2400. Deep Representation Learning for Generating Candidate Solutions in Multi-Agent Negotiation Evolutionary Search
Invited abstract in session WE-11: Heuristics, stream Heuristics, Metaheuristics and Matheuristics.
Wednesday, 16:30-18:00Room: U2-200
Authors (first author is the speaker)
| 1. | Andreas Fink
|
| Chair of Information Systems, Helmut-Schmidt-University | |
| 2. | Jörg Homberger
|
Abstract
We address multi-agent decision-making problems that require the identification of mutually beneficial agreements within formally defined solution spaces. A typical application is the coordination of delivery sequences or machine schedules among autonomous companies in a supply chain. We propose a novel autoencoder-based method for generating candidate solutions in an evolutionary search negotiation framework. In this approach, candidate solutions are constructed iteratively through recombinations in the latent space, guided by the encoder–decoder structure of the trained autoencoder network. We examine the effect of different configurations for application scenarios and demonstrate the potential of the approach to improve solution quality and negotiation efficiency.
Keywords
- Metaheuristics
- Artificial Intelligence
- Agent Systems
Status: accepted
Back to the list of papers