(Slide 1)
Good Afternoon.  I am Dr. Todd Carrico and I will be presenting some information on one 
of ISO’s emerging success stories, the Advanced Logistics Project, or ALP.  The Advanced 
Logistics Project is a joint DARPA and DLA initiative being developed in conjunction with 
US Transcom and the Joint Staff J4.  ALP has the objective of trying to develop and apply 
advanced information technology to get control of the Logistics Pipeline and ultimately the 
entire global logistics business process.

(Slide 2)
To achieve total control of the entire logistics pipeline through advanced technology means 
being able to plan, manage, and have visibility into everything going on in logistics at all 
echelons and through all phases of the operations.  We want to be able to do this 
continuously during both planning and execution.  To achieve this, the advanced logistics 
project is extending the state of the art in agent technology to create a radical new cognitive 
agent architecture for large scale distributed agent systems.  With this architecture as our 
basic building block, we can build communities and ultimately societies of agents that will 
automate and integrate everything going on in logistics to form a complex society of 
specialized agents capable of working together to plan and execute the entire logistics 
business process globally.

(Slide 3)
The Advanced Logistics Project is trying to develop an End-to End Logistics System.  Our 
vision is that at every level operations and logistics would work together during the course 
of action development and planning processes with all of the organizations involved in 
order to build an extremely detailed plan based on real world data.  With this extremely 
detailed plan, level 5 or execution level detail, we can then go into execution with the 
confidence that we can use real world data to monitor execution and understand how we're 
doing in achieving our operational objectives.  We can use technologies such as agent plan 
sentinels to watch critical components of the plan and when, based on real world data, we 
detect deviations in the plan we can immediately trigger replanning.  If you'll notice, our 
replanning is focused on just fixing those pieces of the plan that are affected by the 
deviations that we've detected.  This forms a continuous cycle between execution 
monitoring and continuous replanning, where the system is always in an execution state and 
fixing those components of the plan in which we have detected deviations.

Now the process of global logistics may appear somewhat simplistic, but I assure it is not.  
The reality of just how complex logistics really is keeps many CIOs and CEOs in the 
business community up very late at night.  In DoD it is especially complex for a number of 
reasons.  To begin with, logistics consumes $84 Billion dollars of the defense budget every 
year.  DoD can be called on to engage anywhere in the world at any time - and they expect 
the logistics to be there.  The organizations supporting logistics are globally distributed and 
cut across all the services and across all echelons.  Furthermore, logistics involves not only 
DoD, but also commercial industry.  Our operating environments are very diverse. We must 
be prepared to operate everywhere, from places where we have a very rich infrastructure to 
places where we can't even get clean water.  Furthermore, it involves over a thousand 
stovepipe logistic systems all working to do different pieces of the logistics plan but trying 
to come up with a complete and a cohesive solution.  There are tens of thousands of major 
end items that need to be incorporated into any one logistics plan, and furthermore, millions 
of critical spares or minor items that we need to be able to account for in order to ensure 
operational success.  And finally, all of this needs to be done in a continuous and dynamic 
environment where the real world is changing and all of the information that we are 
planning upon is constantly in flux.  Couple that with the need to accomplish the logistics 
mission with fewer people, less inventory, and less lift capability makes for a complex and 
challenging problem domain.

(Slide 4)
Now I’d like to take a minute to talk about our cognitive agent architecture.

Computers are a fairly new addition to our workplace desktop - they help us manage our 
data and to keep track of our messages and action items.  Before we had computers on our 
desktop we had to rely a lot more on our own internal ability to decompose large problems 
into smaller pieces so that we could solve them.  We have a tremendous amount of sensory 
data in the form of direct orders, e-mail, telephone conversations, briefing materials, 
newspapers, CNN, etc.  All of these items helping us understand the situation and make 
effective decisions.  It's up to each of us as individuals and managers to take all the tasks 
that we have to accomplish and decompose and organize them into smaller tasks which we 
could either accomplish internally or we can give to other people or organizations who are 
supporting us.  This is called the expander role.  The next thing that we have to do is take 
these individual smaller pieces of the problem and allocate our resources against those 
pieces so that we can accomplish the task at hand.  This is called allocation.  After 
allocation we have to assess how we're doing.  This is a continuous process of comparing 
our original objectives to how we're doing against achieving those objectives, and then if 
necessary, replanning in order to correct our course and get the objective achieved.  We've 
been doing this for a lot of years and it's a continuous process that we've become very 
familiar with doing.  The problem is we're a global world now.  We have to handle much 
larger problems with much longer duration but with less time to the planning.  It's becoming 
more and more difficult for the human alone to manage all of these details and 
requirements.  So, when we designed our system we took this into consideration, creating 
an overarching cognitive model which drove the design of our agent architecture.

This is the real innovative, revolutionary technology that is being developed under this 
project.  This architecture models the way humans do planning and execution. This 
architecture, something we call the cluster architecture, contains software instantiations of 
each of these roles that we play as managers, everyday.  It has an expander, an allocator, 
and an assessor.  In addition, it has a data management section which takes the full power 
of the computers ability to very rapidly and accurately manage very large conglomerations 
of data coming in from a large number of external sources.  What makes this cluster unique 
is the plug-ins for each one of these major components.  A plug-in is simply a software 
capability that provides domain expertise to the function of this particular cluster.  So each 
one of these plug-ins would be tailored to make the appropriate decisions in planning and 
scheduling activities that occur for a given organization.  If this cluster, in fact, was at an 
air-base, it might be engaged in the process of scheduling aircraft.  If it were at an inventory 
point for the services, DLA or the Army Materiel Command, it might have had the 
processes embodied in it to allow us to source materials that we need and manage 
inventories, to allow us to sustain our operation.

Now, these plug-ins do not have to be new software.  Some of the software would 
obviously be new, other software may be systems that exist today, only it can't talk to all of 
the other systems that it needs to feed data to.  This architecture would allow an existing 
system today to be wrapped and plugged in to the ALP architecture so that we can take 
100% advantage of the capability of the existing system and use it in the ALP architecture 
that we're building.

Now if you can imagine one of these agents doing a single problem, we can have a number 
of agents that do different but related problems and pull them together in a community.

(Slide 5)
Here we have a community of clusters solving some particular piece of the problem for the 
logistics domain.  In this slide we're trying to illustrate that the different pieces need to 
come together and exchange information in a timely and efficient manner so that together a 
collection of agents, each a specialist at a particular function, together - we can automate 
the business processes of an organization.

Now, if we take this community of clusters which is specialized in one particular aspect of 
the 
logistics and add it to other communities of clusters we have what we call the ALP Society.  

(Slide 6) 
The ALP Society takes communities of clusters working together in a very efficient and 
timely 
manner, to exchange the information and to have very up-to-date, accurate situations so that 
we can quickly respond to the needs and desires of the operational commander.  This 
society is now capable of doing the global business process of logistics in a fully 
automated, distributed, continuous fashion which reached across all of the DoD logistics 
assets, people, and organizations.

(Slide 7) 
This year, we demonstrated a distributed logistics planning and execution system which is 
built 
upon this agent based architecture.  We demonstrated in January how we can use that 
architecture to automatically generate a very detailed logistics plan using bottom up 
generation techniques against this live data and building detailed support and transportation 
schedules for over 10,000 personnel and about 35% of an armored heavy division.  We also 
demonstrated the planning components for deployment of an Air Expeditionary Force.  And 
all of this was accomplished in under an hour. 

(Slide 8)
In the next year we are going to build the larger society, with over 125 clusters over seven 
(7) 
geographic locations.  We're going to have the larger force deployment package.  We're 
going to be looking at bringing in the Marine Expeditionary Force processes and the 
numbered Navy processes in order to add diversity and complexity to our existing society, 
as well as, going deeper on some of the processes in the AEF and the Army division we 
already have.  In addition to the size and complexity of our forces, we're also going to be 
addressing a larger component of the logistics pipeline.  In January we demonstrated the 
detailed planning from the installation through to an administrative load of a ship at the 
port.  In next year's demonstration we're going to be extending that to go from the 
installation all the way through to the Tactical Assembly Area.  Most importantly, next 
year's demonstration will address execution, both the monitoring of that execution in real-
time against real data, and also the detection of deviations in the plans based on those real-
world data streams and correction of just those components of the logistics plan that have 
been affected by the deviation.  

(Slide 9)
Now we've talked several times about the logistics plan.  I want to spend a few minutes and 
talk about the global logistics plan as its key to our total system vision.  The logistics plan 
that we envision is a living entity.  It's a living logistics plan that contains all of the 
information for all of the organizations involved in logistics globally.   It's distributed 
geographically all over the world where pieces of the plan are located with the 
organizations that are doing those components of logistics.  It's tied to the operational plan 
so that we understand the drivers for all of the pieces of the logistics plan over time.  And 
it's also fed continuously by real world data, so that we're constantly updating and refining 
our representation of our plans and of the current world state based on these real world data 
feeds.  Furthermore, this information is all interrelated so we can understand the utilization 
of people and assets and organizations throughout the plan, as well as, the relationships 
between all of these components.

(Slide 10)
This system, though revolutionary, does not ignore the need to evolve our current 
capabilities.  
Instead, it has been designed to integrate all our existing information systems and 
databases, both within the DoD and outside of DoD (like commercial and coalition forces), 
effectively forming the glue for a revolutionary new logistics infrastructure.

The benefit of the cluster design is that we now have the framework for incrementally 
building large distributed heterogeneous systems that can work semi-autonomously, fully-
distributed, platform independent, and work together in a much more efficient manner than 
we're able to do today with our current relatively stovepipe systems.  In addition, we have 
the added benefit of improved cost effectiveness for future system development.  This is a 
very important issue because if we are able to build an architecture in which all we have to 
do in order to add a desired functional capability is to build a domain specific plug-in 
appropriate for that task, we can concentrate our resource dollars on developing new 
capabilities and new functionality and not spend large amounts of our limited resources on 
architecture and infrastructure to support those capabilities.  

So in conclusion, ALP is a proof of concept exploration of an end-to-end logistics system 
which will allow us for the first time to get control of the logistics pipeline across to all 
services, across to all phases of the operation, at all echelons.  Most importantly, it will 
provide us an infrastructure for execution and execution planning that will allow us to plan, 
manage, and have visibility into the entire logistics pipeline.  But in order to do that we've 
had to develop a revolutionary new architecture, the cluster architecture.  This basic 
building block, when specialized and given unique domain behaviors, can work together 
with other clusters to form an agent community that can tackle pieces of the logistics 
business process.  When several of these communities, which often reflect individual 
organizations, come together, we can create a complex society of these clusters that can 
work together to solve the entire logistics business process.  By the end of the Advanced 
Logistics Program, we hope to prove and demonstrate the first large scale distributed agent-
based system, and more importantly an architecture that provides the capabilities to deal 
with the information complexity and diversity of the logistics domain.

This revolutionary new architecture will enable a fundamental change in the way we do 
business in logistics.  This moves us from an environment where we're planning for 
execution, deliberate planning, to an environment where we can do execution, monitoring, 
and replanning in real time against real information.  

ALP plans to demonstrate the first critical steps towards gaining control of the logistics 
pipeline, and will develop and demonstrate a technology which at least holds the potential 
to completely revolutionize the way we do business, and more importantly, to realize the 
Joint Vision 2010 tenant of Focused Logistics.

Thank you very much.