Darwinian Robotic Evolution

A “mother robot” (A) is used for the automatic assembly of candidate agents from active and passive modules

Engineers have created evolving robotic systems that develop, learn from previous experience and create an improving functionality.  The purpose is to create self-learning, improving robots that absorb from their recent experience and build a new generation of improved robots that have learned and adapted from their environment.

This does sound like a science fiction thriller from the nineteen sixties but this is real development that this and the next generation of humans and robots will evolve from.

Teams of engineers from Cambridge and Zurich are working together and have published a paper in PLOS One.
Machines are physically constructed and their performance is analyzed without simulation and human intervention to incrementally improve their functionality. They are at present small child like robots with motors inside which are built by the parent/mother robot which 3Dimentialaly puts the new generation together and sets them off. 
This careful implementation of the variable construction process allows machines to explore a large set of solutions with initially unknown final structures. 

The demonstrations show the feasibility of the model-free evolution of a physical system. The evaluation of a candidate’s fitness is done with a physical robot, producing real data in a time-intensive process. 

Extending the model-free development could allow machines to autonomously and adaptively modify their mechanical structures together with their controls, similar to the animals’ functions observed in Darwinian evolutionary developmental processes.
Plos: http://bit.ly/1N6Tqlg