand and the aggregated production

capacity over the planning horizon. Three cases are considered:

The demand is slightly superior (L1), slightly inferior (L2), or

strongly inferior (L3) to the production capacity.

– The extra capacity ratio (R EXT in) compares the cost of using

the extra transportation capacity with destination-related trans-

portationcosts. It affects thedecisions of whether and how much

extra transportation capacity is required. Three cases are consid-ered: The cost of extra capacity is twice the transportation cost

(L1), five times greater thanthe transportation cost(L2), or highly

prohibitive (L3).

– The inventory ratio (R INV in) is related to the equilibrium

between inventory cost and early supplied penalty cost. For

instance, if the early supplied penalty cost is less than inventory

cost, the manufacturer will prefer to make an early delivery, if it

is allowed by customer. Three cases are considered: The inven-

tory cost is half the early cost (L1), these two costs are balanced

(L2), or the inventory cost is twice the early cost (L3).

– The selling price ratio (R SELLin),in ourstudy case,differs depend-

ing on the business dealings between the manufacturer and its

various customers. It can be interpreted to a certain degree as

the relative priority of products to be served when the produc-tion capacity cannot meet all customer demands. Three cases are

considered: The selling prices of the two customers are equal

(L1), the price of customer 1 is significantly greater than that of

customer 2 (L2), or it is strongly superior (L3).

– The late ratio (R LAT in)-or the early ratio (R EARin)-specifies the

balance between the financial penalties paid by manufacturer to

customers andthe planning change penalties paidby the transport

operator to the manufacturer. Three cases are considered: The

financial penalties are half the planning change penalties (L1),

these penalties are balanced (L2), or the financial penalties are

twice the planning change penalties (L3).

Themainparametersof the planningmodels related tothe man-

ufacturer and the transportoperatorused in these experiments are




Table 3

Fixed-value parameters.

Name Label Definition Value

Profit reduction percentage RG Percentage reduction of the expected profit value 0.99

Profit relaxation percentage pp relax Percentage of production profit tha t manufacturer decides to rela x a fter any iteration 0.01

Table 4

Preliminary experiment results of economic responses.

No. of trial 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

DPP out (%) 0.10 0.07 −0.01 0.17 0.09 0.07 0.19 0.22 0.09 0.89 0.29 0.06 0.35 0.14 0.04

DPT out (%) 1.95 2.28 2.04 1.63 1.63 2.55 1.04 1.05 1.53 4.52 1.56 1.10 10.54 4.21 3.84

No. of trial 16 17 18 19 20 21 22 23 24 25 26 27 Max Min Average

DPP out (%) 0.67 0.22 0.07 −0.54 −10.1 0.01 −0.83 −1.49 0.02 −0.28 −3.03 0.00 0.89 −10.1 −0.46

DPT out (%) 7.53 3.01 3.21 5.97 19.32 0.00 6.55 23.07 0.46 5.97 23.38 1.18 23.38 0.00 5.23

given in Appendix A. This appendix contains the profile of the cus-

tomer’s demands (Table A.1) and all detail parameters involved in

the previous ratio factors: the production capacities (A.2), the sell-

ing prices (A.3), the penalties costs of late/early supply and pickup

(A.4), the destination-related transportation costs and the costs of

using extra capacity (A.5), and the inventory costs (A.6).

Thelower part of Table2contains theremaining factors thatcor-

respond directly to certain parameters of the mathematical models

or the negotiation protocol. These factors were introduced in Sec-

tion 4.4, except parameterTNT in, whichis the totalnumberof trucks

of the transport operator.

5.1.3. Fixed-value parameters required for simulation

To reduce the complexity of the design of experiments, some

parameters in relation with some key determinants of the negoti-

ation protocol mentioned in Table 3 have a fixed value. The value

of these parameters is arbitrarily defined; however, the value of pp-relax is chosen so as to guarantee a slight increase in the size of

the negotiation space for the manufacturer, whichdoes not quicklyrelax itsfinancial claimsin terms of profit.The value of RG is defined

to ensure that the negotiation protocol is triggered if a minimal

benefit can be expected by the transport operator at the end of

negotiation.

5.2. Experimental study

This study is carried out in two steps, presented below.

Step1: Preliminary experiments array

In this first experiment array,the interactions between anycou-

ple of factors are considered to have minor effects on the resultsand therefore are ignored. A fractional orthogonal array is carried

out to significantly reduce the number of trials to be performed

while providing information almost as effective as the full facto-

rial design. The fractional orthogonal array labeled L 27 is chosen

to take into account the 10 factors with three levels. The results

of this experiment must be confirmed through a complementary

trial that aims to check the hypothesis of nonexistent interaction.

Thiscomplementary experiment consistsof settingthe levelof each

factor so that it orients the concerned response (i.e., DPP out andDPT out ) toward the best value. Every response of the initial array is

comparedto thisconfirmationtrial. The synthesisof thetwo confir-

mation trials is reported in Table 5. It shows that the confirmation

success rate is more than 93% forboth responses, andtherefore,the

interactions can be ignored.

The results of these experiments are presented in Table 4. The

graph of Fig. 4 drawn from this table depicts the effects of theinput

parameters on the responses.

Two conclusions can be reached from this first experimental

step:

– After negotiation, the transport operator’s profit is improved

without significantlydecreasing the manufacturer’s profit: There

is a financial exchange in the negotiation. The transport

operator’s profit has an average increase of 5.23%, and the man-

ufacturer’s profit has an average decrease of 0.46%. Moreover,

these gains are reasonable in the field of transportation, where

the profit margins are generally low.

– A list of most influential parameters is set. The selling price

ratio R SELLin, the extra capacity ratio R EXT in, the capacity ratio

R CAP in and the total number of trucks TNT in are identified as

important factors that have significant effects on both responses.

Step 2: Refined experiment array

The refined experiment array focuses on the parameters of

the negotiation protocol. This array involves the three following

factors: the acceptance criterion ACRin, the maximum number of

negotiation MNN in, and the compensation percentage CPP in. To

carry out a full factorial design of the experiments, the number

of levels is reduced to two using the extreme values (L1 and L3) of

these factors. Therefore, array L 8 is chosen for these three factors

at two levels.

The results of these experiments are presented in Table 6. The

graph of Fig. 5 drawn from this table depicts the effects of theinput

parameters on the responses.

The followingconclusionscan be drawn fromthis second exper-

imental step:

– The different values of the current studied parameters do not

significantly affect the manufacturer’s profit.

– The maximum number of negotiations (MNN in) directly affects

the convergence of the negotiation process. If the number of

allowed iterations MNN in increases, the profits of the trans-

port operator are better, and the corresponding profit difference

before and after the negotiation is increased.

– The compensation percentage CPP in, which is a part of profit gap

that the transportoperatorpays to the manufacturer to motivate

the acceptance of a pickup plan, can also increase the transport

operator’s profit.

– The acceptance criterion ACRin does not significantly affect the

production or transportation the profit of the transport operator.




Fig. 4. Factors’ effects on theresponses (DPP out and DPT out ).

Table 5

Confirmation experiment.

Responses Confirmation rate Confirmation rate definition

DPP out 25/27 = 0.93 Number of trials having a response worse than theconfirmationresponse/27

DPT out 26/27 = 0.96

Table 6

Refined experiment.

No. of trial 1 2 3 4 5 6 7 8

DPT out (%) 1.95 1.39 2.44 1.74 1.95 1.39 2.44 1.74

DPP out (%) 0.01 0.02 0.01 0.03 0.01 0.02 0.01 0.03

Fig. 5. Factors’ effects on DPT out response.

These conclusions are consistent with our goal, which is to give

somecooperation flexibilityto the transportoperator; i.e.,the stud-

ied network can be assimilated to a transport-driven supply chain.

In this context, theimportant conclusion is that the transportoper-

ator’s profit can be significantly improved without negative effects

on the manufacturer’s profit. In this study, there is only one trans-

port operator, and consequently, the manufacturer slightly adapts

its production and delivery planning to the limitation of the trans-

port operator.

6. Conclusion

Thispaper focuses on the coordinationof theplanning processes

of independent partners inside a supply chain and on the relation

between manufacturer and transport operators in a decentral-

ized decision-making context. A negotiation protocol is described,

based on the following three concepts: ‘Compensation’, which can

be interpreted as an incentive mechanism used by the transport

operator to persuade the manufacturer to accept a pickup plan;

the ‘acceptance criterion’, which proposes a general mechanism to

decide whether negotiation can lead to a sufficient profit increase

andwhethera planningsolutioncan be estimatedas acceptable for

a partner; andthe ‘relaxation degree’, whichallows fordefining the

lower profit bound that each partner can expect at each step of thenegotiation. Simulations are carried out through an implemented

experimental platform that supports the performance measure-

ment of economic responses (i.e., profit). It has been shown that in

most cases, the transportoperator’s profit canbe increasedwithout

affecting the manufacturer’s profit.

These results enableus to validatethe approachproposedin this

paper. Future developments of this workwill concern the following

points:

• The main limitation of this approach is that the experimenta-

tions are conducted in the context of one manufacturer and one

transportoperator. It is necessaryto extend thiscontributionfirst

to the case of multiple transport operators and then to multiple

manufacturers and multiple transport operators. This extension




Table A.1

Customer’s demands profile.

Product 1 required

resource/unit

Demand (unit) Product 2 required

resource/unit

Demand (unit) Inventory Capacity (ton)

d p, j,t d p, j,t Icapt

Period Customer 1 Customer 2 Period Customer 1 Customer 2

1 0 0 1 0 0 4000

2 0 0 2 0 0 4000

3 130 150 3 110 200 4000

4 150 130 4 120 210 4000

5 170 170 5 110 170 4000

6 130 200 6 130 200 4000

7 170 170 7 130 150 4000

8 179 120 8 108 120 4000

9 200 150 9 140 200 4000

10 170 170 10 180 190 4000

11 150 190 11 140 210 4000

12 200 200 12 110 220 4000

13 230 150 13 100 160 4000

14 201 170 14 130 220 4000

15 150 100 15 105 130 4000

16 160 130 16 123 200 4000

17 130 170 17 125 200 4000

18 121 150 18 132 190 4000

19 150 150 19 125 200 4000

20 169 170 20 155 133 4000

21 100 150 21 200 150 4000

22 151 120 22 132 190 4000

Total 3211 3110 Total 2605 3643 88,000

Table A.2

Production capacities.

R CAP in

L1 L2 L3

1.01 0.96 0.75

Pcapt

931 980 1254

Table A.3

Selling prices.

R SELLin

L1 L2 L3

0.5 1 2

SP p,2 SP p,1

Product 1 270 135 270 540

Product 2 360 180 360 720

needs to study the problem of the splittingand the assignment of

a global delivery request to each individual transport operator.• Regarding the planning method related to the manufacturer and

the transportoperator, it would be necessaryto integratea rolling

planninghorizon, which is commonlyused in industrial practices.

Indeed, the rolling planning mechanism is effective for coping

with the uncertainty of the market, which causes customers to

change their orders.• Finally, concerning the negotiation, it would be useful to take

advantage of some of the principles from cooperative game

theory to improve the efficiency of our protocol. Our ongoingresearch aims to test the relevance of using the Shapley value

in defining a win–win relationship between a manufacturer and

many transport operators.

Appendix A. Input data andmodels parameters

See Tables A.1–A.6.

Table A.4

Unitary late/early supply and pickup penalties costs.

R L AT in

L1 L2 L3

0.5 1 2

BC p, j CR p, j

Customer 1 Customer 2 Customer 1 Customer 2 Customer 1 Customer 2 Customer 1 Customer 2

Product 1 40 45 20 22.5 40 45 80 90

Product 2 50 55 25 27.5 50 55 100 110

R EARin

L1 L2 L3

0.5 1 2

EC p, j CE p, j

Customer 1 Customer 2 Customer 1 Customer 2 Customer 1 Customer 2 Customer 1 Customer 2

Product 1 20 25 10 12.5 20 25 40 50

Product 2 30 35 15 17.5 30 35 60 70




Table A.5

Destination-related transportation costs and using extra capacity costs.

R EXT in

L1 L2 L3

2 5 100

FC j FC extra j

Customer 1 600 1200 3000 60,000

Customer 2 800 1600 4000 80,000

Table A.6

Inventory costs.

R INV in

L1 L2 L3

0.5 1 2

CS p, j

Product 1

Customer 1 11.25 22.5 45

Customer 2 11.25 22.5 45

Product 2

Customer 1 16.25 32.5 65Customer 2 16.25 32.5 65

References

[1] Albrecht M. Supply chain coordination mechanisms: new approaches for col-laborative planning. Lecture notes in economics and mathematical systems,vol. 628. Germany:Paul Hartmann AG; 2010. p. 1–211.

[2] Albrecht M, Stadtler H. Coordinating decentralized linear programs byexchange of primal information.Eur J Oper Res2015;247(3):788–96.

[3] Amrani A, Deschamps JC, Bourrieres JP. The impact of supply contracts onsupply chain product-flow management. J Manuf Syst 2012;31(2):253–66.

[4] Arshinder K, Kanda A, Deshmukh SG. A review on supply chain coordina-tion:coordination mechanisms, managinguncertainty andresearchdirections.In: Supply chain coordination under uncertainty. International handbooks oninformation systems Berlin, Heidelberg: Springer; 2011. p. 39–82.

[5] Barbarosoglu G, Ozgur D. Hierarchical design of an integrated production and

2-echelon distribution system. Eur J Oper Res 1999;118(3):464–84.[6] BenYahiaW, AyadiO, Masmoudi F. A fuzzy-basednegotiationapproachfor col-

laborativeplanning inmanufacturingsupplychains.J Intell Manuf2015;(April).[7] Bo nfi ll A, E spun a A, P uigjane r L. Decis ion s uppor t f ramewor k f or coor -

dinated production and transport scheduling in SCM. Comput Chem Eng2008;32(6):1206–24.

[8] CachonGP, Netessine S. Game theoryin supply chain analysis.In: Simchi-LeviD, W u S D, Sh en M, editors. Handbook of quantitative supply chain analysis:modeling in thee-business era. Boston: Kluwer Academic Publishers; 2004. p.13–66.

[9] Dhaenens-Flipo C, Finke G. An integrated model for an industrialproduction–distribution problem. IIE Trans 2001;33(9):705–15.

[10] Dudek G. Collaborative planning in supply chains – a negotiation basedapproach. Lecture notes in economics and mathematical systems, vol. 533.Springer; 2004.

[11] Dudek G, Stadler H. Negotiation based collaborative planning between supplychains partners.Eur J Oper Res2005;163(3):668–87.

[12] Erengüc SS, Simpson NC, Vakharia AJ. Integrated production/distribution plan-ning in supply chains: an invited review. EurJ Oper Res1999;115:219–36.

[13] Fahimnia B, Zanjirani Farahani R, Marian R, Luong L. A review and critique onintegrated production–distribution planning models and techniques. J Manuf Syst 2013;32(1):1–19.

[14] FinkA. Supplychaincoordinationby meansofautomatednegotiationsbetween

autonomous agents. Multiagent based supply chain management, vol. 28.Springer; 2006. p. 351–72.

[15] Fischer K, Chaib-draa B, Müller JP, Pischel M, Gerber C. A simulation approachbasedon negotiation and cooperation between agents: a casestudy. IEEETransSyst Man Cybernet C: Appl Rev 1999;29(4):531–45.

[16] Forget P, D’A mours S, Frayret J-M, Gaudreault J. Study of the perfor-mance of multi-behaviour agents for supply chain planning. Comput Ind2009;60(9):698–708.

[17] Hernández JE, Mula J, Poler R, Lyons AC. Collaborative planning in multi-tiersupply chains supported by a negotiation-based mechanism and multi-agentsystem. Group Decis Negotiat 2014;23(2):235–69.

[18] Jain V, DeshmukhSG. Dynamic supply chain modeling using a new fuzzy hybridnegotiation mechanism. Int J Prod Econ 2009;122(1):319–28.

[19] Jhaa JK, Shankerb K. An integrated inventory problem with transportationin a diver gen t supply chain under se rvice level con st raint . J Manuf Sy st2014;33(4):462–75.

[20] Jia ZZ [PhD thesis] Planification décentralisée des activités de productionet de transport: coordination par négociation. Bordeaux University; 2012[in English] http://ori-oai.u-bordeaux1.fr/pdf/2012/JIA ZHENZHEN 2012.pdf

[accessed 20.12.12].[21] Jung H, Chen FF, Jeong B. Decentralized supply chain planning framework for

third party logistics partnership. Comput Ind Eng 2008;55(2):348–64.[22] Jung H, Jeong B, Lee C-G. An order quantity negotiation model for distributor-

driven supply chains. Int J Prod Econ 2008;111(1):147–58.[23] Mula J, Peidro D, Díaz-Madronero M, Vicens E. Mathematical programming

models for supply chain production and transport planning. Eur J Oper Res2010;204(3):377–90.

[24] Park YB. An integrated approach for production and distribution planning insupply chain management. Int J Prod Res2005;43(6):1205–24.

[25] Selim H, Araz C, Ozkarahan I. Collaborative production–distribution planningin supply chain: a fuzzy goal programming approach. Transport Res E: LogistTransport Rev 2008;44(3):396–419.

[26] Song H, Hsu VN, Cheung RK. Distribution coordination between suppliers andcustomerswith a consolidation center. J Oper Res2008;56(5):1264–70.

[27] Stadtler H. A framework for collaborative planning and state-of-the-art. OR Spectr 2009;31:5–30.

[28] Taghipour A, Frayret J-M. Negotiation-based coordination in supply chain:

model and discussion. In:2010 IEEE international conference on systems manand cybernetics (SMC). 2010. p. 1643–9.[29] TaghipourA, FrayretJ-M. Coordinationof operations planningin supplychains:

a review. Int J BusPerform Supply Chain Model 2013;5(3):272–307.[30] TaghipourA, Frayret J-M. Dynamic mutual adjustment searchfor supplychain

operations planning co-ordination. Int J Prod Res 2013;51(9):2715–39.[31] ZamarripaMA, Aguirre AM, Méndez CA, Espuna A. Mathematical programming

and game theory optimization-based toolfor supplychain planning in cooper-ative/competitive environments. Chem Eng Res Des 2013;9(1):1588–600.

[32] Zamarripa M, Hjaila K, Silvente J, Espuna A. Tactical management for coordi-nated supply chains. Comput Chem Eng 2014;66:110–23.

http://refhub.elsevier.com/S0278-6125(15)00080-1/sbref0165

















































































http://ori-oai.u-bordeaux1.fr/pdf/2012/JIA_ZHENZHEN_2012.pdf






























































































































































































































































































































































































































































































































































































































































































































































































































































Documents

A Negotiation Protocol to Improve Planning Coordination in Transport-driven Supply Chains