Multi-agent learning & evaluation for open world games - Sam Devlin, Microsoft Research

Описание: The AI Programme at the Turing will host an interactive UK Symposium on Multi-Agent Systems (UK-MAS). The goal of the symposium is to bring together UK-based research labs at universities and industry who have a significant focus on MAS research, to explore the MAS research landscape in the UK.

We define MAS in a broad sense, including multi-agent interaction/learning/planning, negotiation, game theory, mechanism design, opponent modelling, game playing, swarm robotics, HRI/HCI, autonomous driving, social networks, verification, etc.

There will be space for talks by people from academia and industry as well as discussion sessions and networking opportunities. A concrete outcome for this event will be the creation of a virtual map with the locations and research foci of the various UK-based labs, as well as a shared mailing list to disseminate relevant opportunities, with the hope that these will facilitate future collaborations.

---

Multi-agent systems (MAS) are a core area of research of modern artificial intelligence. A multi-agent system consists of multiple decision-making agents which interact in a shared environment to achieve common or conflicting goals. An abundance of applications and problems can be usefully modelled and analysed using MAS methodology.

Examples include markets in which artificial and human trading agents negotiate contracts; autonomous vehicles driving in dense urban traffic; teams of robots and humans working together in Industry 4.0 factories and warehouses; drone swarms used for airport surveillance and patrolling; and game playing agents in commercial or educational computer games.

Research in MAS spans a very broad range of technical problems and uses concepts from artificial intelligence, machine learning, game theory, behavioural sciences, operations research, and simulation/verification, among others.