Both projects are associated with a formal statement of work that I've committed to doing.

SOW: Office of Naval Research

• The formal Statement of Work (SOW) from the Proposal is available. BYU's responsibilities are:

1. Develop Models of Interaction Time (IT) and Neglect Time (IT). Since the operator has limited attention to devote to UV control, she must neglect a UV for some time (Neglect Time) to devote interaction time (IT) to another UV. We will gather data on (NT,IT) characteristics of conventional interaction, and develop models that predict (NT,IT) for various control modes given environmental state.
2. Add Scheduling to Fan-Out Models and Use These Models to Improve Adjustable Autonomy (AA). We will model how AA affects fan-out when a human is managing multiple independent UVs using conventional control (waypoint, shared, and planning-based autonomy). We will identify normative scheduling models that improve AA given (NT,IT) characteristics for conventional UV control.
3. Biologically-Inspired Coordination and Team-Oriented Plans. We will investigate what makes optimization and biologically-inspired coordination algorithms difficult to command, and develop more natural mixed-initiative methods for managing UV teams. We will explore how operators can manage a UV team by explicitly communicating revisions to team goals, role assignments, etc.
4. Robust Communication and Tasking for Large UV Teams. We will investigate techniques for demarking subteams, identifying and addressing potential conflicts, and conveying subteam goals and progress without cluttering the user interface. We will also evaluate the types and frequencies of information sharing between multiple people and UVs, as well as the content of the desired interactions
5. Supporting Multiple Coordinating Humans and Multiple Coordinating UVs. We will compare different coordination parameters and complexity metrics for managing multiple interdependent operators managing multiple teams of coordinating UVs.
6. Research Evaluation. We will evaluate the research through human experimentation in high fidelity simulation environments and with physical UVs. We will develop and evaluate different C2 organizational structures for the M human, N UV system, and identify different performance.

Although these are tasks that are to be accomplished over a 3-year horizon, we are already in the second year and we are far behind on delivering on these objectives. To focus on this year's work, please read the following revised plan sent to the project PI.

• 2010-2011 Plans from report to Katia Sycara.

We are now improving and extending our agent-based simulation environment to include bio-inspired agent teams, and will soon look at ways for humans to manage them. This will be main focus for 2011, meaning that we will emphasize Coordination and Information, Robust Communication and Tasking, and Supporting Multiple Humans from my SOW.

We will focus on what we do best for this work: designing HRI to manage bio-inspired agent teams. Note that I'm not saying “control” these teams since the teams must have enough autonomy to benefit from being part of a bio-inspired teams. Instead, the goal is to manage them, meaning influence the way that the collective behavior emerges. This will require some modeling and metric work: modeling different organizations or humans and different communication models among agents, and creating metrics of effectiveness of managed behavior of bio-inspired teams.

SOW: Robotics CTA Here's a link to the Robotics CTA FY 2010 Task Proposal report. Two tasks are to be accomplished between now and the end of the year. UCF is primarily responsible for accomplishing the first task, BYU is responsible for the second. Philip's and Yisong's research are intended to accomplish BYU's task.

1. analyze the effects of variation on the ability of a multi-agent system to self organize and on the robustness and stability of a multi-agent system
2. extend fan-out-based organizations to allow robust performance in the presence of variability and multiple human operators

• based in reality, but abstracting out the key problem elements
• allow variation in agent behaviors
• blend bio-inspired agent behaviors and human supervisory control
• allow variation in agent roles
• explore various organizations, including designer-specified organizations and organizations that evolve through agent specialization

• urban combat
• Normandy
• small unit action, with blending between members of unit
• spatially distributed agents
• some tasks require multiple generalist agents, some require collaborative specialist agents
• limited observation and communication radius for each agent
• authority relationships

• Information Foraging.
• actions
• observations
• percept history
• influence
• roles
• action selection