AgentLight : a multilateral negotiation strategy in multiple-issues domains
The increasing importance of business to business electronic trading has driven interest in automated negotiation to soaring heights. To support the automated negotiation research, and provide a unique benchmark for evaluating practical negotiation strategies in multi-issue domains, competitions lik...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2016
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/67402 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The increasing importance of business to business electronic trading has driven interest in automated negotiation to soaring heights. To support the automated negotiation research, and provide a unique benchmark for evaluating practical negotiation strategies in multi-issue domains, competitions like the Automated Negotiating Agents Competition (ANAC) have been introduced. This paper presents the negotiation strategy used in AgentLight which is developed in accordance with the regulation of the ANAC 2016.
The agent is first designed for bilateral negotiation using a Tit for Tat strategy which decides actions based on the behavior of the opponents. Asides from the Tit for Tat, the agent also deploys a time based concession function to further control the concession rate. And to analyze an opponent’s model, the agent adopts a frequency based strategy to model the opponent’s preference in linear utility space and analyzes the standard deviation of opponent’s offer in non-linear utility space in order to push the offer near the Pareto Optimal Frontier.
The Agent has been evaluated through the competition with other 5 agents in a tournament setting in bilateral negotiation and with another 5 agents in a tournament setting in multilateral negotiation under the GENIUS simulation environment. Although the results presented a workable agent for multilateral negotiation, some points still should be noted down in improving the agents in the future.
For future enhancement, other negotiation theories proposed for multilateral negotiation and algorithms on opponents modelling can be studied. Further testing is recommended |
|---|