The potential synergy between instance-based pattern recognition and
means-end (possible world) reasoning is explored in the domain of
multi-agent air mission simulations. Graph matching and inductive logic
programming techniques are applied to traces of aeroplane behaviour during
flight manoeuvres. This enables the agents to recognise what other agents
are doing and to abstract about their activity, at the instrumentation
level. The belief, desires and internationality (BDI) model of means-end
reasoning is then used to deliberate about and invoke standard operating
procedures, based on perceived activity. The reasoning model constrains the
recognition process by framing queries according to what a pilot would
expect during the execution of the current plan(s). Of particular interest
is how the processes of graph matching and inductive logic programming have
been coupled for any-time recognition at run time. The importance of
capturing relative information in these multi-agent simulations is
emphasised, including self-agent, agent-agent and agent-environment
relationships.