Employing neural networks to assist negotiating intelligent agents

Artificial intelligence is one of the various disciplines that need to be employed towards the vision of ambient intelligence. Mobile intelligent agents introduce a powerful technology that may assist the market penetration of services and products offered online in intelligent environments. Such agents have the potential to improve the efficiency, proactive behaviour and performance of computing and communication systems in such domains. In this framework, the design and evaluation of agents responsible for handling automated negotiations on behalf of their human or corporate owners is a challenging research field. In this paper these agents are enhanced with learning techniques, in order to better simulate the human intelligence and increase the profits of their owners. As mobile agents have reduced processing capabilities and may need to migrate to foreign network nodes, the learning mechanisms they employ should require minimal resources and be computationally efficient. The proposed learning technique is based on a specially designed neural network (NN), is quite lightweight, and is appropriate for agents that represent clients in automated negotiations in intelligent environments. Exploited by agents that use a fair relative tit-for-tat negotiation strategy, it aims to increase the ratio of successful negotiations and maximize the utility of the client. The designed NN-assisted negotiation strategy has been empirically evaluated via numerous experiments under various conditions