17 th ICCRTS “ Operationalizing C 2 Agility ” Command and Control of Teams of Autonomous Units Topics

Command and Control (C2) has always been the practice of directing teams of autonomous units. These units have included individual soldiers, aircraft directed by a pilot, and ships maneuvered and fought by the combined intelligence of an entire crew. It is into this already populated universe that we are now working to add autonomous unmanned systems (AUS). In this paper, we explore the C2 of teams of autonomous units that include human, human populated, and uninhabited systems. Beyond the simple onefor-one substitution of a manned vehicle by an unmanned vehicle, we consider the reality that AUS share information differently among themselves and their inhabited counterparts. Similarly, humans and human populated units will provide information to their commanders in a different fashion than machines will, because each are uniquely capable of different observations and understandings. This paper also describes the sparse supervisory control that must be exercised over highly autonomous units, and considers what it means for a commander to supervise AUS that employ machine learning and cooperative autonomy. The Boundary of Command and Control and Autonomy Autonomous unmanned systems (AUS) development is based on the idea of having a system operate without continuous human intervention. This autonomy is not the same as avoiding all human direction. Human commanders decide when to deploy systems, when to end their deployment, and for those systems capable of making changes to their plans while deployed, commanders may change their tasks in some way. Teams of heterogeneous autonomous units will require command and control, just as they always have. That the units are now unmanned does not necessarily change that. Controlling such a complex team requires several critical capabilities. First, the goals and constraints for the AUS team must be communicated to the various decision making nodes. These nodes may include all of the AUS, as they all may possess sufficient autonomous capability to decide how to act under many situations given the goals. A central controller, or more generally several distributed controllers, must have confidence (particularly if human operated) that the goals and constraints have been received and correctly interpreted by the autonomous units. Second, the control units must have sufficient situational awareness of the environment and the behaviors of the team members in order to decide if changes to orders are required. The control must have the ability to determine if any error conditions are

[1]  F. Hawkins.  Human factors in aviation. , 1979, Journal of psychosomatic research.

[2]  S. B. Filskov,et al.  Clinical detection of intellectual deterioration associated with brain damage. , 1984, Journal of clinical psychology.

[3]  E A Feigenbaum,et al.  Knowledge Engineering , 1984, Annals of the New York Academy of Sciences.

[4]  Sukhan Lee,et al.  Machine acquisition of skills by neural networks , 1991, IJCNN-91-Seattle International Joint Conference on Neural Networks.

[5]  Raja Parasuraman,et al.  Performance Consequences of Automation-Induced 'Complacency' , 1993 .

[6]  Philip J. Smith,et al.  Design of a Cooperative Problem-Solving System for En-Route Flight Planning: An Empirical Evaluation , 1994 .

[7]  S. King.  Learning to fly. , 1998, Nursing Times.

[8]  Amy E. Henninger,et al.  The Limitations of Static Performance Metrics for Dynamic Tasks Learned Through Observation , 2001 .

[9]  Amy R. Pritchett,et al.  Development and Evaluation of a Cockpit Decision-Aid for Emergency Trajectory Generation , 2001 .

[10]  Kathleen L. Mosier,et al.  Aircrews and Automation Bias: The Advantages of Teamwork? , 2001 .

[11]  Liling Ren,et al.  Interaction of Automation and Time Pressure in a Route Replanning Task , 2002 .

[12]  Heath A. Ruff,et al.  Human Interaction with Levels of Automation and Decision-Aid Fidelity in the Supervisory Control of Multiple Simulated Unmanned Air Vehicles , 2002, Presence: Teleoperators & Virtual Environments.

[13]  David Garlan,et al.  Rainbow: architecture-based self-adaptation with reusable infrastructure , 2004 .

[14]  R. Mooney,et al.  Explanation-Based Learning: An Alternative View , 1986, Machine Learning.

[15]  Avelino J. Gonzalez,et al.  Learning tactical human behavior through observation of human performance , 2006, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics).

[16]  Mary L. Cummings,et al.  Developing Operator Capacity Estimates for Supervisory Control of Autonomous Vehicles , 2007, Hum. Factors.

[17]  Mary L. Cummings,et al.  Predicting Controller Capacity in Supervisory Control of Multiple UAVs , 2008, IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans.

[18]  Christopher D. Wickens,et al.  Humans: Still Vital After All These Years of Automation , 2008, Hum. Factors.

[19]  Kenneth O. Stanley,et al.  Generative encoding for multiagent learning , 2008, GECCO '08.

[20]  A. Gonzalez,et al.  Falconet: force-feedback approach for learning from coaching and observation using natural and experiential training , 2009 .

[21]  Raja Parasuraman,et al.  Adaptive Automation for Human Supervision of Multiple Uninhabited Vehicles: Effects on Change Detection, Situation Awareness, and Mental Workload , 2009 .

[22]  Joel Lehman,et al.  Evolving policy geometry for scalable multiagent learning , 2010, AAMAS.

[23]  Raja Parasuraman,et al.  Complacency and Bias in Human Use of Automation: An Attentional Integration , 2010, Hum. Factors.

[24]  M. O. Rodas,et al.  16th ICCRTS “Collective C2 in Multinational Civil-Military Operations” “Modeling Operator Cognitive Capacity in Complex C2 Environments” Title: Evaluating Unmanned Systems’ Command and Control Technologies under Realistic Assumptions Topic(s): Modeling and Simulation, Collaboration, Shared Awareness , 2011 .