Modification History

• 09:24 - Thu Jun 26, 1996 -rmt- Created
• 16:22 - Tue Jul 9-16, 1996 -rmt- Added docs for time-related and message-related stuff (source files currently in directory Development/).
• 11:10 - Mon Nov 3, 1997 -rmt- Started new version of simulator documentation to add information about trace output and experiment control rules/functions.
• 11:58 - Mon May 11, 1998 -rmt- Converted to a chapter; changes to sync up with new version of simulator.

The MAUV (Multiple Autonomous Underwater Vehicle) projecthttp://cdps.umcs.maine.edu/MAUV is a collaborative project of the University of Maine's Cooperative Distributed Problem Solving (CDPS) Research Grouphttp://cdps.umcs.maine.edu and the Marine Systems Engineering Laboratory (MSEL)http://cdps.umcs.maine.edu/MSEL of The Autonomous Undersea Systems Insitute (AUSI)http://cdps.umcs.maine.edu/AUSI. The MAUV project focuses on ways to control an autonomous oceanographic sampling network (AOSN) [Curtin et al., 1993].

AOSNs are collections of sensor platforms, including autonomous underwater vehicles (AUVs), that have the task of monitoring an area of the ocean for an extended period of time, collecting data for scientists or other end users. See [Turner et al., 1996] for A <a href="/Papers/1996/AUV">paper</a> is also available that provides more details about AOSNs and our assumptions concerning controlling them.

The first two phases of the MAUV project involve building simulators with which to develop and test the protocols and mechanisms for AOSN control. We are currently in the first phase of the project. During this phase, a high-level simulator is being developed that models the aggregate properties of a group of agents operating according to the protocols without implementing individual agents.

Traditionally in AI, new ideas are directly implemented in what would now be called agents and tested out, either in simulation, in the lab, or in the real world. This process, though providing the most realistic test of the ideas, is difficult, time-consuming, and, worst of all, difficult to evaluate via empirical tests. In contrast, the first phase simulator models how agents would act without directly implementing the agents. Though this is a lower-fidelity simulator than one that directly implements agents, it has the advantages of being quick to build, supporting incremental changes leading toward more high-fidelity models, and allowing empirical, repeatable tests. We will use this simulation testbed to develop and test the protocols before devoting the effort needed to implement them directly.

Phase II of the project will see the implementation of the protocols by individual agents, probably including future versions of Orca [Turner, 1994]<a href="/Orca">Orca</a>. Phase III of the project will see the protocols and mechanisms ported to real AUVs for in-water tests prior to turning over the technology to the ocean science community.

This document describes the phase I simulator. It is intended both to give an overview of the simulator to those not involved in the project and to describe it in sufficient detail to orient newcomers to the project.