DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
interactions as a dynamical system <strong>and</strong> the second is<br />
based on moving averages.<br />
The paper is organized as follows. In Section 2, we<br />
discuss the modeling approach adopted to build the<br />
predictors. In Section 3, software implementation of the<br />
predictors as plugins within a Cougaar agent is<br />
discussed. In Section 4, we discuss experimental results<br />
based on the implementation of the predictors on a<br />
logistics system built on Cougaar. Approaches adopted<br />
to tune the predictors based on historical data is also<br />
discussed. Section 5 discusses conclusions <strong>and</strong> possible<br />
future areas of research <strong>and</strong> development.<br />
2 Predictor Design <strong>and</strong> Algorithms<br />
In this section we discuss the predictor algorithms<br />
that were implemented in Cougaar. Before we discuss<br />
the technical details of the algorithms we present a brief<br />
description of the logistics application domain, as some<br />
aspects of the design <strong>and</strong> implementation are specific to<br />
the application.<br />
2.1 Logistics Scenario<br />
The Cougaar multi-agent society considered in this<br />
effort is the Full society that was developed as a part of<br />
<strong>DARPA</strong>’s Ultralog Program (See [1] for more details)<br />
The Full is a military supply chain logistics society that<br />
consists of many different supply classes. Each agent in<br />
the society represents a military unit performing a<br />
certain logistics operation in the supply chain. For e.g.<br />
the TRANSCOM agent represents the transportation<br />
comm<strong>and</strong> authority for the US military. It issues<br />
directives to its subordinate units regarding the<br />
transportation to be provided to a particular agent for a<br />
particular type of shipment. Figure 1 shows the<br />
organizational structure of the prototype Full society.<br />
Figure 1. Full society hierarchical structure<br />
There are five main supply chain threads in the<br />
prototype military logistics society. They are (i)<br />
Ammunition Supply Chain (ii) Petroleum, Oil <strong>and</strong><br />
Lubricants Supply Chain (BulkPOL <strong>and</strong> PackagedPOL)<br />
(iii) Subsistence Supply Chain (Food, Water) (iv) Repair<br />
Parts Supply Chain; <strong>and</strong> (v) Transportation Supply<br />
Chain<br />
Within each supply chain there exists a customersupplier<br />
relationship between various agents. A<br />
customer makes requests for various items (POL,<br />
ammunition etc) to its supplier <strong>and</strong> the supplier in turn<br />
attempts to meet these dem<strong>and</strong>s based on its current<br />
inventory, or forwards the requests up the supply chain<br />
hierarchy. Thus, depending on its positioning in the<br />
hierarchy, a supplier can also be the customer for<br />
another agent.<br />
Figure 2 shows a part of the supply chain. Here, FSB<br />
is a supplier. ARBN <strong>and</strong> INFBN are customers of FSB<br />
(Note: FSB is also a customer of MSB.) These agents<br />
send dem<strong>and</strong> requests to the FSB, which are managed<br />
by its Inventory Manager. Based on the operation plan,<br />
Optempo <strong>and</strong> current inventory each customer (agent)<br />
requests items from its supplier.<br />
Figure 2. Predictor Implementation in Agent<br />
Network