Cooperative Booking of Time-Dependent Travel Services

This paper proposes a new collaboration protocol for procuring travel services within tight deadlines, where personal assistant agents, loaded onto travelers’ mobile phones, collaborate with each other to procure services available at particular time slots, such as hotel rooms, flights and concerts. A simulation model of a travel services marketplace has been built to test if the new protocol is to bring improvements to the success rate for obtaining travel services. Experiments with the model demonstrate that the new collaboration protocol increases the number of successful negotiations. 1. Travel Services through Your Mobile Phone Globalization of trade and tourism increases the number of tourists and business travelers. This means increased consumption of local services at the destination of travel, such as booking hotels, organizing local travel and car hire, and booking tickets for concerts. Automatic procurement of such services will give travelers greater flexibility in their itinerary using the capabilities of a global network of agents such as Agentcities / openNet (www.agentcities.org). The procurement of such services can be conducted by personal assistant agents (PAAs), which can automatically negotiate the best offer according to their traveler’s preferences[2][6]. A traveler would create their PAA before the trip and parameterize it with the general requirements for the trip, such as a type of room, draft itinerary, etc. They would then load the agent onto their mobile phone, thus ensuring global interoperability at the level of GSM and GPRS roaming. The autonomous operation of agents would minimize the need for sophisticated interactions between software and traveler once on the road. At the level of semantic interoperability, a global distributed platform for the procurement of travel services can be provided by Agentcities and its successor openNet. UMIST has started to build a demonstrator of such a service, detailed in Section 2 below. One major factor impeding the successful use of such a technology is the success rate of finding an appropriate service under the constraints of limited time and limited negotiation capabilities. A personal assistant agent, for example, may attempt to book a hotel room upon arrival at the destination, when the time available for this would be several hours. Even when the hotel is to be booked before commencing the journey, completing this part of the booking in the quickest possible time would allow the booking of other elements of the trip which are dependent on it, such as flights or car hire. This problem is further detailed in Section 3 of this paper, where we focus on a particular type of travel services associated with fixed time slots, such as hotel rooms and flights. This paper posits that this problem can be alleviated by the use of a novel collaboration protocol, where personal assistant agents share information about “second best” offers which they are about to reject. An early version of this protocol has been shown to improve success rate within the context of procuring Web Services [1]. Section 4 below describes a version of this collaboration protocol modified to reflect the specific requirements of the travel services scenario. We have created a simulation model of a travel services marketplace to explore the effect of the proposed collaboration protocol on the number of successful room bookings within realistic time limits. The results are presented in Section 5, and they demonstrate an increase of the successful bookings in comparison with a noncollaborative version of the marketplace. The paper concludes with discussion and review of relevant work in Section 6. 2. The Vision of Automatic Procurement of Travel Services It is an early Friday in June 2005. On touchdown at Manchester Airport, you switch on your mobile phone. On its calendar, you have specified an 11am business meeting at the University of Manchester and have noted the northern Lake District as your weekend holiday destination. Your mobile contains an intelligent personal assistant agent (PAA). It knows that you have not booked any transport nor accommodation for the weekend. Whilst you go towards passport control, your PAA: • Connects to the Manchester Agentcity, which acts as a local gateway providing a number of travel services as detailed below. • Checks the train timetable from the airport and calls a radio-taxi since it has detected that otherwise you will be late for the meeting. • Checks the rail connections to the northern Lake District and discovers they would not work for you, so it hires a car from EuroCarHire on a bargain price for the weekend. The car will be delivered to the University reception after 2pm. • Contacts the pool of hotel room suppliers in the northern Lake District, and asks for quotes for a single room for Friday to Sunday evening inclusive. Price negotiation is guided according to a pre-specified bargaining strategy for each party. There are strict deadlines for all parties – you need a room for the night, and hotels need to sell the tonight’s availability of their rooms before midnight when the value of this availability reaches zero. • A room conforming to your requirements is booked and directions for reaching it are stored in your mobile for passing onto the hired car’s GPS-based guide. Various complications of this scenario (room cancellations, changing bookings, etc) have been explored in the literature (e.g. FIPA Personal Travel Assistance Specification [3])to showcase the deliberative reasoning and goal-driven behavior of intelligent agents. For the purposes of this paper, however, we are only interested in one aspect of the service procurement process, which is clearly evident even in the simplified case above: negotiations for travel services are conducted under strict limits in terms of time and scope. Indeed, we can safely assume that a PAA can only conduct a limited number of concurrent negotiations because of limited computation power and comparatively expensive communication bandwidth. Negotiation between service providers and service requesters will in principle be linked to price and quality attributes. A service requester agent will negotiate with several service supplier agents and vice versa. Under ideal circumstances, a service requester will secure several acceptable offers which it will hold onto in the hope of securing an even better offer. When the time runs out, a service requester will commit to the best offer available. A detailed description of this negotiation process is provided in Section 3 of this paper. The system will use the distributed agent-based infrastructure built on the Agentcities.NET project, which consists of a number of platforms (agentcities) dispersed throughout four continents. The personal agent on the mobile phone will be a part of this global system and will therefore interoperate with the service provider agents contained at the destination agentcity. Appropriate interfaces will link the system with non-agent information sources such as hotels and hotel brokers, which will be represented within the agentcity by service provider agents. This setup is illustrated