Student Project Proposal:
Worldwide Competitive Electronic Business Agents
Office: INR 218, 236
Tel: +41 21 693 6679
Office: INR 240
Tel: +41 21 693 6711
Project TypeDiploma project for one student.
The future trends of current economies cannot be entirely predicted but global electronic markets will probably play an important role in it. Such markets will bring together large numbers of producers and consumers represented often by software agents. These systems will be highly dynamic - the price for one good could change in a matter of seconds.
Successful players in these markets will be those that will be able to:
- take the best decision regarding the kind of actions to take (based on possibilities to integrate available services)
- take the best decision regarding the choice of a partner for a given kind of action (based on the reputation of partners).
A number of existing, or in development technologies will possibly be used for achieving this vision. Web Services are a high-profile technology that will enable applications to talk to each-other. Nowadays there are intense research efforts around them as they are seen as the possible next revolution in computing. Currently the technology matures as it is endorsed by all the major industrial players (IBM, Sun, Microsoft) but naturally the approach is bottom-up starting from the lowest and simplest layers and going up on the complexity stack.
Intelligent web services address issues in this higher-level part of the technology spectrum like: service descriptions, service discovery, service integration, service orchestration and service coordination.
Application integration is the process trough which existing applications are glued together to provide new functionality. Current statistics show that software companies spend 80% of the development time for application integrating and only 20% of the time for building new applications. In an Internet based on web services where millions of deployments will be available, integration will become a key critical issue.
Last, but not the least, this new model of electronic interaction is much more susceptible to fraud and deception. Humans exhibit in social interactions behavioral characteristics that are favoring honest behavior. Having purely electronic interactions lowers the social barriers for cheating for humans and thus different mechanisms should be used in order to ensure sufficient cooperation levels in the business environment. Dynamic trusting decisions, based on the agents' reputation (defined as information about the agent's past behavior), provide promising solutions for the problem of security in open distributed multi-agent systems. E-commerce systems like eBay and Amazon implement successful reputation mechanisms which were partly credited for the businesses' success.
Another important facet of reputation is that it also allows to characterize the quality of a service (say provided by an electronic agent) before actually experiencing (and paying) for that service. This property can be used in order to make optimal decisions for service integration or service composition solutions.
- Distributed File Sharing System. There are a number of users in the environment each having a set of resources (music files for example) that she is willing to share for a certain cost with her peers. In such a system, a simple task would be to allow one user to download a file from another user. For this, a user A has to locate the user B who has the desired resource, initiate a negotiation with B, and finally make the transfer: the money is transferred from A to B, and the file is transferred from B to A.
However, a more evolved version of the system will have to accommodate more complicated needs of the users. For example, a user might want to download at once a whole album. This is a basic example of service discovery and integration since (1) first the user needs to discover all the songs in that particular album, (2) the user has to decide on a set of users who have all the required songs, (3) negotiate the price and (4) download the resource. The problem becomes even more complicated if we assume that music files have various quality levels, and different accessibility (in terms of download time for example). This is were the reputation comes in. By taking into consideration reputation information, at step (2) a user can optimize the solution.
- Distributed Delivery Application. Imagine a geographical territory in which a number of agents can deliver packages. Different agents might have different delivery destinations, different prices, and different service quality levels (in terms of delivery time, reliability, customer support, etc). As in the example above, an agent who wants a package delivered from A to B, has to discover the optimal (considering reputation) set of agents that can do the whole job, negotiate a contract and do the exchange: payment against delivery.
- T 1. Scenario Analysis and determination of requirements
- T 2. Design
- T 3. Implementation
- T 4. Final Wrap-up
Ion Constantinescu Last modified: Thu May 13 2004