Developing Maneuvering Behaviors for a Glider UAV Using a Genetic Programming Model

Abstract

[his report describes the methodology for using a genetic programming model to develop maneuvering behaviors for an unmanned aerial vehicle (UAV) that is also a glider. The use of glider UAVs for surveillance and information-gathering operations has become increasingly important in defense and national security applications. Through an evolutionary process similar to that found in nature, the genetic programming model generates a computer program that when downloaded onto a glider UAV's on-board computer will guide the vehicle to find regions of lift for staying aloft while accomplishing an information-gathering task. This report discusses various approaches to developing behaviors using genetic programming and presents the results achieved.

Graduated embodiment for sophisticated agent evolution and optimization.

“…Using a tree-based GP (genetic programming) representation for behavior (Pryor 1998;Barnette et al 2000;Pryor and Barton 2002) allows a great deal of flexibility and can be adapted to many kinds of problems. The behavior programs execute by traversing a tree that is made up of elemental building-block units called nodes, which can either be a function or a terminal.…”

“…It was developed for this project as a generalized and object-oriented code that captures and extends the capabilities of an earlier Sandia GP code (Pryor and Barton 2002) with the added goals of modularity and integration with modeling and simulation tools. One of the DGGP engine's most important attributes is that it has been integrated with Umbra, Sandia's modeling and simulation framework (described in Section 6).…”

Developing close combat behaviors for simulated soldiers using genetic programming techniques.

“…We set out to build a system that develops sophisticated behaviors for soldier agents automatically. Because genetic programming has been successfully used in the past to create intelligent agents in similar domains, such as the RoboCup softbot competition (Luke, 1998), and has proven successful in previous experiments involving agents that must prioritize and balance several goals (Pryor, 2002), we decided to use genetic programming techniques in our system. In this paper, we will document the processes used to create and refine that system, and our accomplishments; it will also describe the directions we would like to take the research, and what we believe should be possible for such a system to do.…”

“…Welldeveloped descriptions of genetic programming can be found in previous papers in this sequence (see Pryor, 1998 and2002), or in any of several books. The seminal work in genetic programming, Genetic Programming: On the Programming of Computers by Means of Natural Selection, was written by John Koza, the methodology's inventor, and is extremely accessible and informative.…”

“…The second phase determined that genetic programs were capable of developing agent behaviors that pursued several goals simultaneously. For more details on that work, involving the creation of behaviors that control UAVs, see Pryor, 2002. The third phase would lead to the actual development of soldier behaviors through genetic programming.…”