DAMN:
A Distributed Architecture 
for Mobile Navigation

Investigator: Julio Rosenblatt

Abstract

In unstructured, unknown, and dynamic environments, planning systems cannot generate a plan a priori that can be expected to perform reasonably in the face of such uncertainty, nor can they anticipate all contingencies that may arise; instead, decision-making must be based on current information and state at all times, proceeding in a data-driven manner, rather than attempting to impose unrealisable plans in a top-down fashion. In addition, the dynamics of the vehicle itself often play an important role in determining which actions may be achieved and which actions are to be avoided. For complex domains such as mobile robot navigation, there also exists the need to combine information from several different sources to be used to perform diverse tasks. A system architecture should be able to accommodate and integrate subsystems that have been developed independently, therefore it must impose minimal restrictions on the nature of the data, representations, and algorithms used by these subsystems designed to achieve their respective tasks. Also, sensors and the functions that process their data each operate at different rates, so they must be allowed to operate asynchronously to maximize the throughput and thus the responsiveness of the system.

The Distributed Architecture for Mobile Navigation (DAMN) is a planning and control architecture in which a collection of independently operating modules collectively determine a robot's actions. DAMN consists of a group of distributed behaviours communicating with a centralized arbiter, either by sending votes in favour of actions that satisfy its objectives, or by indicating the utility of various possible world states. The arbiter is then responsible for combining the behaviours' votes and generating actions which reflects their objectives and priorities. The use of distributed shared control allows multiple levels of planning to be used in decision-making without the need for an hierarchical structure, and the distributed, asynchronous nature of the architecture allows multiple goals and constraints to be fulfilled simultaneously. Thus, DAMN provides coherent, rational, goal-directed behaviour while preserving real-time responsiveness to its immediate physical environment. DAMN also provides a framework for developing and integrating independent decision-making modules communicating with such arbiters, thus facilitating their development and leading to evolutionary creation of robust systems of incrementally greater capabilities.

DAMN has been used to create various systems for mobile robot navigation and active sensor control. Diverse subsystems have been integrated within this architecture to create systems that, for example, perform road following, cross-country navigation, map-based route following, and tele-operation while avoiding obstacles and meeting mission objectives. A Reactive System for Off-Road Autonomous Driving provides a specific example of a system built with DAMN at Carnegie Mellon University. Work is underway in developing DAMN for use in the Autonomous Underwater Vehicles and Cargo Handling projects here at the University of Sydney.

DAMN Papers

Rosenblatt, J. Maximizing Expected Utility for Optimal Action Selection under Uncertainty. Autonomous Robots, vol. 9, no. 1, pp. 17-25, 2000.

Rosenblatt, J. The Distributed Architecture for Mobile Navigation. Journal of Experimental and Theoretical Artificial Intelligence, vol. 9, no. 2/3, pp.339-360, April-September, 1997.

Rosenblatt, J. DAMN: A Distributed Architecture for Mobile Navigation (5.8Mb) Ph.D. Thesis, Carnegie Mellon University Robotics Institute Technical Report CMU-RI-TR-97-01, Pittsburgh, PA, 1997. (postscript 10.1Mb)

You can also find various other journal and conference papers, book chapters, etc. on my Publications Page

Comments and suggestions are always welcome! E-Mail me at julio@acfr.usyd.edu.au