A Deep Reinforcement Learning Approach to Concurrent Bilateral Negotiation

We present a novel negotiation model that allows an agent to learn how to\nnegotiate during concurrent bilateral negotiations in unknown and dynamic\ne-markets. The agent uses an actor-critic architecture with model-free\nreinforcement learning to learn a strategy expressed as a deep neural network.\nWe pre-train the strategy by supervision from synthetic market data, thereby\ndecreasing the exploration time required for learning during negotiation. As a\nresult, we can build automated agents for concurrent negotiations that can\nadapt to different e-market settings without the need to be pre-programmed. Our\nexperimental evaluation shows that our deep reinforcement learning-based agents\noutperform two existing well-known negotiation strategies in one-to-many\nconcurrent bilateral negotiations for a range of e-market settings.\n