Objectives

  • To analyse at a fine scale how larval olfactory behaviour is controlled and altered by associative conditioning with attractive and aversive gustatory stimuli, and how memory-based decision making is organised according to motivational and contextual states. This will be tightly linked to agent-based computational models of the larval behaviour that allow us to test crucial computational principles required to support the observed behavioural phenomena.
  • To build and validate integrative models at the level of neural circuits and link them to the characterisation of neural function in larvae through controlled loss-of-function and gain-of-functionexperiments, functional imaging and functional perturbations through optogenetics. This will make use of unique and novel tools to gain unprecedented correspondence between model and system manipulations.
  • To abstract dynamical principles and derive novel, generalisable algorithms and architectures that can be used to enhance the learning and anticipatory capabilities of machines, and apply them in extended scenarios. This will involve a proof of concept that such autonomous learning capacities in real environments can be achieved with relatively minimal computational power.

The aim is a new form of IT that is complementary to the current dominance of ‘big data’ approaches – our key idea is that small but active systems can, like animals, locally discriminate and remember only the effective cues needed for the ongoing task. Although our main aim is to demonstrate such capabilities in real world robot systems, there may be parallels in the information environment – gradients of a cue that help to locate target information more efficiently. Another exciting area of potential application is to minimalist adaptive interactive devices that can learn associations that allow efficient anticipation of user actions. Looking further into the future, some of the neurogenetic methods we will explore point towards the potential use of larvae themselves as engineered computational devices for critical signal processing tasks.

Expanding on our three current objectives:

  1. Understanding how associative learning is embedded in behavioural control
  2. Understanding the neural circuit
  3. Enhancing the learning capabilities of machines

2 Responses to Objectives

  1. Pingback: This Creepy Mechanical Maggot May Be The Future Of Robotics – Co.Design (blog)

  2. Pingback: This Creepy Mechanical Maggot May Be The Future Of Robotics | Co.Design | business + design – Co.Design (blog) | Screenny

Leave a Reply

Your email address will not be published. Required fields are marked *