‘Motherly robot’ evolves creations to improve performance

Researchers from the University of Cambridge have developed robots capable of reproducing, so to speak-– they have invented a “mother” robot which can design, build, and even test out its own “children,” then use its experience to improve the performance of future generations.

According to Engadget, the original robot can “give birth to” (build) 10 baby cube-bots at a time. Each of these “pint-sized machine children” had a motor and between one and five plastic cubes, moving and performing different tasks under the watchful eye of their robotic mother. Check it out:

The mom leaves the fastest ones on their own, but slower designs are changed in the next generation, similar to the evolutionary process. By passing down the best traits and improving upon the weaker ones, the mother robot was able to double the overall speed of its children by the end of the fifth generation of robots – all without human intervention.

As lead researcher Dr. Fumiya Iida of Cambridge’s Department of Engineering, whose work is detailed in a new PLOS One paper, explained in a statement, “Natural selection is basically reproduction, assessment, reproduction, assessment, and so on. That’s essentially what this robot is doing – we can actually watch the improvement and diversification of the species.”

‘Survival of the fittest’ observed in robots

Dr. Iida and colleagues from ETH Zurich conducted a series of five experiments, each of which involved the mother robot creating and constructing a generation of 10 children. As it observed the created cube-bots, the mother selected the preferential traits and continued to pass them down from one generation to the next, ensuing the survival of the fittest individuals.

Each child robot had unique “genomes” of one to give different genes, each of which contained data about its shape, construction and motor command, the researchers explained. As happens in nature, the evolution in the robots occurred through mutation (modification, addition or deletion) and crossover (the formation of a new genome by merging genes from two individuals).

The mother determined which children were fittest through a trial where the cube-bots were given a set amount of time to travel as far from their starting position as possible. The successful ones remained unchanged so that their abilities could be preserved, while the weaker robots were subject to mutation and crossover in order to improve.

Overall, the study authors found that design variations emerged and overall performance on the time trials improved with each generation. The fastest in the final generation travelled twice the speed, on average, as the fastest from the first generation. This was attributed to tweaked design parameters, and the mother’s use of new shapes and gait patterns for the offspring.

“One of the big questions in biology is how intelligence came about – we’re using robotics to explore this mystery,” Dr. Iida said. “We think of robots as performing repetitive tasks, and they’re typically designed for mass production instead of mass customization, but we want to see robots that are capable of innovation and creativity.”


Image credit: Youtube/Cambridge University