Attack Munition Real Time Assessment (CAMRA) Air
Force Eglin Munitions Lab Grant F08630-03-1-0005
The Cooperative Attack Munition Real Time Assessment (CAMRA) project
is a joint project between the Intelligent Software Agents Lab
in the Robotics Institute and University of Pittsburgh to develop
large teams of autonomous Wide Area Search Munitions (WASMs) that
can be controlled by a small number of human operators. A WASM
is a cross between a unmanned aerial vehicle - it can sense its
environment and react to changing situations - and a smart bomb
- it can destroy targets by hitting them. The Air Force sponsors
of the project envision having hundreds or even thousands of WASMs
flying in support of troops within a hostile battle space. Achieving
this vision requires overcoming some significant technical challenges.
CMU's role in this project is to develop the algorithms required
to achieve cohesive, flexible and robust coordination in the hostile
A number of
intertwined algorithms are required for effective coordination.
One algorithm needs to initiate, monitor and terminate joint plans.
Another algorithm needs to determine which information should
be communicated from one group member to another, based on the
costs, benefits and risks of that communication. Another key algorithm,
especially in dynamic environments, allocates roles to group members
to best leverage the abilities of the group members towards the
joint goal. Additional algorithms are required to manage access
to shared resources or form sub-groups or create plans. Notice,
that all the coordination algorithms are distributed and must
operate despite a noisy, dynamic environment. While algorithms
for each of these problems exist, they typically do not scale
up to the size of the group that we need to coordinate in the
CAMRA project. Thus, we are actively developing and extending
these critical algorithms to ensure they are appropriate for very
the generality of our work we are encapsulating the generic coordination
algorithms in domain independent "proxies" that operate in close
cooperation with domain dependent "control agents". These proxies
are lightweight Java processes that have been specifically designed
to meet the challenges of large scale, highly heterogeneous teams.
Two key areas where these proxies depart dramatically from previous
efforts are in the role allocation mechanism and communication
reasoning. The role allocation algorithm is a highly scalable
algorithm based on ideas from distributed constraint optimization.
The key is to represent roles as tokens and allow only the proxy
currently holding a token to assume that role. Using probabilistic
information about the overall situation, each proxy decides whether
to accept the role represented by the token or pass it on. By
intelligently passing and holding tokens, the group can rapidly
find good allocations and robustly adapt the allocations when
the situation changes. Second, we are developing novel algorithms
for communication reasoning that do not rely on the accurate models
of group members typically relied on in previous work.
- P. Scerri,
J. A. Giampapa, and K. Sycara., "Techniques
and Directions for Building Very Large Agent Teams,"
in International Conference Integration of Knowledge Intensive
Multi-Agent Systems (KIMAS '05), 2005. Invited Paper.
P. and Sycara, K. and Tambe, M, "Adjustable
Autonomy in the Context of Coordination," in AIAA
3rd "Unmanned Unlimited" Technical Conference, Workshop and
Exhibit, 2004. Invited Paper.
P., Farinelli, A., Okamoto, S. and Tambe, M. "Token
Approach for Role Allocation in Extreme Teams: analysis and
experimental evaluation," in Proceedings of 2nd
IEEE International Workshop on Theory and Practice of Open Computational
P., Xu, Yang., Liao, E., Lai, J. and Sycara, K. "Scaling
Teamwork to Very Large Teams," in AAMAS'04, 2004.
- Liao, E.,
Scerri, P. and Sycara, K. "A
Framework for Very Large Teams," in AAMAS'04 Workshop
on Coalitions and Teams, 2004.
- Xu, Y.,
Lewis, M., Sycara, K. and Scerri, P. "Information
Sharing in Large Scale Teams," in AAMAS'04 Workshop on
Challenges in Coordination of Large Scale MultiAgent Systems,
- P. Scerri,
E. Liao, Y. Xu, M. Lewis, G. Lai, and K Sycara, "Coordinating
very large groups of wide area search munitions," in Theory
and Algorithms for Cooperative Systems, World Scientific