Teamwork Coordination

In this paper we examine human tasks in controlling UAV teams by considering their computational complexity in the number of UAVs. This analysis suggests that controlling or following fine grained cooperation among many subteams of UAVs is the most difficult task and probably beyond human capabilities for most sizes of teams. We introduce Machinetta, a teamwork system that automates fine grained level of cooperation through proxies and uses a heuristic approach based on small world networks to allow coordination of large teams. We describe an interface based on the FalconView pfps system that has been successfully used to control small teams of UAVs. We describe likely obstacles to human control of larger teams and introduce a 3 tiered architecture in which a commander performs O(1) control tasks, a group of secondary operators perform O(n) control tasks on a call room basis, and a teamwork system such as Machinetta automates the O(mn) coordination tasks. 1.0 INTRODUCTION Effective military use of robotics will depend on coordination of large teams. Automating an aircraft squadron, for example, would not only require automating the piloting of planes but also leadership functions, support roles, and incidental forms of cooperation such as battle damage assessment or escort functions. To illustrate anticipated scales one scenario proposed by RAND [1] using LOCAAS (low cost autonomous attack system) calls for the use of up to 1,000 UAVs (uninhabited aerial vehicles) in a massed attack.. Because some functions such as selection of targets or authorization to attack may doctrinally require human input evaluating the operator’s span of control as the number of controlled entities scale is critical for designing feasible human-automation control systems. Current human span of control limitations are severe. Miller [2], for example, showed that under expected target densities, a controller who is required to authorize weapon release for a target identified by a UCAV (unmanned combat air vehicle), could control no more than 13 UAVs even in the absence of other tasks. A similar breakpoint of 12 was found by [3] for retargeting Tomahawk missiles. Recent AFRL studies [4] target an even more modest 4 UAVs/operator. Similar numbers (3-9) [5] have been found for UGVs (unmanned ground vehicles). To extend these numbers to large scale teams will require breaking new ground in redefining the role of the operator and taking advantage of new forms of automation. Roles traditionally reserved for a human commander such as verifying each UCAV-found target prior to weapon release [2,6] may become infeasible in rapidly evolving missions as the size of teams increases unless novel control architectures can be developed. To extend operator span of control to large teams we must consider how control difficulty for different control tasks grows with increases in team size. Authorization for weapon release after operator verification of each