The Laboratory for Intelligent Processes and Systems (UT:LIPS) conducts research to advance the fundamental science and application of distributed agent-based systems.  Such systems, often categorized as Distributed Artificial Intelligence (DAI) or Multi-Agent Systems (MAS), provide decision support and operation control in complex domains where decision makers must accomplish goals despite environments characterized by high volatility, information overload, and inherent uncertainty.  UT:LIPS addresses important MAS research challenges in two fundamental areas: (i) Multi-Agent System Design and Runtime Analysis – methods and tools to enable disciplined design and architecting of Multi-Agent Systems as well as techniques for verifying and understanding agent-based implementation behavior and (ii) Advanced Agent Technologies supporting decision making in complex domains – specific capabilities that can be designed into an agent to enhance fundamental agent operations (e.g., belief maintenance, planning, coordination), thereby improving decision-making performance and ensuring resilience in volatile, uncertain, and information-dependent environments.  Specific research efforts under each area are highlighted below.

Multi-Agent System Design and Runtime Analysis

Disciplined Design and Engineering of Multi-Agent Systems

As a basis for the definition of disciplined software engineering processes for MAS development, a fundamental goal of this research is to formalize the capabilities that should be expected of agents as well as the inter-dependencies between these capabilities.  To encourage the rapid development of interoperable, compatible, and flexible agents, tools based on disciplined design and evaluation methodologies are being developed to assist designers in constructing agent architectures and evaluating design decisions.

Verification and Comprehension of Multi-Agent System Behavior through Explanation

A strong motivation for the application of multi-agent systems for decision support in complex domains is the capability enabled by properties such as emergent system behavior, proactive reasoning to achieve goals, autonomy to take action without a human when necessary, etc.  Nonetheless, these properties complicate efforts to understand exactly what agents are doing in an executing MAS and verify that the system and agent behavior observed is as anticipated.  This research contributes a method and accompanying tool to explain agent behavior in an implemented system in terms of general agent concepts and their relationships commonly used in agent-oriented design (e.g., goal, belief, intention, action, event, and message).  These advances facilitate the comprehension, debugging, and testing of agent-based software for designers, developers, and end-users.

Advanced Agent Technologies Supporting Decision Making

Improving Robustness and Fault Tolerance by Assessing the Trustworthiness of Information and Sources

As decision-makers, software agents construct beliefs about themselves, others, and the environment.  These beliefs form the basis for decisions and actions; therefore, a decision maker requires accurate beliefs.  This research equips agents with effective, efficient trustworthiness assessment mechanisms to determine what information to trust, who to trust, and when the information is most trustworthy.  Trustworthiness assessments are justified based on a set of information valuation policies: 1) priority of using maximum information, 2) priority to corroborated information, 3) priority for source certainty, 4) priority to reliable sources, and 5) priority to recent information.

Building Quality, Reliable Information Sharing Networks

Information sharing among agents influences each agent’s ability to achieve its goals.  Selecting information sources often occurs under the existence of uncertainty, motivating an agent to maintain trustworthiness evaluations of its information sources to assist in the evaluation of incoming information quality from those sources. In addition to the notion of trustworthiness, this research considers cost and coverage when narrowing among potential information sources.  Coverage is a measure of an information source’s contribution to an agent’s information needs, whereas the cost of getting information from a source is defined by the timeliness of information delivery from the source.

Facilitating the Coordination of Interacting Agents through Preference-based Action Selection

One strength of multi-agent systems is the ability of agents to aid each other in attaining their respective desires.  In deciding what actions to perform, an agent should consider both the capabilities and the desires of other agents in the system.  The decision model, constructed through both planning and desire analysis, provides a basis from which agents can negotiate to form agreements for coordination or collaboration with other agents.  This research provides the agent with the ability to incorporate the capabilities, intentions, and desires of other agents into its decision model for action selection.

The Laboratory addresses research problems in the context of active collaborations with industrial partners, industry consortia, federal agencies and state agencies. We believe that frequent and meaningful interactions with industry and government partners help to focus the Laboratory energies, hone skills and freshen perspectives as we pursue the Laboratory's research objectives.  As an engineering research element, the applicability, acceptability and extensibility of the research objectives form a cornerstone of the Laboratory's unified effort.

The Laboratory fosters a creative and productive engineering research environment for graduate and undergraduate students.  Organizationally, the Laboratory for Intelligent Processes and Systems resides within the Electrical and Computer Engineering department (ranked 7^th according to U.S. News and World Report).  The Laboratory is located in the Applied Computational Engineering and Sciences (ACES) building, designed for interdisciplinary research and graduate study.  This building brings together faculty, graduate students and research resources from three closely related programs at the University of Texas: Electrical and Computer Engineering, Computer Sciences, and the Institute for Computational Engineering and Sciences, focusing on the state of the art in software engineering, visualization and graphics, intelligent systems, and parallel and distributed systems.  The building infrastructure is specifically designed for bandwidth-intensive computational research and will support changing technology.

Within the ACES building (total 179,000 sq. ft, total 16,500 square feet of lab space), the Laboratory resides in a 2000 sq.ft. laboratory with 500 square feet of additional office space.  Equipment specific to the Laboratory currently includes numerous Intel-based workstations, laptops, and servers provided by grants from DARPA, NSF, IBM, and Allied Signal.   The Laboratory has considerable experience in C, C++, Java, CLIPS, and LISP programming, as well as experience in database technologies and XML, to support both research and potential deployment.