Publications

Robustness Infrastructure for Multi-Agent Systems

Ronald D. Snyder, Douglas C. MacKenzie, and Raymond S. Tomlinson
Proceedings Open Cougaar 2004
New York, July 2004.

When used for mission critical applications multi-agent systems must be capable of sustaining uninterrupted operations in the face of both hardware and software failures. Applications of this type are often distributed over a large number of geographically dispersed hosts, which make them susceptible to a wide range of failures. A fault tolerant infrastructure must be able to detect and adapt to these failures to provide continuity of processing. This paper discusses the approach used to provide such an infrastructure in the Cougaar multi-agent system.

Cougaar Agent Communities

Ronald D. Snyder, and Douglas C. MacKenzie
Proceedings Open Cougaar 2004
New York, July 2004.

The ability to organize agents into abstract groups provides a powerful tool for agent organization and communication in large multi-agent systems. In this paper we outline the fundamental characteristics required of a robust and scalable agent grouping mechanism. These characteristics are then discussed in the context of the Cougaar community infrastructure. This implementation of agent communities illustrates the use of distributed state management, distributed event propagation and abstract messaging in a high-performance agent architecture designed for robustness and scalability.

Usability Evaluation of High-Level User Assistance for Robot Mission Specification

Yoichiro Endo, Douglas C. MacKenzie, and Ronald C. Arkin
IEEE Transactions on Systems, Man, and Cybernetics,
Part C: Appications and Reviews Special Issue on Human-Robot Interactions
Vol 34, No. 2, Pg 168-180. May 2004.

MissionLab is a mission specification system that implements a hybrid deliberative and reactive control architecture for autonomous mobile robots. The user creates and executes the robot mission plans through its graphical user interface. As robot deployments become more common in highly stressful situations, such as in dealing with explosives or biohazards, the usability of their mission specification system becomes critical. To address this need, a mission-planning "wizard" has been recently integrated into MissionLab. By retrieving and adapting past successful mission plans stored in its database, this new feature is designed to simplify the user's planning process. The latest formal usability experiments, reported in this paper, testing for usability improvements in terms of speed of the mission planning process, accuracy of the produced mission plans, and ease of use is conducted. This paper introduces the mission-planning wizard, describes the usability experiments (including design), and discusses the results in detail.

Collaborative Tasking of Tightly Constrained Multi-Robot Missions

Douglas C. MacKenzie
In Multi-Robot Systems: From Swarms to Intelligent Automata (Proceedings Second International Workshop on Multi-Robot Systems),
Washington D.C., A.C. Schultz et al. Editors, Kluwer Academic Publishers. vol. 2, pp. 39-50, 2003.

Collaborative Tasking provides the ability to automatically task and re-task a group of heterogeneous unmanned vehicles, producing a stable, robust, tightly integrated unit while reducing manpower requirements. A variant of a market economy-based approach is used, where each vehicle computes constraint-based estimates of its cost to perform a particular task. These cost projections are sent back to the entity offering the task, which uses a cost minimization algorithm to select the appropriate vehicle(s) and constraint values in a unified manner. Each vehicle's computation of its cost to perform a task takes into account remaining consumables, required effort, its other pending tasks, and user specified preferences. Costs are reported as functions of constraints, such as location and time, to provide a unified approach to assigning time-sensitive, tightly coordinated tasks.

Multistrategy Learning Methods for Multirobot Systems

Ronald C. Arkin, Yoichiro Endo, Brian Lee, Douglas C. MacKenzie, and Eric Martinson.
In Multi-Robot Systems: From Swarms to Intelligent Automata (Proceedings Second International Workshop on Multi-Robot Systems),
Washington D.C., A.C. Schultz et al. Editors, Kluwer Academic Publishers. vol. 2, pp. 137-150, 2003.

This article describes three different methods for introducing machine learning into a hybrid deliberative/reactive architecture for multirobot systems: learning momentum, Q-learning, and CBR wizards. A range of simulation experiments and results are reported using the Georgia Tech MissionLab mission specification system.

Evaluating the Usability of Robot Programming Toolsets

D.C. MacKenzie and R.C. Arkin,
The International Journal of Robotics Research,
Vol. 17, No. 4, pp 381-401, 1998.

This article explores the issues surrounding just how one begins to evaluate the usability of robot programming toolsets used to create and maintain robot mission plans. It is necessary to consider usability early in the develop cycle, but when the application is available, it must be evaluated as to its usability by the target audience. There are four popular procedures to evaluate the usability of software packages in the Human-Computer Interfaces literature. Heuristic evaluation asks interface specialists to study the package and look for aspects that, based on their experience, will be confusing for users. A process called Guidelines has developers rate their system based on a list of good interface design principles. In Cognitive walkthroughs, developers perform software walkthroughs to evaluate the actions required by the toolset based on a cognitive model of how users will expect the interface to work. Usability testing attempts to study and measure how representative users interact with the system while performing realistic tasks. The peculiarities of applying Usability testing to a robot programming toolset are the focus of this article. The desired characteristics of a Robot Programming Toolset are presented along with the specifics of an exemplar system, the MissionLab toolset developed at Georgia Tech, which is used to ground the discussions. Specific techniques which can be used to establish usability criteria for toolsets are discussed and the specific usability criteria established for MissionLab are presented. Designing experiments to generate values for usability criteria is discussed and two specific experiments created to evaluate MissionLab are presented. The evaluation of experimental data is discussed along with the results for the MissionLab experiments.

MultiagentMission Specification and Execution

D.C. MacKenzie, R.C. Arkin, and J. Cameron,
Autonomous Robots, Kluwer Academic Publishers,
Vol. 4, No. 1, pp 29-52, 1997.

Specifying a reactive behavioral configuration for use by a multiagent team requires both a careful choice of the behavior set and the creation of a temporal chain of behaviors which executes the mission. This difficult task is simplified by applying an object-oriented approach to the design of the mission using a construction called an assemblage and a methodology called temporal sequencing. The assemblage construct allows building high level primitives which provide abstractions for the designer. Assemblages consist of groups of basic behaviors and coordination mechanisms that allow the group to be treated as a new coherent behavior. Upon instantiation, the assemblage is parameterized based on the specific mission requirements. Assemblages can be re-parameterized and used in other states within a mission or archived as high level primitives for use in subsequent projects. Temporal sequencing partitions the mission into discrete operating states with perceptual triggers causing transitions between those states. Several smaller independent configurations (assemblages) can then be created which each implement one state. The Societal Agent theory is presented as a basis for constructions of this form. The Configuration Description Language (CDL) is developed to capture the recursive composition of configurations in an architecture- and robot-independent fashion. The MissionLab system, an implementation based on CDL, supports the graphical construction of configurations using a visual editor. Various multiagent missions are demonstrated in simulation and on our Denning robots using these tools.

A Design Methodology for the Configuration of Behavior-Based Robots

D.C. MacKenzie
Ph.D. Dissertation (also Georgia Institute of Technology Tech Report "GIT-CS-97/01").

Behavior-based robotic systems are becoming both more prevalent and more competent. However, operators lacking programming skills are still forced to use canned configurations hand-crafted by experienced roboticists. This inability of ordinary people to specify tasks for robots is inhibiting the spread of robots into everyday life. Even expert roboticists are unable to share solutions in executable forms since there is no commonality of configuration descriptions. Further, a configuration commonly requires significant rework before it can be deployed on a different robot, even one with similar capabilities. The research documented in this dissertation attacks this problem from three fronts.

First, the foundational Societal Agent theory is developed to describe how agents form abstract structures at all levels in a recursive fashion. It provides a uniform view of agents, no matter what their physical embodiment. Agents are treated consistently across the spectrum, from a primitive motor behavior to a configuration coordinating large groups of robots. The recursive nature of the agent construction facilitates information hiding and the creation of high-level primitives.

Secondly, the MissionLab toolset is developed which supports the graphical construction of architecture- and robot-independent configurations. This independence allows users to directly transfer designs to be bound to the specific robots at the recipient's site. The assemblage construction supports the recursive construction of new coherent behaviors from coordinated groups of other behaviors. This allows users to build libraries of increasingly high-level primitives which are directly tailored to their needs. MissionLab support for the graphical construction of state-transition diagrams allows use of temporal sequencing to partition a mission into discrete operating states, with assemblages implementing each state. Support for multiple code generators (currently existing for AuRA and SAUSAGES) ensures that a wide variety of robots can be supported.

Finally, specific usability criteria for toolsets such as MissionLab are established. Three usability studies are defined to allow experimental establishment of values for these criteria. The results of carrying out these studies using the MissionLab toolset are presented, confirming its benefits over conventional techniques.

Behavior-Based Mobile Manipulation for Drum Sampling

D.C. MacKenzie and R.C. Arkin,
Proc. 1996 International Conference on Robotics and Automation,
Minneapolis, MN, April 1996, Vol. 3, pp. 2389-2395.

This paper describes an implementation of a behavior-based mobile manipulator capable of autonomously transferring a sample from one drum to a second in unstructured environments. A major contribution of the project was the coherent integration of the arm and base as a cohesive unit, and not just a mobile base with an arm attached. The support for smooth simultaneous operation of all joints on the vehicle facilitated biologically plausible motions, such as arm preshaping. The behavior-based controller used a pseudo-force model, where behaviors add forces and torques to joints and limbs resulting in coordinated motion. The vehicle Jacobian is used to convert the pseudo-forces into joint torques and a pseudo-damping model converts the joint torques into joint velocities. This process allows rapid control of the manipulator without the use of inverse kinematics. A drum sampling task is presented where the vehicle demonstrates how a sample of material could be moved from one drum to another, illustrating the efficacy of the solution.

Specification and Execution of Multiagent Missions

D.C. MacKenzie, J.M. Cameron, and R.C. Arkin,
Proc. 1995 International Conference on Intelligent Robots and Systems,
Pittsburgh, PA, August 1995, Vol. 3, pp. 51-58.

Specifying a multiagent behavioral configuration requires both a careful choice of the behavior set and creation of a temporal chain executing the mission using those behaviors. This difficult task is simplified by applying an object-oriented approach to the design using a methodology called temporal sequencing to partition the mission into discrete operating states and enumerate the perceptual triggers causing transitions between those states. Several smaller independent configurations (assemblages) can then be created, each implementing one distinct operating state. Each assemblage consists of a collection of behaviors and a suitable coordination mechanism which causes the group to act as a single, coherent behavior. The missions are specified in a structured user-friendly language targeted for military-style scout missions. Various multiagent missions have been demonstrated in simulation and results are shown using our Denning mobile robots.

Io, Ganymede and Callisto - A Multiagent Robot Trash-collecting Team

T. Balch, G. Boone, T. Collins, H. Forbes, D. MacKenzie, J. Santamaria,
AI Magazine, Vol. 16, No. 2, Summer 1995, pp. 39-51.

Georgia Tech won the Office Cleanup Event at the 1994 AAAI Mobile Robot Competition with a multi-robot cooperating team. This paper describes the design and implementation of these reactive trash-collecting robots, including details of multiagent cooperation, color vision for the detection of perceptual object classes, temporal sequencing of behaviors for task completion, and a language for specifying motor schema-based robot behaviors.

Planning to Behave: A Hybrid Deliberative/Reactive Control Architecture for Mobile Manipulation

R.C. Arkin, and D.C. MacKenzie,
Proc. 1994 International Symposium on Robotics and Manufacturing,
Maui, HI, August 1994, pp. 5-12.

Hybrid architectures provide an effective means for integrating world knowledge with reactive control. This paper describes the motivation behind the architectural decision to hybridize, and presents a case study in mobile manipulation in the context of the Autonomous Robot Architecture (AuRA).

Temporal Coordination of Perceptual Algorithms for Mobile Robot Navigation

R.C. Arkin, and D.C. MacKenzie,
IEEE Transactions on Robotics and Automation,
Vol 10, No. 3, June 1994, Pg 276-286.

A methodology for integrating multiple perceptual algorithms within a reactive robotic control system is presented. A model using finite state acceptors is developed as a means for expressing perceptual processing over space and time in the context of a particular motor behavior. This model can be utilized for a wide range of perceptual sequencing problems. The feasibility of this method is demonstrated in two separate implementations. The first is in the context of mobile robot docking where our mobile robot uses four different vision and ultrasonic algorithms to position itself relative to a docking workstation over a long-range course. The second uses vision, IR beacon, and ultrasonic algorithms to park the robot next to a desired plastic pole randomly placed within an arena.

Formal Specification for Behavior-based Mobile Robots

D.C. MacKenzie, and R.C. Arkin,
Proc. SPIE Conference on Mobile Robots VIII,
Sep. 1993, Boston, MA, pp. 94-104.

This paper presents formalisms for describing societies of cooperating behavior-based mobile robots, including the coordination between members of homogeneous teams, members of heterogeneous castes, assemblages of behaviors on individual robots, as well as perceptual strategies within primitive sensorimotor behaviors. This formal language is intended to facilitate proving properties about systems described in it.

Reactive Control for Mobile Manipulation

J. Cameron, D.C. MacKenzie, K. Ward, R.C. Arkin, and W. Book,
Proc. 1993 International Conference on Robotics and Automation,
Atlanta, GA, May 1993, Vol. 3, pp. 228-235.

Research for executing large-scale motions of mobile manipulators is described. Mobile manipulators are mobile bases with an attached arm which function in an integrated manner. Motivation is given for moving the arm while the base is moving. This work applies reactive control concepts to achieve this type of motion. Tools for modeling integrated arm-vehicle kinematics and dynamics are discussed. Simulation results are presented.

Active Avoidance: Escape and Dodging Behaviors for Reactive Control

R.C. Arkin, W. Carter, and D.C. MacKenzie,
International Journal of Pattern Recognition and Artificial Intelligence,
Feb. 1993, Vol. 7, No. 1, pp. 175-192.

New methods for producing avoidance behavior among moving obstacles within the context of reactive robotic control are described. These specifically include escape and dodging behaviors. Dodging is concerned with the avoidance of a ballistic projectile while escape is more useful within the context of chase. The motivation and formulation of these new reactive behaviors are presented. Both simulation and experimental results using a robot in a cluttered and moving world are provided.

Buzz, An Instantiation of a schema-based reactive robotic system

Arkin, R.C., Balch, T., Collins, T., Henshaw, A., MacKenzie, D., Nitz, E., Rodriguez, R., and Ward, K.
Proc. International Conference on Intelligent Autonomous Systems: IAS-3,
Pittsburg, PA., pp. 418-427, 1993.

This paper describes the Georgia Tech Mobile Robot Lab entry in the 1992 AAAI Mobile Robot Competition.

Autonomous Helicopter Position Determination using an On-board Integrated Vision System

D.C. MacKenzie, and R.C. Arkin,
Proc. SME/MVA Applied Machine Vision Conference '92,
Vol. I, SME MS92-162, Atlanta, GA, June 1992.

During the 1990-1991 school year an autonomous unmanned aerial vehicle was developed at the Georgia Institute of Technology to compete in the Autonomous Unmanned Vehicle Society's aerial robot competition. This helicopter carried two Integrated Vision Systems (IVS) on-board for vehicle position determination and for target location. Each IVS consisted of a Charge Coupled Detector (CCD) connected directly to a 68000 based microcomputer. This allowed rapid image acquisition and processing to occur on-board the vehicle, removing the need for high bandwidth video links. Novel real-time image processing algorithms were created to allow rapid adaptive tracking of navigation landmarks and targets from the moving helicopter. The paper describes the commercially available IVS, the unique application of this industrial vision system on the helicopter, and the image processing algorithms developed.

Development of an Autonomous Aerial Vehicle: A Case Study

Baker, N.C., MacKenzie, D.C., and Ingalls, S.A.,
Journal of Applied Intelligence
Vol. 2, pp. 271-297, 1992.

This article describes the processes used to design and build an autonomous unmanned helicopter during the 1990-1991 school year at the Georgia Institute of Technology to compete in the Autonomous Unmanned Vehicle Society's aerial robot competition.