This project will integrate flight scheduling, execution management,
and distributed coordination capabilities to provide an integrated
basis for generating and updating flight schedules in response to new
requirements, negotiating adjustments to resource assignments,
immediately detecting schedule deviations during execution and
alerting users about them, and dynamically revising and reoptimizing
flight schedules when circumstances warrant. Current operations at
the Air Mobility Command (AMC) are hampered by the gap in information
flow and problem solving between planning and execution offices.
We will use machine-understandable schedules and policies to close
this gap. This project is joint work with Carnegie Mellon University.
We are able to close the loop by using machine-understandable
schedules to derive the systems expectations of what should happen,
and using machine-understandable representations of policies and
procedures to react to situations that do not meet these expectations.
These policies and procedures, which encode AMCs desired responses to
execution events, will in turn modify the schedule, thus closing the
While there may be automated responses for small, common deviations,
responses to significant deviations will often be to alert the human
user to the situation and let the user take control of revising the
schedule, as directed by encoded AMC policies.
Our approach builds from a unique combination of three enabling
technologies. The AMC Barrel Allocator (AMC-BA), an incremental
constraint-based scheduler developed by CMU and now actively transitioning
into operations as the day-to-day airlift and tanker scheduling component of
AMCs Consolidated Air Mobility Planning System (CAMPS), will be the core
component for generating and revising flight schedules. AMC-BA is capable of
incrementally revising schedules in response to new or changed input
missions or resource availability conditions. Although currently
incorporated in CAMPS as a planning tool, it is also seen to provide a
direct basis for managing flight schedules in response to execution events.
We will develop an execution assistant, AMC-EA, which will actively monitor
AMC data information sources for expectations it derives from the current
schedule, recognize deviations immediately, and apply policies for
responding to deviations. As described above, responses to significant
deviations will alert the user to take control. Thus, we can view the
AMC-EA as filtering the noise out of data streams, handling small deviations
when so specified by a policy, and recognizing (and alerting on) those
significant deviations that require human-controlled rescheduling. The
AMC-EA will continuously react to new information while interspersing its
proactive pursuit of response procedures. The AMC-EA will build on
execution monitoring technology developed by SRI for the
Defense Advanced Research Projects Agency (DARPA) Small Unit
Operations (SUO) program.
We will develop techniques for collaborative negotiation of resource
allocation decisions between AMC planning cells by building on
technology SRI is developing for incremental negotiation and coalition
formation, in the DARPA Autonomous Negotiating Teams program.