Research programme
A focused research programme studying the mechanisms that allow populations of simple agents to solve difficult problems collectively.
We design optimisation methods that use heterogeneous behaviour, dynamic neighbourhoods and multiple interacting populations to navigate difficult search landscapes. The programme spans standalone, hybrid and ensemble methods.
We investigate swarm and evolutionary approaches to model selection, hyperparameter optimisation, clustering and prediction—particularly where conventional gradient-based methods are constrained.
Our robotics direction explores how distributed agents can coordinate without a single point of control, adapting collectively to uncertainty, resource constraints and changing objectives.
Biological behaviours including altruism, interspecific eavesdropping, quorum decisions and fission–fusion dynamics become computational mechanisms for information exchange and resource allocation.
Research method
Our work begins with a behavioural or organisational principle, translates it into a precise computational mechanism, and tests it against established optimisation benchmarks. Beyond headline performance, we examine why a method works through diversity, topology, information-flow and convergence analysis.
Questions guiding the laboratory
We investigate how local, restricted and heterogeneous information exchange changes collective search behaviour.
Dynamic topologies, multiple swarms and fission–fusion processes allow collectives to adapt their structure during search.
Altruistic transfer and role differentiation provide new ways to allocate computational attention across a population.