Scientists at the University of Cambridge in England have created a mother robot that not only constructs its own children but mimics the process of natural selection to improve their capabilities with each generation. It’s a robotic version of nature’s survival of the fittest. The mother robot analyzes the performance of each of the “children” it creates, and incorporates preferential traits into the next generation. The Mother robot makes each child better than the last. “As the mother creates them and puts them to work, she evaluates how they’re behaving, and she uses data from this behavior to create the next generation of robots,” explained research scientist Andres Rosendo.
The process begins with a robotic arm and a set of five plastic cubes with motors inside. Each cube features a unique “genome” made up of a combination of between one and five genes. This genome gives each cube its own set of attributes, relating to its shape, construction and motor commands. The thinking is that just as species in nature mutate and genes are deleted, added and merged to adapt to different environments, the mother robot would facilitate its own version of evolution. Given only a single command to build a robot capable of movement, without any human intervention or computer simulation, the mother robot did just that.
The robotic overlord was put to use across five separate experiments, in which it designed, created and then tested generations consisting of ten children. It then tested their performance by measuring how far each child was capable of traveling within a certain amount of time. The best performing children advanced to the next generation unchanged, while new genomes were formed for slower ones in the group through mutation and merging genes. This technique proved a success. The researchers report that the information gathered from each test drove the evolution of the following generation. So much so, that the robots in the final generation moved with an average speed of more than twice that of the fastest robots in the first generation. The mother was able to boost their performance by improving the design and inventing new shapes and gait patterns, including some designs that humans would not have been able to conjure up.
There’s no human intervention, except for a computer command to create a robot capable of moving from one place to another. The mother builds its children by gluing together pieces with small motors inside, in different configurations. Then it watches how quickly the children move, keeping the designs of the ones that moved the fastest. “The mother robot can actually build hundreds of child robots and see the performance of these child robots. And if their performance is good, keep their design for the next generation. And if bad, just let it go,” said Fumiya Iida, lead researcher.
The motivation to produce better children is controlled by the research team, which provides an incentive. “We program the robot based on some functions that define the reward the robot is going to get, depending on the construction that they make. They cannot change their own reward. In the case of the child robot the longer the distance the robot walks, the better the reward it receives,” said Rosendo. After several generations, the “children” were running twice as fast. “The mother robot generated 500 robots to see what one is good and which one is bad,” said Iida.
The researchers suggest the machines could be used in an auto plant, for example, where robot cameras scrutinize each car in the assembly line, evaluate any mistakes, and then design a better car. The overall aim of the project is to uncover ways robotics can receive a boost from aspects of the natural world. This means both through more intelligent machines and shaping the way that they move. These smart little blue cubes don’t look likely to mount a challenge to mankind any time soon, but do indicate how robots, when left to their own devices, may be better at crafting machines than we are in the not too distant future.
“Natural selection is basically reproduction, assessment, reproduction, assessment and so on,” says lead researcher Dr Fumiya Iida of Cambridge’s Department of Engineering. “That’s essentially what this robot is doing, we can actually watch the improvement and diversification of the species.”
For more information please visit: https://www.cam.ac.uk
