Intelligent Distributed Group and Team Training Systems RPD-Enabled Agents as Teammates and Decision Aids Team-oriented Agents for Enhancing Fusion Based knowledge for the Objective Force Team-based Fusion Agents for Warrior's Edge Agent-based Collaborative Planning ITR: An Agent-Based Negotiation Framework for the Robust Design of Active-Passive Hybrid Piezoelectric Vibration Control NetworksThe goal of this MURI project is to develop a foundation for efficacious training of complex performance. The two, high-level underlying themes are: 1) enhancing group and team training capabilities through the use of intelligent agents, and 2) building a long term synergistic collaboration for advanced learning technologies between cognitive psychologists and computer scientists.
The two types of intelligent agents being developed to assist team training are partner agents and coaching agents. Coaching agents provide coaching feedback to trainees and their team based on the performance and the process of the team. Partner agents assist individual trainees by taking over the execution of some of the task components, thus allowing the trainee to concentrate on learning specific components, and to assist team training by fulfilling the roles of some team members. Both types of agents require knowledge and intelligence about desired team interactions, which are enabled by CAST, a team-based agent architecture.
Two application domains are being investigated under this research: Space Fortress and a synthetic task related to the AWACS weapon director team.
The Penn State/IST team for this research project is led by Dr. John Yen. The entire MURI project is led by Dr. Richard A. Volz from Department of Computer Science at Texas A&M University (TAMU), and conducted in collaboration with Dr. Thomas Ioerger (TAMU), and Dr. Wayne L. Shebilske and Dr. Pamela Tsang, both from the Department of Psychology at Wright State University.
This research project is funded by Army Research Lab under Advanced Decision Architecture (ADA) of Collaborative Technology Alliance (CTA). The objective of this RT, RPD-Enabled Teaming Agent Research (formerly Expanding the Capabilities of Existing Simulations of Military Command and Control System Performance Using Team-based Agent Models) is to investigate using “smart” agents, agents that operate using the basic principles of recognition-primed decision making (RPD), as simulated teammates or as decision aids. The long-term goal is to assist human decision makers via the use of collaborative RPD-enabled agents. Motivated by research in team-based agents and naturalistic decision making, collaborative RPD-enabled teaming agents should enable improved decision support for distributed human teams.
The RPD-enabled software agents can serve as teammates or as direct decision aids that proactively reason, seek, and exchange relevant and required information to decision makers in dynamic environments. The “cognitively-aware” software teammates or aids, each assigned to a specific functional area, are intended to assist military staff in balancing information requirements against the dynamic and time sensitive decision making process: proactively reasoning across the decision space(s), seeking missing information queues from external information sources (collateral space), exchanging information between teammate, and monitoring/alerting on-going decision against expected outcomes.
The CAST-R work at Penn State has been demonstrated through the prototype Synergy displays developed under the ADA CTA. Thus, the basic capability for software integration and communication has already been established. This RT will build on that demonstration so that experiments can be conducted using CAST-R and Synergy to provide agents serving as teammates or as decision aids in the targeted context of the challenging information demands associated with the command and control of urban operations, including combat and sustain and support operations. The bounding parameters of this teaming and decision-support space for RPD-agent effectiveness will be explored and provided as guidance for development of decision architectures. The prototype environment developed will allow experimentation and demonstration of advanced tactical information exchange, reduced cognitive load, enhanced situation awareness, positive human-agent collaboration, and overall improved system effectiveness.
The basic agent architecture to be used is based on CAST, a teamwork model that enables agents, whether they are software agents or human agents, to anticipate potential information needs among teammates, and exchange information proactively. It especially incorporates a high-level language called Multi-Agent Logic-based Language for Encoding Teamwork (MALLET) for describing team structure, team process, as well as the criteria used by the team to dynamically assign tasks to members of the team. The agent kernel is able to dynamically infer information needs of teammates (based on a computational representation of individual and shared mental models) and proactively send them information relevant to their needs. In R-CAST, this basic capability has been further embellished with an RPD-based framework.
Principle investigators of this project are Dr. John Yen and Dr. Michael McNeese from School of Information Sciences and Technology at the Pennsylvania State University.
The goal of this project is to develop team-oriented agent technologies to support Fusion based Knowledge for the Objective Force Science and Technology Objective (STO). The vision of the Knowledge-based Fusion is that information and knowledge will be accessed, fused, and delivered more effectively. To fulfill this vision, data/information sources among and between the Unit of Action command integration cell, as well as other cells corresponding to various battle function areas (such as information superiority cell, maneuver cells) not only need to be integrated using an agent infrastructure that supports their communications and collaborations, but also need an agent teaming model that enables them to anticipate information needs of others and assist them proactively. Team performance researchers have long identified anticipatory behaviors and proactive assist behaviors as key elements of effective teamwork. They have pointed out that high performance team can accomplish this because they establishes overlapping shared mental models. Therefore, a major thrust of the proposed approach is to empower agents in a team with overlapping shared mental model so that they can conduct and support information/knowledge fusion in a proactive way.
The proposed project will be lead by Dr. John Yen at the Pennsylvania State University. The project will gain leverage from Dr. Yen’s previous and ongoing research in developing intelligent agent technologies. In particular, his current research supported by an AFOSR MURI grant has developed a multi-agent architecture called CAST (Collaborative Agents Simulating Teamwork), which will provide key foundations to the proposed team-oriented agents for knowledge fusion. The proposed project will also utilize expertise and relevant agent technologies developed under the “Agent-based Collaborative Planning” project, supported by Army Research Laboratory through Army Research Office in 2002.
The main deliverable of the project is an agent teaming architecture that supports fusion based knowledge for the objective force. The proposed agent teaming architecture will be integrated through the CoABS Grid concept and/or EMAA framework developed by Lockheed Martin. The participants of the proposed project will collaborate with relevant researchers from Army Research Laboratory in co-authoring one or more joint papers describing the results of the project.
For details, please look at a Project Summary Slide.
The goal of this research is to develop team-based agent technologies for enhancing the information fusion (especially level 2 fusion) of DCGS (Distributed Common Ground Station), which serves not only as the gateway between local units and the global “net” but also as a facilitator for information fusion occurring at different levels. There are at least two issues related to the effectiveness of such a vision. First, information fusion and exchanges between the global net, DCGS, and the local units need to adapt to the dynamic battle fields situations, which includes, but not limited to, new information from other sources, and decisions made by local units. Second, effective information fusion within DCGS needs to combine information collected locally with those obtained globally. Third, DCGS should guide, facilitate, and support information fusion of the local units.
As defined in JDL Data Fusion Process Model, level 2 fusion processing involves understanding the relationships among entities identified by level one processing, the relationship of these entities to the environment, and aggregation of entities for a higher level of understanding and assessment about the situation. More specifically, level 2 fusion includes object aggregation, event/activity aggregation, contextual interpretation, and multi-perspective assessment. Object aggregation analyzes objects that are in geographical proximity to determine if these objects are functionally related. Event/activity aggregation analyzes events and activities in time to identify their associations. Contextual interpretation analyzes the environmental, weather, socio-political context in which level 1 entities are being viewed. Multi-perspective assessment constructs the view of the enemy (i.e., the red view), the “neutral” view of the situation (i.e., the white view), and the view of the friendly forces (i.e., the blue view). These different components of level 2 fusion processing are inter-related. For example, the result of contextual interpretation should be used for object aggregation. Different perspectives of the battlefield are formed by other elements of level 2 processing.
This research addresses these technical challenges in three novel ways. First, we will augment DCGS with a team of level 2 information fusion agents with “shared mental models” between DCGS and the local units, and between DCGS and the global net. These shared mental models will enable the global net to adapt its information fusion based on dynamic needs of the local units. They will also enable and guide the entity/activity centric information fusion activities of the DCGS and local units. Second, based on the key functionalities of level-2 fusion processing, multiple agents will be developed with complementary capabilities. Together, they form a team that can coordinate and cooperate effectively to deal with the inter-relationships between elements of level-2 fusion. Third, to provide a general framework for agent-based level 2 information fusion, these agents will incorporate a novel situation awareness model based on naturalistic decision making.
The proposed research will be lead by Dr. John Yen, Dr. Dave Hall, Dr. Mike McNeese, and Dr. Sashi Phoha.
The goal of this project is to develop intelligent agent technology that helps a distributed mobile TOC (Tactical Operation Center) to collaborate for mission analysis by integrating a graphical display and a text-based collaborative planning tool in an intelligent way. At least two kinds of intelligent agents will be included in the proposed framework: (1) staff assistant agents and (2) team support agents. A staff assistant agent alerts a staff officer regarding events or messages they should pay attention to. Such alert will be generated based on a given context, which includes the roles of the officer as well as the situations of the battle space. A team support agent monitors the collaboration process to identify missing or incomplete steps and to suggest actions for the team to consider.
Both types of agents will utilize knowledge about the structure and the process of the mission analysis team, as well as impacts of a team member's actions on the function of other staff officers. The proposed project will be lead by Dr. John Yen. The project will gain leverage from Dr. Yen's previous and ongoing research in developing team-based agents. In particular, his current research supported by an AFOSR MURI grant has developed a multi-agent architecture called CAST (Collaborative Agents Simulating Teamwork), which will provide key foundations to the proposed agent-based collaborative planning research. The proposed project will also utilize expertise and relevant agent technologies developed under the "Tactical Intelligent Agent" project, supported by Army Research Laboratory from 2000 to 2001.
The objective of the proposed research is to develop an agent-based negotiation framework for the robust design for large-scale active-passive hybrid piezoelectric networks (APPN) for structural vibration control.
The underlying principle for APPN is to combine the active and passive control features of piezoelectric materials and circuits. While vibration suppression using APPN is indeed a very attractive concept, it was also recognized that several critical issues need to be addressed before such a system can be realistically be optimized in a complex smart structure environment for a large-scale and distributed vibration control. These issues, difficult to resolve using conventional methods, include system uncertainties, tradeoff scenarios, and topology variations in design. Through combining the expertise (IT and engineering) of the two investigators and leveraging upon their ongoing research, the proposed effort can address these critical issues and greatly advance the state of the art.
The goal of this research is to investigate an agent-based framework in making high-level design decisions (e.g., the selection of topology) and in analyzing the tradeoff among conflicting design objectives.
The principal investigators are Prof. John Yen from School of Information Sciences and Technology, and Dr. Kon-Well Wang from Department of Mechanical and Nuclear Engineering.