Path integration is a widely used method of navigation in nature whereby an animal continuously tracks its location by integrating its motion over the course of a journey. Many mathematical models of this process exist, as do at least two hand designed neural network models. Two one dimensional distance measuring tasks are here presented as a simplified analogy of path integration- and as a first step towards producing a neuron-based model of full path integration constructed entirely by artificial evolution. Simulated agents are evolved capable of measuring the distance they have travelled along a one dimensional space. The resulting neural mechanisms are analysed and 'discussed, along with the prospects of producing a full model using the same methodology.
展开▼
