Companies involved in air cargo have lagged in recent years to adapt to the needs of their customers. This paper reports the use of software agents as a decision support mechanism for electronic commerce to alleviate this problem. We use software agents as a means to maximize the market value chain and to optimize market transactions between market players. We suggest a DSS approach that promotes the decentralized design and implementation of software agents to support market transactions. Autonomous software agents are designed to perform routine operations automatically and to support market agents to fulfill their trading tasks (from information gathering, to negotiation and settlement).

Keywords: Internet commerce, Decision support systems

1. Introduction

Development of electronic markets (EM) has recently gained increased interests among both academics and practitioners. Information technology is expected to increase market transparency, transaction speed and allocation efficiency while reducing biased decisions and time-intensive routines. However, current EM can at best be described as computerized transactions with focus on automated information delivery (e.g., WEB homepages as information centers) and on data transactions (e.g., WEB browser-based ordering, sales auctions, and customer support). As such, current EM implementation fails to take advantage of the well-proven potential of decision support technology. Furthermore, the design of existing EM follow a centralized approach with all transactions being performed at the server site with traders accessing the EM via remote clients. In this paper, we propose a special type of EM that fundamentally departs from the microeconomic model of market competition (with perfect atomicity among buyers and sellers) and, as will be discussed later, which requires a unique mode of doing business. To address this particular type of electronic market transaction, we propose the use of software agents.

To facilitate the discussion of the use of software agents in the design of EM, Section 2 briefly describes the air cargo business, its market characteristics, and the specific requirements for an effective electronic air cargo market. Section 3 presents a classification of software agents and proposes an agent-based approach to design EM. Section 4 highlights the key implementation aspects of the electronic air cargo market and is followed by a short illustration (Section 5). Lessons learned and directions for future research are presented in the Section 6.

 

2. The Air Cargo Business as an Internet Commerce

Until recently, cargo business has been economically insignificant for most major airlines compared to that of passenger transportation. The volume of goods to be shipped by air is however predicted to double by the year 2000 with approximately 480 carriers worldwide [Alt and Cathomen, 1995]. Airline companies have started to adapt their strategic goals accordingly, i.e., to look at information technology as new mechanisms to respond to their forwarders more competitively. Potential benefits of such an effort would be even greater for the forwarders. As communications by telephone and fax still dominate the coordination process between a forwarder and its airlines, only a very small choice of airlines can be considered for a shipment. Consequently, forwarders tend to stick to a small number of airlines they have successfully done business with. Furthermore, market transactions tend to be biased. More often than not, the choice of an airline transporter is dictated by personal preferences of the officer-in-charge of the requesting forwarder. Thus, the number of business relationships a forwarder can handle during one transaction is very small, resulting in inefficient market transaction (e.g., opportunity loss of business, opportunity gain derived from a wider market). Table 1 highlights some of the major differences between the current business process and the EM-based approach.

Table 1. Migrating Air Cargo Business to Electronic Market Place

For most of the major airlines, cargo business has been insignificant compared to that of passenger transportation. The volume of goods to be shipped by air is, however, predicted to double by the year 2000 with approximately 480 carriers worldwide [Alt, Cathomen, 1995]. Airline companies have started to adapt their strategic goals accordingly, i.e., to find new computer-supported market mechanisms to competitively respond to their forwarders.

The business process itself will be affected by the introduction of computer support. Potential benefits of such an effort would be even greater for the forwarders. So far transactions in the cargo business have relied mainly on informal relationships between officers-in-charge on either side of the market. Long-term agreements have dictated the general conditions, including prices, under which bilateral negotiations have taken place. As communication by telephone and fax still dominates the coordination process between a forwarder and its airlines, only a very small choice of airlines are considered for a shipment. Consequently, forwarders tend to stick to airlines they know and have successfully done business with before. As such, preferences of officers-in-charge of the requesting forwarder dominate the choice of an airline transporter instead of the goal to find an efficient fit between supply and demand. Thus, the reduced number of interaction relationships a forwarder can handle during one transaction is very small, resulting in potential loss of business.

A special demand condition of the cargo business is tracking and tracing of a shipment, so that the forwarder always knows where his shipment is located. This is currently achieved by telephone and fax calls performed by the officers-in-charge.

The idea of a virtual corporation (see for example [Davidow and Malone, 1992], [Gamble, 1992], [Goldman, 1994], [Klein, 1994]), connecting partners, whose combined core competencies guarantee the best possible result for a transaction, has not been picked up in air cargo business, yet. It is this type of organization that can improve service quality and - combined with modern IT - can minimize ineffective communication and coordination. Time lags caused by call-backs, misunderstandings and double-checking can be avoided. In addition, integrated computerized support will prevent business partners from repeated data inputs that often cause errors.

When comparing the needs of an electronic air cargo market to those of other electronic coordination mechanisms, e.g., like electronic stock markets or computerized reservation systems (CRS), a number of major differences can be made.

The stock markets as well as the computerized reservation systems in the passenger transportation sector are characterized by a great number of potential buyers and sellers germane to the concept of many-to-many relationship in database design. In the air cargo market, however, the uniqueness of each single shipment requires a special treatment so that it is usually the sender who will call for offers for the shipment in question. As this is done with all the shipments of a particular sender as well as the shipments of all the other senders on the market, we can portray this situation as a multiple one-to-many relationship. In other words, the relationship between airlines and forwarders will not be a perfect market in the neoclassical sense. We therefore contend that existing coordination mechanisms as applied in common competitive markets do not suit well the specific needs of air cargo business.

In a centralized market, information, negotiation, and settlement procedures are controlled by a centralized system. A CRS, for example, consists of a core system where all requests and actions are executed.

Common competitive markets are centrally structured as well. All data are typically stored in a central database. Access and retrieval mechanisms are provided by the system owner, resulting in enormous system investments. A centralized market structure limits flexibility of negotiation considerably since no direct, personalized communication is possible.

A decentralized market system coordinates market transactions by locally driven procedures, e.g., by systems implemented on the supplier’s side and on the customer’s side.

Data are stored in the decentralized systems. Communication and coordination take place in a one-to-many way, as opposed to a many-to-many communication in a centralized competitive market. This leads to smaller investments and greater flexibility.

As the air cargo market is not a perfect market, the structure of a support system has to adapt to its particular characteristics. Due to the fact that the air cargo business is a customer driven market dealing with customized services, the initiative has to come from the forwarders, addressing not all potential suppliers but only those that can meet the demand. Additionally, the special needs in the settlement phase call for a market structure where bilateral relations can easily be realized as well.

In order to communicate with market partners other than via telephone or fax, the Internet can be applied without major infrastructure investments. With this as a working assumption, we argue that smart and autonomous software programs called software agents can accomplish numerous routine tasks and, when needed, provide "intelligent" decision support as well.

As mentioned earlier, the needs of an air cargo market resemble the concepts of virtual organizations, since both forwarders and airlines enjoy a great degree of autonomy. Their interaction can very well be described as highly temporal business relationships. Decentralized structures provide a high degree of flexibility for the market partners and enable more personalized communication for bilateral business relationships. Decentralized coordination procedures can be supported easily by software agents guaranteeing the high flexibility desired. The implementation of these agents on the suppliers' and customers' sides is by far less expensive than the installation of a centralized clearinghouse.

1. Electronic Market Architecture Using Software Agents

1. Software Agents

Software agents are the result of a paradigm shift in developing application software. Software is no longer regarded as a tool. Rather, it is considered as an autonomous assistant to the users - hence the word Personal Assistant in the software engineering literature. In other words, the traditional approach to software development is a reactive one in that the computer is programmed to react to the user’s instruction. Instead, the software agent approach is a pro-active one, in that the user specifies what he/she wants the computer to accomplish, and the latter performs tasks on behalf of the users the way it does best. By analogy, a software agent mimics the role of an intelligent, dedicated and competent personal assistant.

An intelligent software agent should be able to perform the following tasks:

• Search and retrieve information,
• Manage information overload,
• Perform repetitive office activity,
• Assist in mundane personal activity, and
• Refer to domain expert.

To accomplish these tasks, it should possess the following general abilities [Kalakota et al. 1996]:

• Independence,
• Learning,
• Cooperation,
• Reasoning, and
• Intelligence.

A generic architecture of the software agent, adapted from Bui [1997] is depicted in Figure 1. For the air cargo business, and as described later in more details, the user (e.g., forwarder or airliner) interact with their agents via the User-Agent Interface or Adapter. The interface role is to pass the agent’s data in the user’s format. The agent uses its embedded procedures (Agent’s processing engine) and its data (Agent’s Repository) to perform tasks and exchange information via the View/Framework (e.g., electronic mailbox). Agents are seamlessly connected together both horizontally and vertically.

Figure 1. A Generic Structure of a Software Agent

1. Software Agents and Transaction Flow

1. Air Cargo Transaction Flow: A Software Agent Perspective

Transactions in the air cargo business involve several business partners. The most important are airlines and forwarders. Other stakeholders implicated in the process are typically banks, custom agencies, and airport companies. In the following, we will concentrate on the supplier-customer-relationship between airlines and forwarders. The architecture of cooperating software agents is introduced (see Figure 3).

 



Figure 2: Software Agents Supporting Transactions between Airlines and Forwarders

In a networked organization, software agents can be regarded as teleworkers as they are assigned specific tasks and sent to the field to interact with other software or human agents [Bui and Sankaran, 1997]. In an electronic air cargo market, three clusters of agents can be distinguished according to the ownership. Forwarders as initiators of a coordination process own user agents providing the interface between the system and the officers-in-charge at the front end. At the back end there are supervising agents that control the processes and intervene when necessary. Also, as back end support coordination agents perform the inter-organizational transaction tasks.

The airlines' officers-in-charge access the system through user agents. Inquiries are handled by the airlines' reply agents. Retrieval agents implemented in a public database environment provide access to routing information and suggested rates. These databases are managed by neutral organizations, like for example IATA (International Air Transport Association).

2. Software Agent Activities

Market transactions are classically divided into three main phases: information, negotiation and settlement phase (compare [Williamson, 1985], [Schmid/Zbornik, 1991], [Schmid, 1993]). In the following, these phases are subdivided (see Table 1). In the table data input, market overview and offer collection can be assigned to the information phase. Offer evaluation and auction can be assigned to the negotiation phase. Document dispatch and tracking and tracing can be assigned to the settlement phase.

Table 2: Processes and their Agents

Marks in brackets denote agent support not necessary but optionally involved in the
process.

An entire transaction with software agent support can be detailed into several steps, refining the activities (represented by columns in Table 2).

1. After the forwarder received an order to send a shipment he calls the user agent, thus initializing a transaction.
2. The user agent receives fundamental specifications of the shipment from the officer-in-charge (e.g., size and weight, destination, time window of departure, time window of arrival, documents to be included, additional attributes like "live animals", "poisonous goods", "dangerous goods", etc) using a standardized form.
3. The user agent asks for the human agent's preferences to be considered. Preferences could be the exclusion of certain airlines, airports, routings, etc. In addition, qualitative requirements can be specified by weighting the needs for punctuality, reliability, minimization of reloads, etc. The officer-in-charge specifies the deadline for offer collection.
4. The user agent initializes the supervising agent giving information on this deadline.

Market Overview

5. The user agent instructs the coordination agent to retrieve information about possible routings and corresponding airlines meeting the constraints of steps 2 and 3.
6. The coordination agent hands over the inquiry to the retrieval agent.
7. The retrieval agent selects appropriate routings from the ABC database and returns routing and airline information to the coordination agent.

Offer Collection

8. The coordination agent sends the calls for offers to the reply agents of the selected airlines.
9. The reply agents automatically provide offers for standard shipments. In more complicated cases they return their offers after checking with the human officer-in-charge.
10. The supervising agent keeps an eye on the given deadline. In case of no reactions it causes the coordination agent to send reminders to the reply agents concerned.
11. The coordination agent sends reminders to the reply agents trying to get the required information.
12. The supervising agent collects the results from the coordination agent. As soon as all airlines answered the call for offers the information phase is terminated. If this goal cannot be achieved, the deadline introduced in Step 3 terminates this phase. The supervising agent informs the user agent about the offer collection results.
13. The user agent forwards the information to the officer-in-charge. Details included are the number of airlines asked for an offer and the number of airlines that answered the call for offers. In the same step, the human agent (i.e., the officer-in-charge) is requested to fix the deadline for the following auction.
14. The user agent forwards the deadline to the supervising agent.

Offer Evaluation

15. The user agent triggers the coordination agent to calculate net value prices for each offer considering quantitative and qualitative factors.
16. The coordination agent calculates net value prices and produces a ranking of the offers including information about the relative differences between the offers.

Auction

17. The coordination agent distributes this ranking to all of the airlines involved.
18. The airlines' reply agents receive the ranking and contact the user agent handing over this information.
19. The airline's user agent asks the officer-in-charge whether or how to adapt the offer to the market situation. The result is given back to the reply agent.
20. The coordination agent collects new offers from the reply agents and recalculates net value prices.
21. The supervising agent terminates the negotiation process and instructs subsequently the coordination agent not to accept any further offers.
22. The coordination agent hands over the final ranking to the user agent.
23. The user agent displays the results to the officer-in-charge and leaves the decision about the market partner for this transaction to the human agent. This closing of a contract terminates the negotiation phase.

Document Dispatch

24. In the settlement phase the forwarder's user agent collects data for the necessary documents from the officer-in-charge and hands them over to the coordination agent.
25. The coordination agent transfers the documents to the recipients, e.g., airlines involved, custom agencies, or handling companies.
26. Documents are reviewed and approved by the partner's reply agent.
27. The whole process to include the order of the documents is handled by the supervising agent.

Tracking and Tracing

28. At the beginning of the actual transport process the user agent hands over milestones of the shipment to the supervising agent. These data are fed into the system in step 23 when the decision of the routing is made.
29. According to these milestones the supervising agent triggers the coordination agent to request a status report from the airline.
30. The coordination agent demands this status report from the airline's reply agent and delivers it to the supervising agent.
31. If expected and actual status differ the supervising agent sends this information to the user agent, which reports this exception to the officer-in-charge and saves the information in the local quality database.
32. Once the last milestone verified by the coordination agent, the user agent informs the human agent about the closing of the transaction.

1. Implementation

The heterogeneity of the applied systems in air cargo led to the decision to use Smalltalk as a programming language in the ParcPlace© version of VisualWorks© 2.52 with compilers for all different platforms, including MS Windows© 3.x, 95 and NT, Sun Solaris© , OS/2© and AIX©.

The system's interface needs to be user-friendly. This is why prototype implementations have adopted the common paper forms, applied in the cargo business, to the computer system. First experiences show that this requirement has been achieved. To assure access to all common databases, ParcPlace© offers various gateways so that existing software can easily be integrated.

To pave the way for a fast diffusion of decentralized market mechanisms, the communication system of our choice is the Internet thanks to its interoperability. The gateway for information to be sent via Internet out of the local VisualWorks© environment can be realized with VisualWave©. This tool produces html-code from given data. Forms that need to be filled in are provided with the help of CGI (common gateway interface). To date, prototypes of user and coordination agents have been implemented with the tools mentioned (see [Robra, 1996] and [Wiesner, 1996]).

 

3. An Example

 



When a forwarder receives an order from a client to send a shipment, he needs to feed the relevant data into the access system of the electronic air cargo market. The user agent leads him through this process. Data screens are designed paper forms the interface for this task is shown in Figure 3. Buttons provide access to more detailed, specialized information. The button On Market allows this order be offered to other forwarders on a market with horizontal coordination structures. This can be necessary when the forwarder has no resources left to settle the order himself due to a lack of manpower or competence.

The menu Dispatch leaves functions disabled which - according to the transaction process - are not yet next in line, to assure that a certain order is kept.

 



In the next step the officer-in-charge defines the special requirements of a shipment concerning place and time information for departure, destination and arrival (see Figure 3). The screen also offers access to more detailed specifications considering special qualitative issues (Further Requirements) and time restrictions for the coordination process (Answer Time) instructing the supervising agent to close offer acceptance at a certain time.



According to the special needs of a shipment, the forwarder can indicate his preferences for qualitative factors like punctuality, duration or reliability. Mandatory criteria about which countries, cities and airports to avoid, as well as which airlines not to consider can also be fixed here. Additionally, the number of reloads might be of importance for the choice of a market partner for a certain shipment. The system's supervisor can integrate additional quality criterion (the last scroll bar in Figure 5) as he sees fit.

When the user agent has received all necessary data from the officer-in-charge, he inquires basic routing information. The coordination agent forwards this information demand to the ABC database by communicating with the database retrieval agent. The coordination agent then contacts the airlines that could possibly handle the shipment and asks their reply agents for offers, thus initializing the auction during the negotiation process. This process continues until final offers from the airlines are accepted and recalculated into the net value price. This results in a final ranking which is delivered to the user agent and processed for the officer-in-charge (see Figure 6). He is the one to finally decide which offer to pick.

 



According to the milestones determined by the routing, supervising and coordination agents dispatch the documents to the places where and when they are needed. Only in special situation interference of the officer-in-charge is necessary. Tracking and tracing are performed autonomously as well, handing over exception reports to the officer-in-charge when necessary. All relevant data are stored in the quality database without any time expenditures on the side of the officer-in-charge.

4. Lessons Learned

The idea of an electronic market for the air cargo business was adopted quickly by all the parties involved. The decentralized design is an attractive concept for both the forwarders and the airlines. Of particular interest for all market agents is the integration of tracking and tracing capabilities, a very time consuming process in the actual communication between forwarders and airlines.

1. Impacts on Forwarders

The most apparent advantage for the forwarders is the reduction of telephone and fax calls. Transaction costs can be reduced as time can be saved for routine communication on information collection, negotiation and status requests. Additionally, cycle times for transactions are cut back through integration of the whole process and the ability of software agents to communicate with other agents much more efficiently compared to personal communication. Data, once put into the system, can be made available for all market partners avoiding multiple data input. The fact that communication can be simplified through the implementation of software agents allows expansion of market partners. The increase of market transparency enlarges the choice for the forwarders, resulting in better matching of problem solutions and better prices due to increased competition.

The approach to electronically support negotiation on prices and services was accepted reluctantly. Forwarders fear that the computer system might take over their jobs and doubt that their knowledge about the influencing factors on the decision about the market partner can be properly represented. On the other hand, the forwarders are willing to accept a decision support tool that does not deprive them from free decisions but structures influencing patterns on a certain shipment and helps finding an appropriate market partner.

2. Impacts on Airlines

On the airline's side, transaction costs are expected to decrease as well since repetitive data inputs can be avoided. This saves time and reduces the risk of input errors. Since most of the communication with forwarders takes place during the information exchange phase (offer inquiries) and during the settlement phase (tracking and tracing), the implementation of software agents that take over these tasks increases efficiency on the supplier side.

After a hesitant first reaction on the use of an electronic, the airlines realized that an electronic market is a valuable tool for them. Market transparency enables airlines to better advertise their core competencies since qualitative issues like reliability or punctuality gain in importance compared to the status quo. In addition new markets can be developed and the presence in an electronic air cargo market will lead to competitive advantages when the added services are useful for the forwarders.

5. Summary

Considering the airlines-forwarder-relationship and neglecting the presence of other market partners (banks, customs, etc.) of an electronic air cargo market advantages for both sides, namely suppliers (airlines) as well as customers (forwarders) seem to be obvious. An enhanced market transparency will increase efficiency from the customers’ points of view and give the airlines a better chance to acquire new customers in and outside their home countries where they have traditionally been strong. Instead of regional preferences the forwarders’ choice will move towards the more suited supplier for their current requirements. The airlines' success will more and more depend on their service qualities and less on market inefficiencies.

It is expected that the introduction of Internet commerce with decision support technology will decrease transaction costs while increasing transaction speed. Time consuming tasks can be moved from human to software agents, saving time on the one hand and decreasing biased decisions on the other hand. Due to the distributed architecture, investments are minimal. The implementation of an electronic market in the air cargo business completely changes existing structures, creating new opportunities for both market sides without fundamentally changing business processes.

6. References