REACTIVE CONTROL FOR MULTI-AGENT ROBOTIC SYSTEMS IN HOSTILE ENVIRONMENTS

ARPA Task #A447 Summary
July, 1995

Outline:

Project Summary
Significant Events
Presentation Charts

=== PROJECT SUMMARY ===

ORGANIZATION: Georgia Institute of Technology

SUBCONTRACTORS: None

PRINCIPAL INVESTIGATOR: Prof. Ronald C. Arkin

TEAM MEMBERS / GRADUATE STUDENTS:: Dr. Jonathan Cameron, Research Scientist
Khaled Ali, Ph.D. Student
Tucker Balch, Ph.D. Student
Douglas MacKenzie, Ph.D. Student

TITLE OF EFFORT: Reactive Control for Multi-Agent Robotic Systems in Hostile Environments

EXECUTIVE SUMMARY: The application of behavior-based robotics to a team of military vehicles is being conducted through the development of 3 major research products: formation control software to enforce spatial relationships between team members; MissionLab mission specification software to facilitate straightforward end-user premission planning capabilities; and team teleautonomy to provide seamless run-time human commander/robot team guidance.

OBJECTIVE: Our objective is to provide robust coordinated performance for teams of robotic systems operating in hostile environments through the use of a flexible reactive control system. Demonstrations are being conducted, both within our laboratory and in the context of ARPA's UGV Demo II Program. Test domains are assumed to be hostile, where individual agents may perish and electronic countermeasures may be in effect, potentially disrupting communications between the robots.

APPROACH:

Formation Control:: Specific behaviors for line, column, wedge, and diamond formations have been developed for use in military scout mission by extending our schema-based control paradigm. Various referencing schemes, including unit-centered and leader-referenced, are used to establish the basis for the spatial relationships between vehicles.
Mission Specification:: MissionLab is an integrated development environment for the construction, evaluation, reuse, and control of multiagent robot control software. Developing a multiagent robot configuration involves three steps: determining an appropriate set of skills for each individual agent; translating these skills into sets of suitable behavioral assemblages; and the construction of suitable coordination mechanisms to ensure that the desired assemblages are deployed correctly over the life of the mission.
Team Teleautonomy:: Two different means by which a human operator can interact with a team of vehicles are provided: Basic Teleautonomy where the operator appears to the control system as another reactive behavior, biasing the team to move in the direction specified while maintaining all other active behaviors such as obstacle avoidance; and Advanced Teleautonomy where the operator is permitted to alter, on the fly, the relative behavioral composition of the entire robotic team, for example by making it more aggressive when required by the situation.

PROGRESS: Substantial progress was achieved within the last year. Various formation control behaviors have been developed and tested both in simulation and within robotic vehicles. A preliminary version of MissionLab software has been developed and will be subjected to usability studies in the upcoming year for further refinement. Basic team teleautonomy has been completed and a preliminary version of advanced teleautonomy is under development.

PRODUCTS: Software delivered to Lockheed-Martin for integration into UGV Demo C includes formation control and basic teleautonomy. Both of these have been integrated into their system and will be used in Demo C. MissionLab has been made available to the ARPA UGV Demo II Community and has been used by the University of Texas at Arlington to develop and test their sensor pointing algorithms.

FY95 ACCOMPLISHMENTS:

Formation control: Quantitative analysis of formation performance was undertaken and documented in a conference publication (see below). Formation behaviors were extended to include non-holonomic vehicles (e.g., cars and trucks). A rule-based formation expert was developed to assist military mission planners in selecting appropriate formations for a given mission.
MissionLab mission specification software has been released via Internet to the research community. (See MissionLab.) A conversion to the ARPA UGV standard IPT communications package was completed, Two internal compilers, CDL (configuration description language) and CNL (configuration network language) were finished and will be demonstrated at UGV Demo C.
Team Teleautonomy: Conversion of the schema-based controllers into a form compatible with the ARPA UGV arbitration system was completed. Integration of the graphical user interface into the Hughes STX premission planning software was accomplished. This software will be shown in a tech demo at Demo C.
Technology insertion of formation control into Lockheed-Martin's Demo C vehicles. Integration of formation expert and basic teleautonomy into Hughes STX premission planning software.

FY96 EVENTS:

In preparation for ARPA UGV Demo II (schdeuled for 8/96) Georgia Tech will make significant improvements in their current software for control of teams of robotics vehicles, including a more complete set of formation control software and improved teleautonomous control by the operator. This will include other software components such as an upgraded formation expert. The current early versions of these systems are important to the success of Demo C (July 1995). It is essential to incorporate the lessons learned from these initial demonstrations and to produce improved, more capable, and more robust versions of this software for inclusion in Demo II. Georgia Tech plans to participate in Demo II at Fort Hood, Texas.
The MissionLab system will be continuously developed and improved and is central to one Ph.D. dissertation. Another release is planned for 12/95. This version will incorporate advanced mission specification techniques, robot software configuration tools, and reusable control software libraries, as well as advanced team teleautonomous control concepts. Initial versions of many of these components constitute part of the release of MissionLab planned for 7/95. The subsequent release in 12/95 will improve and complete the existing components, adding many new capabilities and simplifying the duties of the operator and making the system more powerful and robust.
Several usability studies will be conducted during FY96 to evaluate the usefulness of the mission specification tools and the teleautonomous control concepts and interfaces. In order to obtain more reliable and helpful feedback, part of these evaluations will include testing by military personnel and military students here at Georgia Tech.
Test some of our ideas on our AM General HMMVW ("Hummer"):

TECHNOLOGY TRANSITION: In both ARPA UGV Demo C (7/95) and Demo II (8/96), Georgia Tech will provide software that is essential to the success of these important demonstrations. This includes vehicle formation control software, formation expert software for premission planning, and teleautonomous control software for simple operator control of teams of vehicles. This software will be incorporated into the HMMWV vehicles and operator workstation software by Lockheed-Martin. Portions of this software has been used by other groups. For instance, the formation expert has been used by the researchers at the University of Michigan as part of their premission planning system. It has also been incorporated into the Hughes STXmcu operator workstation. Plans for FY96 include expanding and improving all of these software components to help make the UGV Demo II systems more capable and robust.

The MissionLab mission specification and configuration software will be demonstrated as part of a Technical Demo during ARPA UGV Demo C in July 1995. This demo will involve collaboration of Georgia Tech and the University of Texas at Arlington (UTA) in a joint demonstration. Note that UTA has already used MissionLab software to help verify their sensing algorithms for this joint tech demo. Additional requests to use MissionLab have come from the Naval Research Laboratories and a joint US/Mexico research effort. The system is available through the internet (MissionLab). Plans for the FY96 include releasing improved and extended versions of MissionLab.

PUBLICATIONS:

Balch, T. and Arkin, R.C., "Motor Schema-based Formation Control for Multiagent Robot Teams", 1995 International Conference on Multiagent Systems, San Francisco, CA, pp. 10-16.
MacKenzie, D., Cameron, J., Arkin, R., "Specification and Execution of Multiagent Missions", to appear Proc. 1995 Int. Conf. on Intelligent Robotics and Systems (IROS '95), Pittsburg, PA.
Arkin, R.C. and Balch, T., "AuRA: Principles and Practice", submitted to Journal of Experimental and Theoretical Artificial Intelligence, 1995.
Arkin, R.C., "Reactive Robotic Systems", article in Handbook of Brain Theory and Neural Networks, ed. M. Arbib, MIT Press, pp. 793-796, 1995.
Arkin, R.C. and Balch, T., "Communication and Coordination in Reactive Robotic Teams", to appear in Coordination Theory and Collaboration Technology, ed. G. Olson, J.B. Smith, and T. Malone, 1995.
Arkin, R.C. and Ali, K., "Integration of Reactive and Telerobotic Control in Multi-agent Robotic Systems", Proc. Third International Conference on Simulation of Adaptive Behavior, (SAB94) [From Animals to Animats], Brighton, England, Aug. 1994, pp. 473-478.
MacKenzie, D., Cameron, J., and Arkin, R.C., "Specification and Execution of Multiagent Missions", Technical Report GIT-COGSCI-95/02, Georgia Tech, 1994.
Balch, T. and Arkin, R.C., "Communication in Reactive Multiagent Robotic Systems", Autonomous Robots, Vol. 1, No. 1, pp. 27-52, 1994.

DATE PREPARED: June 28, 1995

=== SIGNIFICANT EVENTS ===

MissionLab

The MissionLab system has developed significantly this past year and will have a new public release (including source code) within the next few weeks. This system has several components which address efficient multiagent mission specification, configuration, and execution. The MissionLab configuration editor allows a non-expert operator to specify robot missions and automatically configure and compile multiagent vehicle software. The basis for this achievement is a better theoretical framework for the expression of multiagent systems and missions and more effective ways to reuse previously developed software components. The MissionLab system can execute a wide range of military scout missions using simulated or real robotic vehicles simply and effectively. MissionLab also incorporates tools necessary to control teams of robotic vehicles moving in formation or under operator teleautonomous control.

Participation in ARPA UGV Demo C

Software developed under this contract has been incorporated into the ARPA UGV Demo II program, particularly Demo C which occurs in July, 1995. Several control components for multi-vehicle motion have been developed, supplied to Lockheed-Martin, and integrated into the Hughes STX vehicle control and operator workstation. Software for maintaining the relative positions of several vehicles (in other words, their formation) has been integrated and has proven so successful that it will be featured in several different demonstrations during UGV Demo C. Basic teleautonomy software has also been installed and will be demonstrated in two Tech Demo components during Demo C, allowing the operator to directly control a team of HMMWV vehicles.

=== PRESENTATION CHARTS ===

SCHEDULE:

[07/95] Release version 0.85 of MissionLab.
[07/95] Participate in UGV Demo C.
[12/95] Release improved version of MissionLab (v1.0)
[02/96] Finish full suite of formation control behaviors.
[05/96] Complete usability studies on teleautonomous system.
[05/96-08/96] Coordinate with LMC for UGV Demo II demos.
[08/96] Participate in UGV Demo II.
[11/96] Final release of MissionLab.

BACKGROUND / HISTORY:

This project resulted from earlier developments in schema-based reactive control systems developed by Ronald Arkin at Georgia Tech. In particular, recent work, funded by the National Science Foundation, in multiagent robotic control serves as the basis for this research. Reactive control systems have consistently proven useful in dynamic and hazardous environments as typified by the battlefield.

TECHNICAL APPROACH:

Develop more improved and robust formation behaviors through innovative spatial strategies.
Extend conceptual framework of team specification to simplify the creation and execution of complex missions.
Expand reuse of software components by extending the library of software for various robotic skills and behaviors.
Improve teleautonomous techniques for control of teams of robotic vehicles.

OBJECTIVES:

Provide robust motion control for teams of robotic vehicles.
PAYOFF: Improved ability to field robotic systems for dangerous military missions as demonstrated in the ARPA UGV Demo II program.
Extended mission creation, reuse, and execution capabilities.
PAYOFF: Enhanced ability for a single non-expert operator to create a mission semi-automatically that controls a team of robotic vehicles.
Efficient teleautonomous control of robotic team behavior.
PAYOFF: Improved ability of a non-expert operator to directly affect the performance of the vehicle team in unpredictable situations.

TECHNICAL CHALLENGE:

Controlling and specifying a single robotic agent in a hazardous environment is a difficult task. Controlling teams of them further compounds the problem. In order to provide robust coordinated control, new ways for envisioning a team of robotic agents as a single unit are applied. These behavioral control and specification methods provide the basis for all the research conducted under this contract. By viewing a human operator as another internal robotic behavior spread throughout many robotic agents, simpler teleoperation is permitted significantly reducing cognitive overload.

MissionLab Home Page

Mobile Robot Laboratory Home Page

College of Computing Home Page

Georgia Tech Home Page

Contact:: Prof. Ronald C. Arkin

Mobile Robot Laboratory

College of Computing

Georgia Institute of Technology

Atlanta, GA 30332-0280

Phone: (404) 894-8209

Fax: (404) 853-0957

email: arkin@cc.gatech.edu