Evolving organisms that can reach for objects

An evolutionary method based on selective reproduction and random mutation was used to evolve neural networks that control two types of simple organisms which can reach for objects using their 2-segment arm. One kind of organism does not move and can only capture an object if it is at reaching distance; the other can displace itself and therefore it first approaches an object and then captures it. Individual learning during lifetime to predict changes in the position of an object or of the hand relative to the organism's body helps in the evolution of the object reaching capacity, although inheritance of the weight matrix is strictly Darwinian. Finally. a more sofisticated fitness criterion which penalizes arm movements causes the more complex organism to move its arm only when an object is at reaching distance.