Related Experiment Videos
Learning to cooperate in solving the traveling salesman problem
1Lamar University, Computer Science Department, PO Box 10056, Beaumont, Texas 77706, USA. dqi@cs.lamar.edu
International Journal of Neural Systems
|May 25, 2005
Summary
Cooperative teams of self-interested agents can outperform individuals. This study introduces a bidding approach using reinforcement learning to achieve effective team cooperation while preserving agent autonomy.
Area of Science:
- Artificial Intelligence
- Multi-Agent Systems
- Reinforcement Learning
Background:
- Cooperation enhances task performance in multi-agent systems.
- Achieving cooperation among self-interested agents is a significant challenge.
- Maintaining individual agent autonomy is crucial for flexibility and generality.
Purpose of the Study:
- To present a novel approach for fostering cooperation among self-interested agents.
- To enable teams to outperform the best single agent and the average performance of individual agents.
Main Methods:
- An approach based on bidding is proposed.
- Reinforcement values are acquired through reinforcement learning.
- The approach was tested and analyzed to evaluate its effectiveness.
Main Results:
- The cooperative team demonstrated superior performance compared to the best single agent.
- The team's performance also surpassed the average performance of individual agents.
- The bidding approach facilitated effective cooperation.
Conclusions:
- The proposed bidding strategy effectively promotes cooperation in multi-agent systems.
- Agent autonomy is maintained while achieving enhanced team performance.
- This method offers a viable solution for coordinating self-interested agents.