Trip-planner: An Agent Framework for Collaborative Trip Planning

Planning a trip to a particular destination can be a time consuming and daunting task. There is a wide range of information sources available, each claiming to present the best deals and the most complete tourist information. This information as well as the current weather, budget constraints, time constraints, and personal preferences all need to be taken into account when planning the perfect trip. It would de nitely be nice to have an agent handle the vast amounts of information gathering and interpretation necessary during trip planning. The ideal system would also allow the user to interact with the system to take advantage of the unique strengths of each. This suggests a mixed initiative system. An example travel situation best illustrates the need and usefulness of a mixed initiative system in the travel planning domain. Suppose a traveler is planning on taking a trip to Chicago, Illinois. The trip will last ve days, the maximum cost can be three thousand dollars, and the preferred airline for the trip is United Airlines. The traveler would like to watch both a Chicago Cubs' game at Wrigley Field and the Chicago Symphony Orchestra. A round trip plane ticket needs to be purchased, a hotel booked, an itinerary planned (with tourist sites and restaurants), and car rented, if necessary. All of these tasks need to be done according to the budget, time, and personal preference constraints of the traveler. Problems could arise when trying to schedule the two events the traveler wishes to see. Scheduling a successful baseball outing requires that baseball is in season, the Cubs are in town, tickets are available, and that the weather forecast does not call for rain. The orchestra outing requires that the orchestra is in season and that tickets are available. It is possible that the orchestra season and the baseball season do not coincide and it is therefore impossible for the travel agent to completely satisfy the traveler's preferences. The same argument could apply to any of the constraints. At this point it becomes necessary for the travel agent to collaborate with the traveler to either relax the constraints, select from a series of agent provided suggestions, or have the traveler specify a new choice. Hence, a mixed initiative system is preferred over a system that nishes the trip planning in one user interaction. System Architecture and Functionality System Architecture From the example introduced above, one can argue the need for a system that not only collaborates amongst itself, but also collaborates with the human user during the planning process. An architecture that implements this can be seen in gure 1. The architecture can be broken down into two distinct components, the personal travel assistant (PTA) and the facilitator agents. An example of a centralized travel agent and a series of facilitator agents can be found in Ndumu et al (Ndumu 1998).