Better Living Through Chemistry: Evolving GasNets for Robot Control

Abstract

This paper introduces a new type of artificial neural network (GasNets) and shows that it is possible to use evolutionary computing techniques to find robot controllers based on them. The controllers are built from networks inspired by the modulatory effects of freely diffusing gases, especially nitric oxide, in real neuronal networks. Evolutionary robotics techniques were used to develop control networks and visual morphologies to enable a robot to achieve a target discrimination task under very noisy lighting conditions. A series of evolutionary runs with and without the gas modulation active demonstrated that networks incorporating modulation by diffusing gases evolved to produce successful controllers considerably faster than networks without this mechanism. GasNets also consistently achieved evolutionary success in far fewer evaluations than were needed when using more conventional connectionist style networks.2

Synchronisation effects on the behavioural performance and information dynamics of a simulated minimally cognitive robotic agent

“…A single parent is then chosen through rank-based roulette selection, and the mutation operators are applied, producing a new individual, which is evaluated and placed back in the mating pool in a position determined by inverse rank-based roulette selection. For further details about the genetic algorithm, the reader should refer to Husbands et al (1998).…”

“…A geographically distributed genetic algorithm with local selection and replacement (Husbands et al, 1998) is used to determine the parameters of the system: the frequency of each of the 15 nodes, w i ∈ [0, 10], the 7 input weights z i ∈ [−5, 5] to odd nodes (0 otherwise), the matrix W N −1,o with 14 × 2 = 28 elements in the interval [−1, 1], a motor output weight s ∈ [0, 100] and the network update time t u ∈ [0, 50] (the number of time steps Equation 7 is updated before the phase-differences are used to calculate the motor output), resulting in a genotype of length 52 for a network with N = 15.…”

The Nonclassical Mind: Cognitive Science and Nonclassical Computing

“…The traditional model of connectionist networks used in cognitive science abstracts from brain physiology at the level of the discrete neuron. However, recent models such as the GasNets model, devised for robot control and related problems [15,30] include the non-local effects of neurotransmitters such as nitrous oxide, where the molecules are small enough to diffuse throughout the brain. This provides a level of non-synaptic communication within the brain.…”

Innately adaptive robotics through embodied evolution

“…Hornby evolved L-Systems to generate complicated multi-component virtual creatures [14]. Husbands used evolutionary robotics techniques to develop control networks and visual morphologies to enable a robot to achieve target discrimination tasks under very noisy lighting conditions [17]. Mihalachi and Munerato created a robot snake able to move like an earthworm [27][28].…”