• Home
• Bio
• Experience
• Publications
• Links
• Schedule
• SWAT
• Drakontas
• Chat

Kyle Usbeck

---

Masters Student, Research Assistant
Department of Computer Science
Drexel University
3141 Chestnut Street
Philadelphia, Pennsylvania, USA, 19104
Email:

---

Software Engineer
Drakontas LLC.
Suite #1
115 East Glenside Avenue
Glenside, Pennsylvania, USA, 19038
Email:

---

Phone:

Masters Thesis

Network-Centric Automated Planning and Execution

Abstract

Web services provide interoperability to network hosts with different capabilities. Complex tasks can be performed by composing services, assuming sufficient service descriptions are provided. Researchers are just beginning to realize the importance of accounting for network properties during automated service composition. The work presented in this thesis considers dynamic, heterogeneous networks --- one type of network-centric environment.

The purpose of this research is to improve network-centric service composition. This is accomplished by converting the service composition problem to an automated planning under uncertainty problem and by reasoning about network properties at various stages of the planning process. This thesis presents a method of improving the agents' ability to construct, execute, and monitor plans in network-centric environments.

There are two main contributions of this thesis: 1) generating qualitatively-different plans and 2) creating network-aware agents. As part of the former contribution, this thesis presents a comparison of methods used to create classical planning domains for distributed service composition problems. The other part of this contribution is an algorithm for guiding a plan-space planner to create qualitatively-different plans based on domain-dependent and network-centric plan evaluations. The second contribution pertains to network-awareness, which agents exhibit by reacting to changes in network conditions. This thesis describes methods of incorporating network-awareness into agents that 1) create plans, 2) execute plans, and 3) monitor plan execution.

Experiments to validate the aforementioned contributions are presented in the context of an Improvised Explosive Device (IED) detection scenario. Several locations are monitored for IEDs using a variety of techniques including manual searching and visual change detection, as well as a variety of resources including humans, robots, and unmanned aerial vehicles (UAVs). Empirical results indicate that incorporating network-awareness into agents in dynamic, heterogeneous networks improves the overall service composition performance and effectiveness.

Presentation