CAEDA Cognitive Architecture

Academic poster for CAEDA presented at Projects Day 2011, University of Johannesburg

Distributed cognitive architecture for multi-agent learning

The CAEDA Cognitive Architecture is a cognitive architecture developed by Tristan Barnett that is designed to support multiple agents which must cooperate to solve a learning problem.

Background

A key question in machine learning is how can multiple agents learn how to work together to solve a problem? Cognitive architecture is inspired by the natural phenomenon of cognition, which has evolved over millennia to cope effectively in solving difficult problems. However, more research is needed on cognitive architecture which supports multiple agents sharing knowledge and learning together efficiently.

Approach

This research has developed the CAEDA Cognitive Architecture, which supports cooperative multi-agent learning.[1][2] CAEDA has been applied to simulated search and rescue robot tasks. In the simulations, tank-tracked robots learn efficient methods to cooperatively locate missing persons inside a collapsed building.

Conclusion

CAEDA is shown to offer a scalable and flexible solution to multi-agent learning in complex environments and may prove to be a useful tool in robotic learning.

References

  • [1] Tristan D. Barnett and Ehlers, E.M., April 2010. Cloud Computing for Synergized Emotional Model Evolution in Multi-agent Learning Systems. In Proceedings of the 8th International Symposium on Tools and Methods of Competitive Engineering, TMCE 2010, Ancona, Italy, April 18–22, 2010. pp. 841-854.
  • [2] Tristan D. Barnett and Ehlers, E.M., 2010. Cloud Computing for Synergized Emotional Model Evolution in Multi-Agent Learning Systems. Strojniški vestnik – Journal of Mechanical Engineering, vol. 56, no. 11, pp. 718–727. Available online: http://en.sv-jme.eu/archive/sv-jme-volume-2010/sv-jme-56-11-2010/