NSF/USDOT: Context-Aware Software Agents for Multi-Modal Travel

该项目致力于在高度个性化的智能高级旅行信息系统(ATIS)中使用软件代理，为旅行者提供集成的实时交通建议。旅行者将使用功能越来越强大的个人和车载无线设备，这些设备可以监控用户的背景、位置和运动(通过GPS和其他传感器)、偏好、优先级、日程安排和情况。美国农业部的研究表明，根据旅行者的特定需求和情况定制的信息将有助于做出更好的交通决策，调整时间表、路线和出行模式(汽车、公共汽车、火车等)，并与其他旅行者、同事和家人进行协调。这应该会提高整个交通系统的便利性、利用率和安全性。软件代理是松散耦合的软件元素，非常适合高度灵活、动态和复杂的系统。代理是自主的，执行任务议程，并通过消息与其他代理协作。ATIS代理代表旅行者的目标、偏好和计划，监控旅行，调整计划以适应不断变化的情况，并确定指导和委托行动的优先顺序。研究目标是确定如何有效地创建、演化和使用旅行者环境、计划、路线和服务的模型，使多个智能体能够自主地维护他们的个人信息，并与其他智能体协商和组合信息以获得有用和及时的旅行建议。研究成果包括：1)用户和旅行信息的概念、关系和属性；2)驱动代理行为和协作的模型和表示；3)集成规则、机器学习和信息检索的代理“智能”引擎；4)在原型ATIS试验台上对该技术进行评估。更广泛地说，该研究解决了智能多代理系统开发中的基本问题，对其他上下文感知的基于代理的应用和研究具有参考价值。结果将使许多面向用户的系统得到改进，包括航天、国防、医疗保健、制造、金融、电子商务和交通运输，减少用户错误，提高用户满意度。研究生和本科生将参与这项研究，结果将被纳入不同学生群体的教育中，关键软件组件将作为开源分发。

This project addresses the use of software agents in a highly personalized intelligent Advanced Travel Information System (ATIS), providing integrated real-time traffic advice to travelers. Travelers will use increasingly powerful personal and in-vehicle wireless appliances, which monitor the user's context, location and motion (via GPS and other sensors), preferences, priorities, schedule and situation. USDOT studies suggest that information customized to a traveler's specific needs and situation will enable better transit decisions, adjusting schedule, routing, and travel modes (car, bus, train, etc.), and coordinating with other travelers, colleagues and family. This should improve the overall transport system convenience, utilization and safety. Software agents are loosely coupled software elements, ideal for highly flexible, dynamic, and complex systems. Agents are autonomous, pursuing a task agenda and collaborating with other agents via messages. ATIS agents represent traveler goals, preferences and plans, monitor trips, adapt plans to changing circumstances, and prioritize guidance and delegated actions. Agents combine information from multiple sources on routes, congestion, incidents, weather, transit modes, and schedules to make real-time, traveler-specific recommendations.The research goal is to determine how to effectively create, evolve and use models of traveler context, plans, routes and services which enable multiple agents to autonomously maintain their individual information, and to negotiate and combine information with other agents for useful and timely travel recommendations. Research outcomes include: 1) concepts, relationships and attributes for user and travel information; 2) models and representations to drive agent behavior and collaboration; 3) an agent "intelligence" engine integrating rules, machine learning, and information retrieval; and, 4) an evaluation of the technology in a prototype ATIS test bed.More broadly, the research addresses fundamental issues in intelligent multi-agent system development, valuable to other context-aware, agent-based applications and research. The results will enable improvements in numerous user-oriented systems, including space, defense, health care, manufacturing, finance, e-commerce, and transportation, reducing user errors and increasing user satisfaction. Graduate and undergraduate students will participate in the research, results will be incorporated into the education of a diverse student body, and key software components will be distributed as open source.