Autonomous and Intelligent Systems Group

The Autonomous and Intelligent Systems (AIS) group conducts research into the development of technologies where autonomy, intelligence, and interaction are key factors for the computer systems of the future in industry, engineering or commerce.

Research in AIS spans three broad, overlapping themes, each of which is described here along with research programmes within each theme, plus details of AIS research publications and funding.

• Machine Optimisation and LEarning (MOLE);
• Software Agents (SA);
• Intelligent Computing Environments (ICE).

AIS actively collaborates with other research groups and centres in School of Informatics and elsewhere, including the Music Informatics Research Group in the study of artificial intelligence, and the world-leading Centre for Interactive Systems Research in the application of these technologies to information retrieval.

The AIS group is pleased to be a node of both EVONET and AgentLink, the EU networks of excellence in Evolutionary Computation and Agent-Based Computing. AIS members are also involved in KDnet, Knowledge Discovery Network, Agentcities.UK, the Interdisciplinary Scheduling Network and ARTIST: Network for Artifical Immune Systems.

AIS members are also active in the UK AI societies, serving on the committees of both the Society of the Study of Artificial Intelligence and the Simulation of Behaviour (AISB) and the British Computer Society's Specialist Group on Artificial Intelligence (BCS SGAI).

Machine Optimisation and LEarning (MOLE)

This research theme focuses on the principled application and development of intelligent systems-based technologies for the following tasks.

• Optimisation: locating a "best" or "good enough" solution to a problem out of a (vast) number of possibilities with time and computing resource constraints (e.g. timetabling). Real-world optimisation problems are often NP-hard thus requiring efficient optimisation solvers to be tractable.
• Machine Learning: deriving a predictive and/or descriptive model of a set of data that contains examples we wish to learn from. The application of machine learning is often called "Data Mining" or "Knowledge Discovery from Databases" (KDD).

The AIS group is a node of EVONET, the EU's network of excellence in Evolutionary Computing. Research programmes within the MOLE theme include the following.

• Knowledge-Based Design of Evolutionary/Local Search Optimisers
• Operational Research Applications
• Genetic Programming
• Neural Networks
• Bioinformatics

Software Agents

The development and application of agent technology is a active research theme of the AIS group. Agents are software-controlled systems situated in an environment which are:

• autonomous: agents operate without direct outside intervention, and have some control over their actions and internal state;
• social: agents interact with other agents and possibly humans;
• reactive: agents perceive their environment, and respond effectively to changes that occur in it;
• pro-active: agents do not simply respond to their environment, but exhibit goal-directed behaviour and take the initiative.

The AIS group is a node of AgentLink, the EU's network of excellence in Agent-based Computing, and helps coordinate the AgentLink SIG, Agents that Learn, Adapt and Discover (ALAD). Research programmes include:

• Computational and Animal Learning: On the one hand, we are interested in developing reinforcement learning algorithms based on current theories and models of associative learning. On the other hand, we are building computational models of classical and instrumental conditioning along with psychologists from UCL, Nottingham and Leicester. For this purpose we have recently formed the Computation and Animal Learning SIG along with Dr Esther Mondragon (UCL)
• Co-ordination in Multi-Agent Systems: We study how agents in complex, dynamic MAS scenarios (e.g., intelligent ambience, business re-engineering, traffic control, and geographic simulation) get effectively co-ordinated using new concepts such as rights and intelligent modeling.
• Communication in Multi-Agent Systems: How agents do retrieve the correct communication meaning when interacting is a though problem in Agent Communication Languages (ACLs). We are building and evaluating different algorithms based on Generalized Implicatures.

Intelligent Computing Environments (ICE)

The main theme of Intelligent Computing Environments (ICE) is to develop virtual electronic environements that are embedded in ordinary social environments, in order to augment and support what ordinary people do in their everyday activities. The approach taken is to:

• build computing technologies that interpret information from sensors;
• allow the objects and entities contained to interact in distributed and possibly ad-hoc networks;
• use computational logic to specify the reasoning and intelligence required;
• use the notion of games as a metaphor for the interactions;
• apply the notion of artificial societies to support people's interactions.

The work of ICE contributes towards the EU Global Computing initiative.