Research Interests

Research Interests

  • Memetics – In his 1976 book The Selfish Gene, Richard Dawkins proposed that a new replicator was born – the meme – when animals gained the capacity to imitate each other. This new replicator is not submissive to the replicators that gave rise to them – the genes – but instead is quite separate and posses their own replicator power. I’m interested in how memes propogate themselves (imitation), how memes are mutated (creativity and combining memes to make new ones) and how memes interact with other replicators like genes (meme-gene coevolution)
  • Robotic Imitation – Ultimately I would like to build a population of robots that are able to evolve memetically. This could be a new direction for evolutionary robotics and developmental robots, evolving behaviour patterns instead of the robots physical or neural structure. The advantage of memetic evolution is that a meme does not have to be constructed, they are simply produced “for free” because certain behaviour patterns spread through the population as robots imitate each others behaviours, and no imitation strategy is perfect. Before robots can evolve memetically several issues need to be addressed:
    1. We need a very simple and computationally cheap way of imitating behaviour
    2. How are imitated behaviours stored in memory? Especially if using neural networks
    3. How do robots switch between imitation and autonomous behaviour?
    4. How do robots discriminate between behaviours? (which robot should I copy?)
    5. How do robots decide which part of a behaviour to imitate (what should I copy?)
    6. Do robots need to understand each others intentions in order to evolve memetically?

Current Work
After reading a paper on evolving a neuro-controller for imitation (Pini G, Tuci E 2008) I was inspired to investigate the ability for a CTRNN to imitate and memorize new behaviours. I am currently running experiments in which a simple 2-wheeled robot with 8 range sensors and a 16-node CTRNN is evolved for it’s ability to imitate a ‘dummy’ robot performing random movements along a single 1D plane.

Post a Comment