Home / Blog / Evolutionary Robotics

Evolutionary Robotics

The closed loop for evolutionary robotics. A robot is designed (1), then it is “manufactured” and starts its life (2). It learns and reach a point when it duplicates (3). Image credit: Milan Jelisavcic et al. Frontiers in Robotics and AI

OK, today’s is April Fools’ day so here I go with a topic that looks like a prank, but it is not (or is it?).

(some) Robots will be able to replicate themselves and generate offsprings. More than that. Their replica might be a bit better, in evolutionary terms, than its parent. A recent article on Wired pointed out that robots have been completely un-interested, so far, in replication. But this is bound to change.

Replication for robots can be quite different from the one life on this Planet chose:

  1. for a robot there is no need to find another robot, have sex and generate an offspring, however there might be cases where a robot will replicate by taking into consideration several other robots
  2. A robot does not need to create a baby robot, rather it can create directly an “adult” robot, that could be a clone or represent an evolution with respect to its parent
  3. robotic evolution does not necessarily need to be “darwinian” -although for soft bots , that might be a strategy, it could also be Lamarckian

OK, so just for fun, it is April 1st, let’s take a look at a robot-mom.  There are actually two quite separate types of robots if we are looking at replication: hard bots (what you would usually consider a robot) and soft bots, robots made by bits only. The first category has an additional challenge to face: finding the hard stuff that can be used to create a robot. So if a robot is made of plastic and steel and decides to create a copy of itself it would have first to find the raw material.

An approach that some researchers have followed is to create a multitude of basic building blocks and then empower a robot to assemble them to create another robot. At MIT and Harvard researchers have demonstrated a self assembling robot (see clip) and a parent robot might follow a similar approach to create a offspring, provided the needed raw material can be found.

In addition to get the raw materials, all robot-parents to be need to have a willingness to replicate (all life, as far as we can tell, is programmed to replicate and finds replication a stimulating activity). As it happens with life, a robot needs to transfer to its offspring the “code of life”, i.e. the instructions that make that new robot functional and possibly able to further replicate. The generation of a baby vs generating a grown up is a matter of convenience. Life has found more convenient to generate babies, since it is cheaper -energy wise- and let the baby grow on its own (with more or less support from its parents, generally with no support at all). However, for a robot it may make more sense to generate an adult copy, since, in general, robot are as they are, they haven’t learnt to grow (physically, from a cognitive point of view modern robots grow, and quite a bit).

Soft bots clearly have no incentive in creating babies, it cost the same to create adult versions and it is much more efficient (and easy).

An intriguing question is if by replication a robot can create a better robot. Having learnt from experience what works well and what could be better, one could imagine that a robot will be able to “design” and then create an offspring that is better than itself. In this sense we can say that robotic evolution can be Lamarckian, rather than Darwinian, it follows a plan and learns from experience of the parent, it is not the result of blind selection processes.
This is a new area, evolutionary robotics. Last week, March 26-29, the first conference on Evolutionary Robotics and Artificial Life was held in Madrid.

We are taking the very first steps on this new path but I expect that the convergence of smart materials, artificial intelligence and robotics will pave the way to robots that will be able to create smarter and smarter offsprings far outpacing Nature in terms of evolution pace.

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the Industry Advisory Board within the Future Directions Committee and co-chairs the Digital Reality Initiative. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.