Access the full text.
Sign up today, get DeepDyve free for 14 days.
N. Sadeh, David Hildum, Dag Kjenstad (2003)
Agent-Based E-Supply Chain Decision SupportJournal of Organizational Computing and Electronic Commerce, 13
H.-H. Wu, J.-Y. Liu (2008)
A capacity available-to-promise model for drum-buffer-rope systemsInternational Journal of Production Research, 46
T. Rupp, M. Ristic (2004)
Determination and exchange of supply information for co-operation in complex production networksRobotics Auton. Syst., 49
D. Emerson, S. Piramuthu (2004)
Agent-based framework for dynamic supply chain configuration37th Annual Hawaii International Conference on System Sciences, 2004. Proceedings of the
T. Bui, M. Shakun (1996)
Negotiation processes, Evolutionary Systems Design, and NEGOTIATORGroup Decision and Negotiation, 5
N. Jennings, K. Sycara, M. Wooldridge (2004)
A Roadmap of Agent Research and DevelopmentAutonomous Agents and Multi-Agent Systems, 1
(2005)
Demand fulfillment and ATP. In: Stadtler H, Kilger C (eds) Supply chain management and advanced planning, 3rd edn
Kitae Shin, C. Leem (2002)
A reference system for internet based inter-enterprise electronic commerceJ. Syst. Softw., 60
R. Fung, Tsiushuang Chen (2005)
A multiagent supply chain planning and coordination architectureThe International Journal of Advanced Manufacturing Technology, 25
T. Bui, Alexandre Gachet, H. Sebastian (2006)
Web Services for Negotiation and Bargaining in Electronic Markets: Design Requirements, Proof-of-Concepts, and Potential Applications to e-ProcurementGroup Decision and Negotiation, 15
(1998)
ARBAS: an action-resource based argumentation systems
T. Bui, Jintae Lee (1999)
An agent-based framework for building decision support systemsDecis. Support Syst., 25
Jérémie Gallien, Y. Tallec, Tor Schoenmeyr (2004)
A Model for Make-To-Order Revenue Management 1
C. Makatsoris, Y. Chang (2008)
Beyond CRM: A System to Bridge the Gap between the Customer and the Global Manufacturing Supply ChainProceedings of the 41st Annual Hawaii International Conference on System Sciences (HICSS 2008)
Y Yan, J Yen, T Bui (2002)
A multi-agent-based negotiation support system for distributed transmission cost allocationInt J Intell Syst Acc Financ Manag, 10
(1987)
Co-oP: a multiple-criteria group decision support system
D. Frey, Tim Stockheim, Peer-Oliver Woelk, R. Zimmermann (2003)
Integrated multi-agent-based supply chain managementWET ICE 2003. Proceedings. Twelfth IEEE International Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprises, 2003.
Ann Bixby, Brian Downs, Mike Self (2006)
A Scheduling and Capable-to-Promise Application for Swift & CompanyInterfaces, 36
Niels Agatz, M. Fleischmann, J. Nunen (2006)
E-fulfillment and multi-channel distribution - A reviewEur. J. Oper. Res., 187
C. Vidal, M. Goetschalckx (1997)
Strategic production-distribution models: A critical review with emphasis on global supply chain modelsEuropean Journal of Operational Research, 98
J. Conklin, M. Begeman (1988)
gIBIS: a hypertext tool for exploratory policy discussionACM Trans. Inf. Syst., 6
B. Fulkerson (1997)
A response to dynamic change in the market placeDecis. Support Syst., 21
T. Bui, H. Sebastian (2010)
Integration of Multi-Criteria Decision Analysis and Negotiation in Order Promising2010 43rd Hawaii International Conference on System Sciences
Zhenying Zhao, M. Ball, Masahiro Kotake (2005)
Optimization-Based Available-To-Promise with Multi-Stage Resource AvailabilityAnnals of Operations Research, 135
Martin Grieger (2003)
Electronic marketplaces: A literature review and a call for supply chain management researchEur. J. Oper. Res., 144
Staffan Larsson (2002)
Issue-based Dialogue Management
H. Stadtler (2005)
Supply chain management and advanced planning--basics, overview and challengesEur. J. Oper. Res., 163
J. Olhager, Björn Östlund (1990)
An integrated push-pull manufacturing strategyEuropean Journal of Operational Research, 45
C Kilger, H Meyr (2008)
Supply chain management and advanced planning, 4th edn
G. Dudek, H. Stadtler (2005)
Negotiation-based collaborative planning between supply chains partnersEur. J. Oper. Res., 163
D. Moodie (1999)
Due Date Demand Management : Negotiating the Trade-off Between Price and DeliveryInternational Journal of Production Research, 37
T. Bui, F. Bodart, P. Ma (1997)
ARBAS: A Formal Language to Support Argumentation in Network-Based OrganizationsJ. Manag. Inf. Syst., 14
D. Simchi-Levi, S. Wu, Z. Shen (2004)
Handbook of Quantitative Supply Chain Analysis: Modeling in the E-Business Era (International Series in Operations Research & Management Science)
Morris Cohen, Hau Lee (1988)
Strategic Analysis of Integrated Production-Distribution Systems: Models and MethodsOper. Res., 36
Chien-Yu Chen, Zhenying Zhao, M. Ball (2001)
Quantity and Due Date Quoting Available to PromiseInformation Systems Frontiers, 3
N. Karacapilidis, D. Papadias (2001)
Computer supported argumentation and collaborative decision making: the HERMES systemInf. Syst., 26
P. Kellett (2006)
Conflict Dialogue: Working With Layers of Meaning for Productive Relationships
M. Ball, Chien-Yu Chen, Zhenying Zhao (2004)
Available to Promise
B. Reinig, Robert Briggs, Morgan Shepherd, Jerome Yen, J. Nunamaker (1995)
Developing and validating an instrument to measure the impact of group support technology on affective rewardProceedings of the Twenty-Eighth Annual Hawaii International Conference on System Sciences, 4
M. Bichler, G. Kersten, Stefan Strecker (2003)
Towards a Structured Design of Electronic NegotiationsGroup Decision and Negotiation, 12
ME Fischer (2001)
Available-to-promise: Aufgaben und Verfahren im Rahmen Des supply chain managament
G. Tan, M. Shaw, B. Fulkerson (2000)
Web-based Supply Chain ManagementInformation Systems Frontiers, 2
H. Meyr (2004)
Supply chain planning in the German automotive industryOR Spectrum, 26
T. Bui, H. Sebastian, Christoph Hempsch, T. Bosse (2009)
A multi-agent simulation framework for automated negotiation in order promising
Xiaohong Zhang, J. Ou, S. Gilbert (2008)
Coordination of stocking decisions in an assemble-to-order environmentEur. J. Oper. Res., 189
A. Azevedo, J. Sousa (2000)
A component-based approach to support order planning in a distributed manufacturing enterpriseJournal of Materials Processing Technology, 107
MO Ball, C-Y Chen, Z-Y Zhao (2004)
Handbook of quantitative supply chain analysis: modeling in the E-business era
H. Richards, H. Dudenhausen, C. Makatseris, L. Ridder (1997)
Flow of orders through a virtual enterprise-their proactive planning and scheduling, and reactive controlComputing & Control Engineering Journal, 8
M. Louta, I. Roussaki, L. Pechlivanos (2008)
An intelligent agent negotiation strategy in the electronic marketplace environmentEur. J. Oper. Res., 187
(2013)
811–819 40 Logist
H. Stadtler, C. Kilger, H. Meyr (2000)
Supply Chain Management and Advanced Planning
(2004)
A model for make - to - order revenue management , Working Paper 23 . GNU Linear Programming Kit , Version 4
Scott Moses, Hank Grant, L. Gruenwald, S. Pulat (2004)
Real-time due-date promising by build-to-order environmentsInternational Journal of Production Research, 42
N. Julka, R. Srinivasan, I. Karimi (2002)
Agent-based supply chain management—1: frameworkComputers & Chemical Engineering, 26
Logist. Res. (2013) 6:25–41 DOI 10.1007/s12159-012-0094-9 OR IGINAL PAPER Solving the order promising impasse using multi-criteria decision analysis and negotiation • • Christoph Hempsch Hans-Ju ¨ rgen Sebastian Tung Bui Received: 21 December 2011 / Accepted: 30 October 2012 / Published online: 1 December 2012 Springer-Verlag Berlin Heidelberg 2012 Abstract In an inter-firm make-to-order production 1 Introduction environment, it is not always possible to satisfy the avail- able-to-promise (ATP) and capable-to-promise (CTP) Despite the adoption of seamless integration of supply conditions due to untimely and unmatched production chain management and order management applications capacity and customer demand. Therefore, it is important using near real-time data in many core industries, finding to quickly explore alternate solutions that would satisfy an alternate solution to an initial order that cannot be sat- both the customers and the suppliers. The purpose of this isfied remains a challenge to manufacturers. For a supply paper is to present a multi-agent-based system for auto- chain planner, an effective sourcing strategy in a volatile mated multi-attribute negotiation in order promising. Our market is critical to ensure cost-effective replenishment of proposed solution is based on concepts of evolutionary products or services between participating firms in their system design that advocates for continued exploration of global operations. As an extension to the available-to- new solutions until a satisfactory solution is found. Based promise (ATP) function, capable-to-promise (CTP) takes on a number of real-life ordering situations—changes of into consideration capacity information in the supply chain delivery date, price adjustments, and addition/modifica- to generate order promising in a short-term, order-based tions of value-added services as part of the order package, production environment. For incoming customer orders, we embed multi-attribute multi-utility simulations into a CTP decides whether or not it is possible to fulfill the linear program to search for a negotiated solution when a desired order quantity and delivery date. With the use of typical ATP/CTP function of a supply chain management online systems, an order promise depends on a number of system fails to fulfill a customer order. complex factors: increasing number of products and ser- vices requested by an increasing number of customers, Keywords Supply chain management Multiple-criteria possibility of product customization during the ordering analysis Multi-agent systems Heuristics Group decision process, expected shorter cycle time and life cycles, and and negotiations Simulation flexible pricing (e.g., Kilger and Meyr [31]. Under scarcity of raw materials and constrained production capacity, orders may be rejected or not fulfilled based on the initial terms of the customers. When an initial order is not ful- C. Hempsch filled, the planner needs to quickly find an alternate and Deutsche Post DHL, Bonn, Germany mutually agreeable solution between the supplier(s) and the e-mail: christoph.hempsch@deutschepost.de customer(s). Yet, a major weakness of the most CTP functions currently available in supply chain systems is the H.-J. Sebastian (&) RWTH Aachen, Aachen, Germany difficulty in finding an alternate solution. The planner is e-mail: sebastian@or.rwth-aachen.de typically given a set of attributes to specify (or to relax), such as the amount of backward/forward consumption to T. Bui cover the current shortage or the time fence. This CTP University of Hawaii, Honolulu, HI, USA procedure is arbitrary and time-consuming. It requires a e-mail: tungb@hawaii.edu 123 26 Logist. Res. (2013) 6:25–41 skillful and experienced supply chain planner to success- or extended before the desired date of delivery, the pro- fully negotiate an alternate order. ducer has to decide on the following: This article combines and extends approaches presented • quantity, by the co-authors at the AMCIS conference 2009 [11] and • due date, and the HICSS 43 conference 2010 [12]. The paper shows how • price a multi-attribute negotiation process can be embedded in a to commit to each customer order [30]. linear program to solve the severe limitation of current ATP and CTP functions are widely discussed in the computer-supported ATP functions. It discusses how to supply chain literature (e.g., Ball et al. [3, Kilger and design a negotiation-augmented supply chain network to Schneeweiss 30, Shin and Leem 44]. ATP can be defined support CTP communication, bargaining, and negotiation as a simple function that looks up the producers’ finished activities. In the absence of an initial, negotiation heuristics products inventory and, if available, reserves the quantity are integrated in optimization routines to generate alter- to satisfy an incoming customer’s order. CTP in turn takes native offers. the whole production process into consideration to look The paper is organized as follows. First, we briefly ahead what quantity may be available within a certain time introduce the domain of order promising in the context of a frame. (Some authors denote the functionality of CTP as make-to-order production environment and suggest how Advanced ATP [13].) negotiation can be applied to ATP/CTP calculations. We Figure 1 depicts a basic workflow of the ATP/CTP next extend our analysis to a production–distribution net- functions. The process is triggered by a customer order to work. Proposed negotiation concepts are operationalized the producer. A customer order typically contains a set of using multi-attribute utility functions. products (with order positions), the desired quantities, and a delivery due date. Sometimes, a price is specified with the order, but price is usually used as a negotiation item. To 2 Order promising in operational supply chain check whether the order can be fulfilled, the ATP and CTP management functions are executed consecutively. If the ATP function is able to fulfill the ordered quantity from existing inven- 2.1 The standard case of ATP/CTP functions tories in the supply chain, the requisition is created and the products should be delivered on time. If the order cannot be As previously introduced, available-to-promise (ATP) and satisfied, the CTP function checks whether the ordered capable-to-promise (CTP) are known activities in the products can be produced within the time delivery con- management of inter-organizational supply chains. Within straint. If this is the case, the CTP function creates a pro- a make-to-order or configure-to-order production environ- duction and delivery plan to fulfill the order. Otherwise, the ment, production or configuration is not initiated until the customer is notified that his order cannot be executed, producer receives a customer order demanding the specific unless some modification to that must be negotiated. product. Due to the fact that the quantity of materials or Given the ATP/CTP functions, Fig. 2 shows how these components in stock or the production resources may be functions operate in a simple pull-based make-to-order limited at a given point in time and cannot be replenished production where one single producer receives customer Fig. 1 Basic workflow of available-to-promise (ATP) and a capable-to-promise (CTP) functions 123 Logist. Res. (2013) 6:25–41 27 Fig. 2 Basic scenario of an aggregated view of a supply chain from the perspective of ATP/CTP functionality orders. The producer P has an attached warehouse W for finished products. We will come back to this standard case in Sect. 2.3. 2.2 Order promising in a production–distribution network In supply chain management, order promising refers to the ability of the SC planner to a global view of supply and Fig. 3 An example of a production–distribution network demand in order for him to answer to customer orders. The layer (S ,…, S ) and the customer layer of the supply chain 1 4 global supply chain network consists of several production in Fig. 3 will not be considered in the discussion below. The sites linked together by a distribution network. The logistics configuration of the production–distribution subsystem of of these networks depend on the number and type of the supply chain is characterized as follows: products (functional or innovative products), the type of production (make-to-stock, make-to-order, and configure- • There is one producer P (the focal enterprise of the focal to-order), and the location of the push–pull boundary within SC) with n manufacturing facilities at different loca- the network. From the perspective of the supplier(s), the tions PL ,i = 1,…, n. Without loss of generalizability, objectives of order promising are to meet customer orders in we consider one functional product manufactured at all such way that minimizes costs (e.g., avoiding production n production locations. Of course, there are several delays or overtime, missing or excessive inventory, addi- variants of the considered product, such as different tional transportation costs due to last-minute shipping configurations of a laptop computer, but we do not deal needs), increasing revenues by achieving higher capture in this model with a configure-to-order production rates given current production capacity, and ultimately environment. We assume to have a make-to-stock improving customer satisfaction by keeping order promises production environment, and we assume that the order accurate and by keeping lead time as short as possible. penetration point within the supply chain is given. For example, generic products are manufactured at the 2.2.1 A dynamic model of a production–distribution production sites, and the places where the variants are built are the warehouses. network We consider a simple case of a pure push-oriented system (e.g., Olhager and Ostlund [39]. In this make-to- In this Sect. 2.2.1 we explain the network and define parameters in particular orders, decision variables, and stock environment, the variants are manufactured at the production locations using demand forecasts and dis- constraints. This is the basis of a collection of submodels which will be introduced in Sect. 2.2.2 by selection of a tributed to the retailers. However, as seen later, the included negotiation process will result in a combina- subset of constraints from Sect. 2.2.1 and by adding two objective functions. tion of make-to-stock and make-to-order. For the sake of clarity, we focus on a production–dis- The production locations PL have given capacities C i i tribution network for make-to-order products with con- (number of items produced per time unit) and given vergent manufacturing process. In this logistic network, production costs K (unit per item), i = 1,…, n. Each manufacturing facility PL has an inventory at its both the supplier stage and the end-customer stage of the entire supply chain are not considered explicitly (see location (attached stock) with a stock level I (PL ), t i i = 1,…, n, at time t which is available for direct Fig. 3). The end-customer demand is modeled by customer orders which are generated by retailers. delivery to the warehouses or to the retailers. • There are retailers RT , k = 1,…, l, that represent ‘‘the Similar systems have been considered in the literature (e.g., Cohen et al. [14, Vidal and Goetschalckx 46]). Thus, customers’’ in the supply chain. Each retailer predicts the demand of its customers and generates ‘‘customer and for the purpose of this paper that focuses more on the negotiation heuristics at the first level of actors, the supplier orders’’ to the supply chain. 123 28 Logist. Res. (2013) 6:25–41 • The distribution system consists of warehouses WL , at time t to the supply chain. To respond to this stream of j = 1,…, m, and transportation links. orders, the focal enterprise P can: focal • The inventory level of the considered product at time • accept all orders, t is given by I (WL ), j = 1,…, m, measured in number t j • accept only a subset of orders and reject the remaining, of items. or • If we denote P = {PL |i = 1,…, n}, W = {WL |j = i j • start negotiation with the retailers to modify the 1,…, m} and R = {RT |k = 1,…,l}, we can represent existing orders (see Table 1). the transportation relations or links as follows: A reasonable decision-making using a CTP system can be • Producer–warehouse links: ðPL ; WLÞ2 A i j 1 based on a quantitative model that takes into consideration P W both the production–distribution system introduced above • Links between warehouses: ðWL ; WL Þ2 A j1 j2 2 and the order-stream. The detailed consideration of the W W network gives much more degrees of freedom (many • Warehouse–retailer links: ðWL ; RT Þ2 A j k 3 decision variables) for such a network-CTP in comparison W R with a standard-CTP described above. For example, the • Direct transportation links: ðPL ; RT Þ2 A i k 4 decision space of a network-CTP includes the following: P R • delivery from warehouses (from which warehouses, how many items) Each transportation link has an assigned transportation • direct delivery from the producers inventories (from time, which is assumed to be deterministic and known. The which producers, how many items) respective sets of links are denoted by A , A , A , and A . 1 2 3 4 (k) • where to produce (how many items of the product), and An order O of retailer RT at time t is represented by: t k • replenishment for the warehouses. ðkÞ ðÞ k ðÞ k O ¼ðQ ; RLT Þ; t t t This leads to the definition of decision variables in (k) where Q is quantity of the order of RT at time t and t k Table 2. (k) RLT , requested lead time of the order t With the decision variables defined in Table 2, the (We use (k) in order to assign the order to the retailer decision model as the basis for the network-CTP can now RT and to avoid confusion when considering the notations k be formulated as follows: in Table 1. In Table 1 we introduce three types of order Constraints: sets by aggregating with respect to the set of retailers. Warehouse replenishment: Therefore, k is not needed within the respective notations.) x ðtÞ yðÞ t þ tðÞ x ðÞ t x ðtÞ; where t ðx ðÞ t Þ ¼ At time t, the supply chain is faced with the following ð1Þ i;j m i i m i j¼1 subsets of orders (Table 1). Note that in the following we do not distinguish between for all i = 1,…, n and t = 0, 1, 2, … the different sets of orders in order to keep the model Delivery assignment of retailers: tractable. Adding more variables to the supply chain net- m n X X work would unnecessarily increase the complexity in ðkÞ w ðtÞþ p ðtÞ Q ð2Þ j;k i;k comparison with the aggregated basic scenario in Fig. 2. j¼1 i¼1 However, a subsequent extension of the model is possible. for all k = 1,…, l and t = 0, 1, 2, … To trigger the process, assume that the retailers RT , Inventory dynamics—warehouses and producers: which represent the customers, generate a stream of orders I WL ¼ I WL þ y t d tþ1 j t j ij ij i:ðÞ PL ; WL 2A i j 1 Table 1 Possible order sets w ðtÞð3Þ ðÞ k jk ‘‘New set’’ of orders at time t O ¼fO jk ¼ 1; ...; lg t t k:ðWL ; RT Þ2A S j k 3 ac 0 ac O Accepted orders at time t \ t, O : 0 0 t \t t \t t \t for all j = 1,…, m and t = 0, 1, 2,…. accepted orders before t S X re 0 re O Rejected orders at time t \ t, O : 0 0 t \t I ðÞ PL ¼ IðÞ PL þ y ðÞ t t \t tþ1 i t i ij t \t rejected orders before t j: PL ; WL 2A ðÞ i j 1 neg 0 Orders arrived at time t and selected for O 0 p ðÞ t þ x ðtÞð4Þ ik i t \t S neg negotiation, O 0 : orders selected t \t k:ðPL ; RT Þ2A i k 4 t \t for negotiation before t for all i = 1,…, n and t = 0, 1, 2, … 123 Logist. Res. (2013) 6:25–41 29 Table 2 Decision variables Decision Variable Description Delivery from warehouse stock w (t) Quantity sent from warehouse WL to RT starting at time t j,k j k (If transportation time is d , then the quantity arrives at the retailers place at time point t þ d ) j;k j;k Direct delivery from the p (t) Quantity sent from producer PL to RT directly starting at time t i,k i k 0 0 manufacturer’s stock (If transportation time is d , then it arrives at time t þ d i;k i;k at the retailers location) Production orders for the x (t) Quantity which producer PL starts to manufacture at time t (With C number of items produced by i i i manufacturers PL per time unit, the manufacturing time for any single order can be computed.) Distribution (replenishment) y (t) Quantity sent from producer PL to warehouse WL i,j i j decisions (in order to store the amount of goods there) at time t Distribution decision constraints: w ðtÞ I ðWL Þ for all j and k j;k t j ð8Þ X X p ðtÞ I ðPL Þ for all i and k i;k t i w ðÞ t I WL þ y t d jk t j ij ij ð5Þ k:ðÞ WL ; RT 2A i:ðÞ PL ; WL 2A j k 3 i j 1 (b) To support multistage decision-making, our model is dynamic in nature with discrete time periods and state for all j = 1,…,m X X variables. y ðÞ t þ p ðÞ t IðÞ PL þ x ðÞ t ij ik t i i ð6Þ k:ðÞ PL ; RT2A j: PL ; WL 2A i k 4 Equation (3) shows how the inventory levels of the ðÞ i j 1 warehouses are changing over time. Formula (4) models for all i = 1, 2, …, m and, both, (5) and (6) for all t = 0, 1, the state transformation for the producer inventories. 2, … (c) If we use stock variables I (PL ) and I (WL ) to model Now, we discuss the constraints (1)to (6). PL produces t i t j producer and warehouse inventories, we have to make a quantity x (t) starting at time t. Then, the manufacturing sure that these variables remain nonnegative over time t of this quantity x (t) (denoted by t (x (t)) becomes: m i m i time. Therefore, from (3) and (4) we derive (5) and x ðtÞ tðÞ x ðÞ t ¼ ð7Þ (6) with I ðWL Þ 0 and I ðPL Þ 0. With initial m i tþ1 j tþ1 i values I ðWL Þ 0, I ðPL Þ 0, and inequalities (5) 0 j 0 i We further assume that transportation of the produced good and (6) fulfilled, nonnegative inventory levels are to the warehouses does not start before the whole pro- guaranteed. duction order quantity x (t) has been produced and that Also, we have to take into account the capacity there is no inventory available at the production site. Then, constraints: we get the constraint (1). Equation (1) shows that con- 0 x ðtÞ C straints are becoming more difficult to formulate if we i i assume a continuous time parameter t and if we explicitly where C are maximal production capacities per time consider time lags caused by production capacity con- interval. straints and the missing inventories from the producers. Therefore, we assume discrete time t, t = 0, 1, 2, …, and 2.2.2 A simple strategic procedure for a CTP network inventories are integrated to the producers PL . We also assume x ðtÞ C . Then, we replace constraint (1) with the i i In 2.2.1, we developed a dynamic model for the production– constraints (4) and (6). distribution network as a subsystem of a supply chain. This We also define the constraints related to an order model might become very large in terms of the number of (k) quantity Q . time-dependent variables and constraints. This model can be used in order to simulate the processes (i.e., production, (a) Delivery from warehouse stock or direct delivery storage, transportation) of the supply chain. For a given set of from manufacturers stock to a retailer RT should not (1) (l) (k) orders O = {O ,…, O } at time t, we can use the data be bigger than the ordered quantity Q of this t t t (k) (k) (k) O = (Q , RLT ), respectively. The ordered quantity and retailer. Formula (2) assumes that the respective links t t t requested lead time define possible situations. Then we in the network exist. If not, the respective decision define a strategy which means mainly a sequence in which variable is set to 0. The delivery starts at time t situations are checked. Perhaps, the most intuitive strategy provided that the inventory levels at time t allow the related to the network is the four-step procedure below: following: 123 30 Logist. Res. (2013) 6:25–41 no m I. If possible, fulfill the order-set from warehouse inven- X ðkÞ ðkÞ LT ¼ max d jw ðÞ t [ 0 and w ðÞ t ¼ Q ; ð11Þ t jk jk t tories only. jk j¼1;...; m j¼1 II. If I is not possible, try to fulfill the order-set by using both the warehouse inventories and the goods which where d denotes the transportation time from warehouse jk are in transport from the producers to the warehouse WL to retailer RT . j k (regular replenishment processes of the warehouses) (Lead time refers to the maximal transportation time III. If II is not possible, additionally select direct delivery from the warehouses to RT , provided the entire ordered from producers to the retailers. quantity is delivered.) IV. If III is not possible, additionally create new In sum, we have l ? 1 objectives, if we do not aggregate production at time t. lead times of the retailers. If we define an overall lead time by: Each situation (I–IV) corresponds with a submodel that can ðkÞ be built from the set of decision variables and constraints LT ¼ max LT ð12Þ t t we have introduced before. It is obvious that model I. is the simplest one (the network-ATP) and the other models we get two objectives where both—costs and lead time— become subsequently more complex in the sequence I, II, need to be minimized. III, and IV. The task at hand for producer P , as the focal enter- focal To illustrate the proposed approach, we will first con- prise in the supply chain, is to set the problem up as a series sider situation I (see Fig. 4). of multi-criteria problems and use the solutions of these problems initial offers to engage in a negotiation process m l X X ðkÞ Situation I is given; if I ðWL Þ Q ð9Þ between the producer P = P (the Supply Chain Agent) focal t j t j¼1 k¼1 and the retailers RT ,…,RT (the Retailer Agents). In order 1 l for us to embed negotiation in the model, we provide in the In this case, it is possible to fulfill the overall ordered next section a brief discussion and justification of using quantity by the aggregated warehouse stock I ðWL Þ t j j¼1 negotiation processes in order promising. at time t. However, inequality (9) does not check whether the delivery from warehouse inventories is possible or partly possible within the requested lead times. Also, there 3 Negotiation in order promising is no decision which quantities should be delivered from which warehouse to which retailer. In order to come up 3.1 Order promising as a compromise with optimal decisions from the supply chain from the between producers and consumers point of view of focal enterprise (the producer P ), we focal define two objective functions by considering the cost K of The main objective of the producer in order promising, of delivery from the warehouses, and the lead time LT for course, is to maximize revenue and earnings by manufac- this delivery process. turing and selling as many products to as many customers m l as possible. As we are considering a pull-based production XX K ¼ K w ðtÞð10Þ t jk jk environment, customer satisfaction is of high importance in j¼1 k¼1 the long run. In general, there are three critical factors that determine the quality of an order promising system from a where K denotes transportation cost per item with regard jk customer satisfaction point of view: to the relation WL ? RT . (If the link WL ? RT is not j k j k included in the network, we have w (t) = 0 by definition.) jk • Reaction time of the system: The duration of the (k) (k) (k) We define the lead time for order O = (Q , RLT ) t t t decision-making process should be as short as possible of retailer RT by: to get the orders to the customers. Fig. 4 The network-ATP (situation I) 123 Logist. Res. (2013) 6:25–41 31 • Quality of promised due date: The customer desires a information with a variety of appropriate decision methods. short delivery time and a reliable prediction on it. NMS implement negotiation processes between multiple • Order acceptance rate: Only a small number of entities. Their aim is to improve the efficiency of the customer orders should be rejected unless the selection negotiation processes through communications support and of accepted orders is solely based on short-term profit assistance toward integrative bargaining. maximization considerations. A customer with a Most NSS seek to improve the outcome of the party that rejected order may choose to buy the product from uses the system. In contrast, and as their name suggests, another producer—and may not come back in the NMS are used to help the negotiation party to gain a more future. effective result. In this paper, we define negotiation in its broadest context, that is, any activity that helps avoid a There are multiple decisions apart from the ones solution impasse, or better yet, one that would yield a win– mentioned above that are commonly incorporated into win situation to both customers and suppliers when the ATP and CTP to achieve these objectives. For example, to initial order by the customers cannot be met by suppliers. be able to accept more customer orders, order splitting or Acknowledging the existence of more than one issue in a quantity splitting may be considered. Order splitting allows typical negotiation, the general literature in multiple-attri- the delivery of order positions of a customer order at bute utility theory advocates for the continued exploration different dates. Quantity splitting allows splitting the of solution until a compromise is found (e.g., Bui [6]). The ordered quantity into multiple orders and delivering these exploration of ‘‘solution space’’ can be achieved by looking orders at different delivery dates. The workflow for the for new solutions that had not been thought of, adjusting or CTP function described above is rather common and is refocusing on views of the problem, or adding/replacing widely discussed in the literature. The issue in this paper, actors. This concept is known as evolutionary in the design one that is not yet studied in depth in the current literature, of negotiation processes [10]. As shown in the next section, is how to proceed once a customer order is rejected by the the consideration to split the quantity of an order that CTP function. This step requires some form of negotiation cannot be fulfilled or the adding of some additional ser- with the customers until an alternate solution is satisfactory vices to a late delivery are examples of evolving the initial for all parties. solutions to a new feasible set of possible solutions that are acceptable to all involved parties. 3.2 Automated negotiation The notion of automated negotiation implies that some aspects of a negotiation are either conducted or at least Research in negotiation support tends to focus on two supported by autonomous computer agents or parties [8]. In major areas: communication support and bargaining and the context of ATP or CTP, this automated negotiation group decision support. Experience has shown that the could be of routine procedures (e.g., fast and expanded more antagonists engage in exchanging information and search of ‘‘matching solutions,’’ quick estimation of expression of their positions in a clear and concise manner, delivery time, or instantaneous reporting of inventory the more likely that they will move toward a solution that is levels). Furthermore, the agents could also be pre-pro- acceptable to all. The underlying principle is that to steer or grammed to act as a trained mediator looking for heuristic- redirect communication leads to conflict to one that based problem solving (e.g., Emerson and Piramuthu [17]). encourages conflict resolution, and better yet, collaboration For example, the first procedural rule of an automated [29]. A number of researchers propose the creation of agent would be to immediately acknowledge the reception computer-based platforms to support communications of a customer’s order and the generation of an alternate through structured language such as argumentation lan- solution should the initial order cannot be satisfied. In a guage (e.g., Bui et al. [7, Karacapilidis and Papdias 28]) or distributed platform linking customers to suppliers, the language to help structure negotiation issues (e.g., issue- automation of negotiation processes could be implemented based dialogue management) [15, 32]. Another area of by a series of simple to more functional agents, thus a research is to search for techniques to improve the nego- multi-agent system. tiation outcomes (e.g., Bui [6], Yan et al. [48]). These techniques range from optimization to heuristics, from 3.3 Potential benefits of introducing negotiation game theory to simulation (e.g., Bichler et al. [4] for a capabilities to CTP review). Among negotiation systems, there are negotiation support systems (NSS) and negotiation mediation systems As mentioned above, the success of an order promising (NMS). NSS are designed to assist negotiators in reaching system depends on three critical factors, that is, short mutually satisfactory decisions by providing a means reaction time, quality of promised due date, and a high of communication and through analysis of available acceptance rate. Unfortunately, the objectives of producers 123 32 Logist. Res. (2013) 6:25–41 Table 3 Multiple issues in CTP and conflicting objectives partnership and not a situation in which producers and customers are confronted with divergent objectives. Order Attributes Producer Customer Thus, there is potential for further research on enhancing Due date Late Early the efficiency and effectiveness of order promising systems Quantity High Ordered amount by introducing negotiation concepts and systems. Table 5 Price Low Low is a brief review of relevant literature supporting our Value-added services (VAS) Low High research work. It classifies recent research literature on ATP/CTP, negotiation or Multi-Agent Systems (MAS), and, most relevant to this research, work that attempts to and customers regarding the order attributes are, at least in integrate these four topics. some cases, divergent (see Table 3). This discrepancy needs to be taken into consideration by the focal producer 3.4 Implementation of a multi-agent simulation whenever a customer order is rejected by the order prom- framework for automated negotiation in order ising system and a counteroffer is needed. As discussed promising earlier, any order that is not fulfilled is an opportunity loss. Table 4 shows suggested strategies on four typical nego- Multi-agent systems (MAS) are information systems that tiable order attributes for the producer. Obviously, nego- have been of great interest in research over the last years. tiation concepts especially a negotiation support system for They consist of several intelligent agents that can exchange the computation of counteroffers may enhance the overall information or objects with each other. By doing so, agents efficiency of the order promising system. can be designed to address complex problems which would The potential need for, and benefit of, introducing be very difficult or impossible to solve with a single negotiation support to the domain of order promising is intelligent agent. In a distributed environment, MAS and discussed in recent literature (e.g., Rupp and Ristic [41]). their agents are naturally well suited to replicate real-world Yet, most authors consider negotiation just for contracting organizations or units. Agent-based technology can today before the ATP or CTP functions are executed, that is, be found in a wide range of applications like disaster producer and customer settle on fixed values or intervals response and modeling social systems (e.g., Jennings et al. for quantities and due dates (e.g., Sadeh et al. [42, Shin and [26]). The intelligent agents of a MAS share some impor- Leem 43]). Other authors claim in their research that tant characteristics: They are mostly autonomous. They negotiation processes have been implemented, but they do only have a limited, local view of the global environment. not explain or even formalize these processes in details And there is no single agent that is able to control all the (e.g., Makatsoris et al. [34]). To our knowledge and at the others. The agents are defined by their objectives, attri- time of this writing, very few specific negotiation processes butes, and behavior [27]. or systems have been proposed to support post-ATP/CTP We have developed a multi-agent system (MAS) pro- negotiation. Dudek and Stadtler [16] discuss a system for totype that consists of different agents representing the negotiation-based collaborative planning between supply retailers business and its customers as shown in Fig. 2. The chain partners, which consists of a supplier and a buyer. implemented MAS focuses on the decision column of The supplier offers an initial quantity that can be revised by Table 4, but it can be extended to support activities during the buyer. Yet, their work assumes a collaborative the pre- and post-decision phases. Table 4 Suggested negotiation strategies to deal with CTP issues Decision Pre-decision Decision Post-decision attributes Due Date Forecast arriving orders and build Produce in advance or negotiate later date Evaluate forecast accuracy, and stock and production capacity if necessary adjust forecast accordingly techniques Quantity Forecast incoming orders and build Reduce quantity or split it Evaluate forecast accuracy, and stock and production capacity if necessary adjust forecast accordingly techniques Price Conduct market research on Reduce price to compensate for later due date and/ Check whether pricing was right competitive pricing or smaller quantity Value-added Build up competence in customer Offer customers value-added services to Assess customer satisfaction services services and preferences research compensate for late delivery and/or delivery with (VAS) smaller quantity 123 Logist. Res. (2013) 6:25–41 33 Table 5 Selected literature on SCM Order promising/ATP/CTP Negotiation support MAS ATP/CTP, MAS and negotiation SCM/ATP/CTP Agatz et al. [1]x x Azevedo and Sousa [2]x x Bixby et al. [5]x x Chen et al. [13]x x Fischer [18]x x Meyr [35]x x Moses et al. [37]x x Rupp and Ristic [41]x x Kilger and Schneeweiss [30]x x Kilger and Meyr [31]x x Vidal and Goetschalckx [46]x Wu and Liu [47]x x Zhao et al. [50]x x Argumentation and negotiation Bichler et al. [4]x x Bui et al. [8]x x Bui and Shakun [10]x Karacapilidis and Papdias [28]x Larsson [32]x Louta et al. [33]x x MAS Bui and Lee [9] x Julka et al. [27]x x SCM and MAS Fulkerson [20]x x x Sadeh et al. [42]x x x SCM and negotiation Gallien et al. [22]x x x Grieger [24]x x Shin and Leem [43]x x x Moodie and Bobrowski [36]x x x Zhang et al. [49]x x x SCM, MAS and negotiation Bui et al. [11]x x x x Dudek and Stadtler [44]x x x Frey et al. [19]x x x Fung and Chen [21]x x x x Makatsoris and Chang [34]x x x x Richards et al. [40]x x x x Stadtler [44]x x x x Tan et al. [45]x x x x 3.5 System architecture ordered products, it has little choice but adopting a make- to-order environment. To illustrate our negotiation framework, we use the case of The Order Collection agent receives customer orders a computer retailer (producer). The customers place orders and passes them on to the Supply Chain Central Control that typically consist of a specified number of computer unit. This agent communicates with Production and systems. Since the retailer cannot accurately predict these Inventory Control to get the necessary information to call orders with specific configurations and reliability of the CTP solver. A linear program is used to decide whether 123 34 Logist. Res. (2013) 6:25–41 4 Outline of a negotiation approach within a CTP environment Bui and Shakun [10] published a negotiation approach and software tool (NEGOTIATOR) to support distributed negotiation using multiple-criteria Pareto optimization. As discussed earlier, the underlying principle is the Evolutionary System Design approach that advocates for a systematic search of new solutions until a satisfactory is found. Although this evolutionary approach was originally applied to fields such as corporate strategies or Fig. 5 System architecture of a negotiation-assisted make-to-order public policies, the technique lends itself well to a CTP environment environment in supply chain management. Therefore, in order to show how a negotiation approach could be or not the orders are accepted or rejected. These decisions applied to a CTP environment, we explain the approach are in turn returned to the Central Control agent. The latter by Bui and Shakun [10] using an example. First, we attempts to derive alternatives for the rejected orders. apply the general concepts of the method to this appli- These counteroffers are then passed on to the Order cation area. Negotiation agent that uses an algorithm described later to modify the counteroffers using new price and value-added 4.1 A combined optimization: negotiation approach services as terms of negotiation with the hope that they will illustrated by an example be considered and accepted by the customer. The Order Negotiation then offers the counteroffers to the customers We consider the production–distribution network example who are asked to take positions. In Fig. 5 a system archi- from 2.2 with only one retailer. tecture is given which is able to realize the negotiation- assisted make-to-order approach described above. Finally, 4.1.1 Definition of values, goal variables, control the distribution agent takes care on realizing the decisions variables, and weights within the physical distribution network. A UML sequence diagram of this CTP and negotiation process shows the life • General values are high performance, reliability in span of and communication between the agent processes in delivery, and safety. [11]. • Operational expressions of the general values are formulated by goal variables, delivered quantity (of 3.6 Software framework and tools the requested good), lead time, cost, and price • Notations of the control variables of the example- The MAS was implemented using the Repast Simphony Supply Chain: framework (North et al.’s Web site [38]). This Java-based environment provides a graphical user interface for running Q —ordered quantity by the retailer (party A) simulations within a MAS. The different agents are OQ —offered quantity by the SC in reply to the order implemented using plain Java classes. The GNU Linear (SC—agent (this means the focal enterprise P )is focal Programming Kit (GLPK Website) [23] and its Java party B of negotiation) interface (GLPKJNI website) [25] are used to solve the K —cost of the ordered quantity Q t t CTP model. P —price of the ordered quantity Q t t The system details as well as the results of computa- LT —lead time of the ordered quantity Q t t tional experiments are published in [11]. As a proof of t—is the time index t [ {0, 1, 2, …} concept, the simulation of the automated negotiation The SC-agent computes cost and lead time using an opti- showed that the number of rejected orders could be reduced mization submodel (introduced in 2.2.2) and calculates a while the overall revenue increased. price P . The successful simulation experiments encouraged us to continue the integration of multi-criteria approaches and • Control variables of party A (retailer) are as follows: negotiation techniques to the production–distribution net- ordered quantity (number of items), price (e.g., EURO, work (see Fig. 3) and the model-based approach outlined in Dollars), and lead time (e.g., days, weeks) Sect. 2.2.2. The novelty here is to use interactive instead of • Control variables of party B (Retailer) are as follows: automated negotiation and to develop a hybrid approach. offered quantity (number of items), price, and lead time 123 Logist. Res. (2013) 6:25–41 35 4.1.2 The first round of negotiation importance of these two attributes using a cardinal scale from [1,10] [10]. Weights can be normalized onto [0,1]. In To illustrate the first round of negotiation triggered by the example, w and w denote the normalized weights, A B initial offers, and for the sake of clarity, we consider the respectively, for Parties A and B following special numerical example: Weights of Party A: Weights of Party B: Party A places an order of 500 units. The buyer (party A) does not propose a price, but the SC-agent (party B) has Weights | Normalized Weights Weights | Normalized Weights to. The SC-agent seeks to satisfy this order using price and Price: 5, w (P ) = 0.357 Price: 10, w (P ) =0.666 A t B t lead time as decision variables. Each agent starts the Lead-time: 9, w (LT ) = 0.643 Lead-time: 5, w (LT ) = 0.333 A t B t negotiation with an initial offer. The initial offer of party B (Supply Chain) is derived from the solution of an optimi- zation problem (see 2.2.2) which is considered next. 4.1.3 Ranges of the values of control variables (attributes) Because the SC has to meet the ordered quantity of 500 items (or to reject the order), it uses submodel I and selects The ranges (intervals) of the control variables can be a solution which is closest to the requested lead time 4. derived by solving an optimization problem (Fig. 7). For Figure 6 shows a simple numerical example. example, the SC-agent solves submodel I with the ordered Initial offer from party A (Retailer) Initial counteroffer from party B (Supply Chain) quantity Q (= 500) as input and gets. Ordered Quantity: 500 Offered Quantity: 500 This generates intervals K K K ðMaxÞ and t t t LT LT LT ðMaxÞ for the control variables, provided t t t Price: – Price: $5,500 the ordered set is Q (= 500) and this quantity Q (= 500) can t t Lead-time: 4 Lead-time: 6 be realized within submodel I. The SC-agent derives a price interval Pt B P B P (Max) t t from the given cost interval. In this numerical example, the requested lead time is smaller than the minimal lead time. The condition for submodel I, IðÞ WL þ IðÞ WL þ t 1 t 2 Therefore, the SC-agent proposes the minimal lead time 6 I ðWL Þ Q , is fulfilled for Q = 500. t 3 t t and a price corresponding to K (Max). b b The minimal-lead-time solution is w ¼ 400, w ¼ 11 21 If we do not require that the ordered set of Q (= 500) to be 100, wb ¼ 0 completely realized by submodel I, we can explore other with related costs: k 400 þ k 100 ¼ KðÞ Max ¼ 11 21 t combination of prices and lead times based on utility functions. We assume now that there are known intervals: Therefore, the initial-offer delivery of the whole order of PðÞ Min P PðÞ Max and t t t 500 items for a price of $5,500 with lead time of 6 weeks LTðÞ Min LT LTðÞ Max t t t was made. The lead time of 6 for the whole order is much longer 4.1.4 Utility functions than the requested lead time of 4 weeks. Therefore, a negotiation with respect to lead time and price is required Conditional utility functions: to avoid an impasse. Since negotiation parties have The method is based on the use of utility functions of different views, we introduce weights that express the each party A and B. Fig. 6 Computation of an initial offer; lead-time minimal solution 123 36 Logist. Res. (2013) 6:25–41 Then we get for the weighted utilities u ðP ; LT jQ Þ¼ w ðP Þ u ðP jQ Þþ w ðLT Þ t t t A t A t t A t u ðLT jQ Þ A t t u ðP ; LT jQ Þ¼ w ðP Þ u ðP jQ Þþ w ðLT Þ t t t B t B t t B t u ðLT jQ Þ B t t Of course, we cannot expect that the weighted utility functions are identical with the conditional (two dimen- sional) utility functions u , u . Details of this weighted A B utility approach are discussed in [10]. Fig. 7 Ranges of control variables In the following discussion, we consider simple exam- ples related to the numbers in the initial offers and the respective submodel I (Fig. 9). Joint utility functions: Multiplication of u ,u with normalized weights and A B adding the resulting curves of A and B results in ‘‘joint utilities of A and B with respect to each of the attributes price and lead time’’. Price u ðP jQ ¼ 500Þ¼ w ðP Þ u ðP jQ ¼ 500Þ t t A t A t t Joint þ w ðP Þ u ðP jQ ¼ 500Þ B t B t t leadtime u ðLT jQ ¼ 500Þ¼ w ðLT Þ u ðLT jQ ¼ 500Þ t t A t A t t Joint þ w ðLT Þ u ðLT jQ ¼ 500Þ B t B t t Using the weights given in Sect. 4.1.2 we get (Figs. 10,11): Price u ¼ 0:357 u ðP j500Þþ 0:666 u ðP j500Þ A t B t Joint leadtime u ¼ 0:643 u ðLT j500Þþ 0:333 u ðLT j500Þ A t B t Joint Bui and Shakun (1996) use a simple additive function to Fig. 8 Two-dimensional conditional utility function derive the maximum of the joint or social utility functions Each party defines a weighted and conditional utility in order to get a solution which maximizes the sum of function (under the condition of ordered quantity Q (= 500)). weighted utility functions with respect to one attribute. In our example (see Figs. 10, 11), the highest joint utility for u (P ,LT |Q ), in our example u (P ,LT |500) A t t t A t t price is given by any value of the interval [P *, P (Max)] t t u (P ,LT |Q ), in our example u (P ,LT |500) B t t t B t t (Fig. 10). The highest joint utility for lead time lies in the u , u : utility of party A, B of a price P and lead time A B t interval [LT (Min),LT *]where LT * = 5 in our example. t t t LT under the condition that the order is Q (= 500) t t For P * = 6,000 and P (Max) = 7,000, (P (Min) = t t t (defined over the ranges introduced in 4.1.3.) 4,800), we find from Figs 10 and 11 that any pair It is very important to understand, that in this example, ðÞ P ; LT 2½ 6000; 7000½4; 5 maximizes the joint utili- t t both utilities u and u depend on two variables P and LT . A B t t ties for prize and lead time as well and is, therefore, a It means that it is not possible to define the utility of a candidate for a compromise solution. particular value of the price variable without knowing the This result is based on the one-dimensional utility lead time values (Fig. 8). analysis prescribed above. If we deal with two-dimensional utilities, weights are, of Another approach would be to use the joint conditional course, not needed. two-dimensional utility functions by maximizing Weighted utility function: Max u ðP ; LT jQ Þþ u ðP ; LT jQ Þ A t t t B t t t We get the weighted utility functions, if we introduce one-dimensional utility functions u (P |Q ) and u (LT |Q ) with respect to (P,LT)over [LT (Min), LT (Max)] 9 t t t t A t t A t t * * [P (Min), P (Max)] and using the optimal solution (P ,LT ) of the party A (and analogously for party B) with respect to t t t t for further negotiation. In that case, weights are not needed. only one variable P or LT , respectively (Fig. 9). t t 123 Logist. Res. (2013) 6:25–41 37 Fig. 9 Examples of one- u (P | Q = 500) u (P | Q = 500) B t t A t t dimensional utility functions P (Min) P (Min) P (Max) P (Max) t t t t (Independent of lead-time) u (LT | Q = 500) A t t u (LT | Q = 500) B t t LT (Min)=4 5 LT =6 LT (Min)=4 LT =6 LT (Max)=8 7 LT (Max)=8 t t t t t t (Independent of price) will be delivered. (There are of course other possibilities, Price: [$6,000, $7,000] e.g., changing Q , splitting Q in subquantities with differ- t t lead-time: [4, 5]weeks ent lead times.) This concludes the first round of negotiation. The negotiation support system offers a compromise solution in case that the whole ordered quantity 500 items Fig. 11 Joint utility for lead time Fig. 10 Joint utility for price 123 38 Logist. Res. (2013) 6:25–41 4.1.5 Second round of negotiation using submodel III That means, the constraints for submodel III are now fulfilled. Party B tries to fulfill the required lead time = 5 The compromise reached in the first round of negotiation with minimal cost. seems to be very reasonable for the SC (party A), because After pre-processing using the retailer requirement: lead it is very close in terms of lead time compared to the initial time = 5, we can set: w = 0,w = 0. 21 31 offer. For party B, the new solution is still far away from Therefore, we get: w ðÞ t þ p ðÞ t þ p ðÞ t ¼ 500 11 11 21 the initial request (lead time of 6), because lead time of 5 is 0 w 400; 0 p 200; 0 p 400 11 11 21 not possible in the ATP scenario (submodel I). However, such that 10 w ðtÞþ 20 p ðtÞþ 15 p ðtÞ! Min: the retailer (party A) only knows that the interval 11 11 21 4 B LT B 5 has also for the SC maximal utility. For t The minimum-cost solution is: wb ðÞ t ¼ 400; pb example, party A offers a price of $6,000 for Q = 500 ðÞ t ¼ 0; pb ðÞ t ¼ 100; with lead time of 5. Therefore, let us assume that the and the assigned minimal cost value is: 10 400þ retailer agent does not accept the compromise in terms of 15 100 ¼ 4; 000 þ 1; 500 ¼ $5; 500 lead time but is willing to pay more instead. With this new solution, the SC-agent will perhaps accept That means the SC-agent is faced with an order of 500 the retailer offer without further negotiation, because it for a price of $6,000 and a lead time of 5 weeks. As seen in realizes maximal joint utility. Fig. 6, it is clear that a lead time of 5 is not possible within Let us now consider another case: d ¼ 6 (instead of submodel I. The SC-agent needs now to look for other d ¼ 5, see Fig. 12). submodels to find a solution. Then we get p ðtÞ¼ 0, and therefore, the optimization To illustrate the approach, we assume Q 21 problem becomes: ðÞ In Transport ðÞ t ¼ 0 and consider submodel III. The respective subnetwork looks as follows (Fig. 12): w ðÞ t þ p ðÞ t ¼ 500 11 11 We analyze this network under the requirements (1) 0 w ðtÞ 400; 0 p ðtÞ 200 11 11 Q = 500 and lead time = 5. With Inv(Prod) we denote t t such that 10 w þ 20 p ! Min: the total inventory of all producers at time t. 11 11 Then, we have The optimal solution is: wb ðÞ t ¼ 400, pb ðÞ t ¼ 100, and minimal cost is $6,000. ð1Þ IðÞ WL ¼ 400\Q ¼ 500; QðÞ In Transport¼ 0; t 1 t A In this scenario, the SC-agent cannot offer a price of InvðProdÞ ¼ 600 $6,000. Therefore, parties might need to reconsider the problem (i.e., re-examine the values of the decision out- and therefore: comes—price and lead time in their utility functions), and ðÞ 1 IðÞ WL þ QðÞ In Transportþ InvðProdÞ Q ¼ 500 another round of negotiation is needed. t 1 A t 1 t Fig. 12 Network for submodel III 123 Logist. Res. (2013) 6:25–41 39 The main reason is that the intervals for price and lead time of 5 is not possible given the overall ordered quantity, time are in a strong sense valid only for each of the sub- and a lead time of 4 is not feasible as well. Therefore, instead models. If we change the submodels, we need to define new of defining u ðLT j500Þ¼ 0for 4 LT 5, we could use B t t intervals and utilities. For example, within submodel I, a lead negative utilities for agent B (see Fig. 13). Fig. 13 Using negative utilities–utility function for lead time in the case of submodel I (lead times 4 and 5 are not possible within the scenario of submodel I) Fig. 14 An agent-based multi-attribute negotiation procedure for order promising 123 40 Logist. Res. (2013) 6:25–41 An utility function with negative values for an undesired References subinterval is better than a smaller interval, because it gives 1. Agatz NAH, Fleischmann M, van Nunen JAEE (2008) E-fulfill- more space for modified order fulfillment and negotiation. ment and multi-channel distribution—a review. Eur J Oper Res This illustrates the approach. Of course, the description 187:339–356 is not complete yet. After a positive result of the negotia- 2. Azevedo AL, Sousa JP (2000) A component-based approach to tion, a re-optimization of the whole supply chain is needed. support order planning in a distributed manufacturing enterprise. J Mater Process Technol 107:431–438 Also, other types of offers are possible, for example, 3. Ball MO, Chen C-Y, Zhao Z-Y (2004) Available to Promise. In: splitting the ordered quantity into parts with different lead Simchi-Levi D, Wu D, Shen M (eds) Handbook of quantitative times, etc. supply chain analysis: modeling in the E-business era. Kluwer, In Fig. 14, we outline an algorithmic approach of the Boston, pp 447–484 4. Bichler M, Kersten G, Strecker S (2003) Towards a structured combined optimization–negotiation method we have design of electronic negotiations. Group Decis Negotiat 12:311– introduced by a simple numeric example above. 5. Bixby A, Downs B, Self M (2006) A scheduling and capable-to- promise application for Swift and company. Interfaces 36(1):69– 5 Conclusion 6. Bui T (1987) Co-oP: a multiple-criteria group decision support system. Springer, NewYork With the increasing adoption of industry-wide supply chain 7. Bui T, Bodard F, Ma PC (1998) ARBAS: an action-resource systems, there is a real and urgent need to find negotiated based argumentation systems. J Manag Inf Syst 14(3):223–237 solutions between suppliers and customers instead of 8. Bui T, Gachet A, Sebastian H-J (2006) Web services for nego- tiation and bargaining in electronic markets: design requirements, rejecting orders that appear to be unable to match with proof-of-concepts, and potential applications to e-procurement. supply capacity. In the context of available-to-promise Group Decis Negotiat 15:469–490 (ATP) and capable-to-promise (CTP) functions, experience 9. Bui T, Lee J (1999) An agent-based framework for building has shown that order promising has turned out to be a decision support systems. Decis Support Syst 25:225–237 10. Bui TX, Shakun MF (1996) Negotiation processes, evolutionary rather difficult process. It is rather common that, in an systems design and NEGOTIATOR. Group Decis Negotiat inter-firm supply chain, finding a solution (that includes 5(4–6):339–353 delivery of available stocks and make-to-order production) 11. Bui T, Bosse T, Hempsch C, Sebastian H-J (2009) A multi-agent to an incoming customer order is not possible. More often simulation framework for automated negotiation in order prom- ising. In: Proceedings of AMCIS conference, San Francisco, than not, orders are rejected, thus failing customers and August 2009 incurring lost revenues. Acknowledging the impossibility 12. Bui T, Sebastian H-J (2010) Integration of multi-criteria decision of using optimization techniques to find an alternate solu- analysis and negotiation in order promising. In: Proceedings of tion, we have presented in this paper a multi-attribute, the 43rd HICCS conference, Kauai, January 2010 13. Chen C-Y, Zhao Z-Y, Ball MO (2001) Quantity and due date multi-utility model that can be integrated in a CTP solver quoting available to promise. Inform Syst Front 3(4):477–488 to explore alternate solutions. First, we set up a basic 14. Cohen M, Lee H (1988) Strategic analysis of integrated pro- supply chain model with the usual constraints: production, duction-distribution systems: models and methods. Oper Res warehouse replenishment, and delivery and distribution. 36(2):216–228 15. Conklin J, Begeman M (1988) gBIS: a hypertext tool for Next, we introduced the concept of weighted utilities for exploratory policy discussion. ACM Trans Inform Syst 6(4):303– the suppliers and buyers to help them explore a variety of feasible solutions based on price, lead time, and quantity 16. Dudek G, Stadler H (2005) Negotiation-based collaborative and demonstrated how the Pareto optimization can be used planning between supply chain partners. Eur J Oper Res 163:668– to search for solutions that maximize joint utilities. The 17. Emerson D, Piramuthu S (2002) Agent-based framework for proposed model has been implemented and tested using dynamic supply chain configuration. In: Proceedings of the 37th life-like scenarios. Results of our simulations suggest that Hawaii international conference on system sciences, IEEE, negotiation procedures did reduce the number of rejected 0-7695-2056-1/04 18. Fischer ME (2001) Available-to-promise: Aufgaben und Ver- orders and increase the overall revenues when negotiation fahren im Rahmen Des supply chain managament. S. Roderer concepts and algorithms are applied. A major benefit of Verlag, Regensburg this research is that, unlike many suggested approaches in 19. Frey D, Stockheim T, Woelk P-O, Zimmermann R (2003) Inte- the recent literature, the proposed solution to solve order grated multi-agent-based supply chain management. In: Pro- ceedings of the twelfth IEEE international workshop on enabling promising impasse is intuitive enough for supply chain technologies: infrastructure for collaborative enterprises participants to understand and appreciate and tractable 20. Fulkerson B (1997) A response to dynamic change in the mar- enough to be implemented in a multi-agent system. ketplace. Decis Support Syst 21:199–214 21. Fung RYK, Chen T (2005) A multiagent supply chain planning Acknowledgments This research is supported by the DAAD and coordination architecture. Int J Adv Manuf Technol 25: (German Academic Exchange Service) through the PPP-program. 811–819 123 Logist. Res. (2013) 6:25–41 41 22. Gallien J, Tallec YL, Schoenmeyr T (2004) A model for make-to- 37. Moses S, Grant H, Gruenwald L, Pulat S (2004) Real-time due- order revenue management, Working Paper date promising by build-to-order environments. Int J Prod Res 23. GNU Linear Programming Kit, Version 4. http://www.gnu.org/ 42(20):4253–4375 software/glpk/ 38. North M, Howe NCT, Howe HT, Vos R The Repast simphony 24. Grieger M (2003) Electronic marketplaces: a literature review runtime system. http://repast.sourceforge.net/ and a call for supply chain management research. Eur J Oper Res 39. Olhager J, Ostlund B (1990) An integrated push-pull manufac- 144:280–294 turing strategy. Eur J Oper Res 45(2–3):135–142 25. Interface, GLPK 4.8 J.: http://bjoern.dapnet.de/glpk/ 40. Richards HD, Dudenhausen HM, Makatsoris C, de Ridder L 26. Jennings NR, Sycara K, Wooldridge M (1998) A roadmap of (1997) Flow of orders through a virtual enterprise—their proac- agent research and development. Int J Auton Agents Multi-Agent tive planning and scheduling, and reactive control, Comput Syst 1(1):7–38 Control Eng J 173–179 27. Julka N, Srinivasan R, Karimi I (2002) Agent-based supply chain 41. Rupp T, Ristic M (2004) Determination and exchange of supply management-1: framework. Comput Chem Eng 26(12):1755– information for co-operation in complex production networks. 1769 (15) Robot Auton Syst 49:181–191 28. Karacapilidis N, Papdias D (1999) Computer-supported argu- 42. Sadeh N, Hildum D, Kjenstad D (2003) Agent-based E-supply mentation and collaborative decision-making: the HERMES chain decision support. J Org Comput Electr Commer 13(3): system. Inform Syst 26(4):259–277 225–241 29. Kellett PM (2007) Conflict dialogue: working with layers of 43. Shin K, Leem CS (2002) A reference system for internet based meaning for productive relationships. Sage, New York inter-enterprise electronic commerce. J Syst Softw 60:195–209 30. Kilger C, Schneeweiss L (2005) Demand fulfillment and ATP. In: 44. Stadtler H (2005) Supply chain management and advanced Stadtler H, Kilger C (eds) Supply chain management and planning—basics, overview and challenges. Eur J Oper Res 163: advanced planning, 3rd edn. Springer, New York, pp 179–195 575–588 31. Kilger C, Meyr H (2008) Supply chain management and 45. Tan GW, Shaw MJ, Fulkerson B (2000) Web-based supply chain advanced planning, 4th edn. Springer, New York management. Inform Syst Front 2(1):41–55 32. Larsson S (2002) Issue-based dialogue management. Working 46. Vidal CJ, Goetschalckx M (1997) Strategic production-distribu- paper, Department of Linguistics, Goteborg University, Sweden tion models: a critical review with emphasis on global supply 33. Louta M, Roussaki I, Pechlivanos L (2008) An intelligent agent chain models. Eur J Oper Res 98:1–18 negotiation strategy in the electronic marketplace environment. 47. Wu H–H, Liu J-Y (2008) A capacity available-to-promise model Eur J Oper Res 187:1327–1345 for drum-buffer-rope systems. Int J Prod Res 46(8):2255–2274 34. Makatsoris C, Chang YS (2008) Beyond CRM: a system to 48. Yan Y, Yen J, Bui T (2002) A multi-agent-based negotiation bridge the gap between the customer and the global manufac- support system for distributed transmission cost allocation. Int J turing supply chain. In: Proceedings of the 41st Hawaii interna- Intell Syst Acc Financ Manag 10(3):187–200 tional conference on system sciences 49. Zhang X, Ou J, Gilbert SM (2008) Coordination of stocking 35. Meyr H (2004) Supply chain planning in the german automotive decisions in an assemble-to-order environment. Eur J Oper Res industry. OR Spectrum 26:447–470 189:540–558 36. Moodie DR, Bobrowski PM (1999) Due date demand manage- 50. Zhao Z, Ball MO, Kotake M (2005) Optimization-based avail- ment: negotiating the trade-off between price and delivery. Int J able-to-promise with multi-stage resource availability. Ann Oper Prod Res 37(5):997–1021 Res 135:65–85
Logistics Research – Springer Journals
Published: Dec 1, 2012
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.