Page 41 - GCN, June/July 2018

function. Fitness is a score that tells us how well the robot is performing — for instance, how many wombats it counts per hour. The robots attempt the task, which gives us a fitness score for each robot. Some robots will be fitter than others, and the fittest robots are more likely to be picked as “parents.”
• Variation. Our parents create children. Crossover combines genes from both parents into a new “child” genome. Mutation slightly changes some of these genes as they pass from parent to child. We can make the children and get a fitness score for each, as before. Because the genomes of the children are similar to the genomes of fit parents, they are likely to be good at the task.
• Generations. We perform this loop many times, incrementally optimizing the population of robots to be suited to their task and environment.
‘Endless forms
most beautiful’
This is an exciting time to be evolving robots. Advances in 3D printing help us create a huge range of different robot designs, and new types of sensors and actuators only add to this burgeoning field of research.
Although we aren’t yet at a stage to match the wonderful complexity of natural life, artificial evolution can generate simulated robots that are adapted to their environments, perform simple tasks and quickly adjust to damage during a mission.
Evolution can also assist a designer by quickly finding good combinations of a robot’s physical attributes.
Crucially, artificial evolution gives us a tool to adapt robots to their environmental niches, which might be the key to successfully deploying robots for useful tasks in the harshest environments.•
David Howard is a research scientist at CSIRO’s Data61, Australia’s data innovation network. This article first appeared on The Conversation.
Testing robots in simulated underground environments by Matt Leonard
The Defense Advanced Research Projects Agency wants to be able to autonomously map and explore underground environments. The agency plans to run a Subterranean (SubT) Challenge to explore new ways to quickly map, navigate and search underground environments, such as human-made tunnel systems, underground mass transit and municipal infrastructure, and naturally occurring cave networks.
Teams are invited to propose novel methods for tackling time- critical scenarios in mapping underground networks under conditions that are too hazardous for human first responders.
The submissions will be tested
in physical environments, but DARPA will also provide simulated environments to help competitors develop their technology, according to the agency’s sole-source notice.
Open Robotics will develop
a simulation platform to mirror
the physical, real-world test environment as closely as possible. It will be built with the Robot Operating System, a set of software libraries and tools for creating robot applications, and Gazebo, a platform that allows developers to accurately simulate robots in complex indoor and outdoor environments. The test bed will run on CloudSim, which provides robot simulation as a web application. Open Robotics will make the simulation software available on an open-source basis.
The Gazebo robotics simulator was used in NASA’s Space Robotics Challenge to improve autonomy for and manipulation of the R5 Valkyrie robot.
The final competition in the SubT Challenge is expected to take place in 2021. •
GCN JUNE/JULY 2018 • GCN.COM 41