I am a PhD student under the supervision of Dr. Inman Harvey in the Center for Computational Neuroscience and Robotics at the University of Sussex, UK. I am currently a ‘visiting research associate’ in the new Dynamical Systems and Robotics lab led by Dr. Randall Beer at the University of Indiana in Bloomington.

The idea of this blog is to keep notes on my research. A key factor that has pushed me towards maintaining such a thing seriously has been Eldan Goldenberg’s own research blog. Although the notes are for myself, I find it an interesting experiment to make them public. Ideally this could help me interact with other people working on similar ideas.

So, what I am working on? There has been much progress towards understanding behavior as the interaction between brains, bodies and environments in the alife field. However, learning behaviour continues to be a process understood as taking place ‘inside the brain’, with a particular attachment to synaptic plasticity in neuronal networks.

My investigations into the mechanisms underlying learning and memory behavior follow the evolutionary robotics methodology. The key strenght of which is to facilitate the exploration of a broader set of possible mechanisms, by minimising the a priori assumptions built into our models and allowing for evolutionary search over a broad space of possibilities.

Why learning and memory? One of the greatest challenges faced by autonomous robotics today is generating (and understanding) robots that can adaptively modify their behaviour according to changes in their environment. The motivation for my work is to evolve and analyse agents that learn during their lifetime without introducing learning mechanisms a priori into the agent’s structure or internal mechanisms.

In my work, that which is to be learned is a feature within a continuum from the environment. The tasks are loosely inspired on imprinting behaviour (in for example C. elegans, birds, etc.) and they are cognitively relevant because they involve memorization and decision-making in very simple models. The interest in the synthesis of such agents is in the analysis of the evolutionary dynamics as well as the evolved mechanisms and interactions using the language of dynamical systems theory. The main outcome of this thesis will be to provide an understanding of learning and memory behaviour that helps shift the focus from ‘accurately representing’ the environment to dynamically engaging it (with a body) so as to generate coordinated patterns of behaviour.

I will be coming back to make this clearer as I make more progress on my work.

Eduardo