I wanted to post an update about our ongoing DARPA SubT Challenge blog post series. This week the team has put together a really fantastic tutorial that demonstrates how to add some very basic perception capabilities to a simulated robot. If you are unfamiliar with DARPA SubT Challenge, both the virtual and systems teams compete to find artifacts in dangerous subterranean environments. There are a variety of artifacts including backpacks, cell phones, heated dummies that represent a human survivor, and more abstract things like a simulated gas leaks and vents. I covered some of this in my SubT Challenge Urban Circuit post.
In this tutorial we walk you through how to add the ROS Darknet package to your robot and modify its deep learning capabilities to detect a few of the artifacts. We also demonstrate how to take the raw detection output and precisely pinpoint its position using the robot’s SLAM system. This is a great tutorial for those who are learning about perception and SLAM and want to tie the two together.
We also recently pushed new features and tutorials for the SubT Virtual Testbed to simulate marsupial robots and WiFi breadcrumbs in Ignition Gazebo. A marsupial robot is exactly as it sounds – a big robot carrying a smaller one. Teams can explore more efficiently by launching aerial robots (which have lower battery life) deep underground from the backs of their larger ground robots. See how here.
Wireless breadcrumbs are one of the more innovative aspects of the SubT Challenge where competition robots drop small wireless repeaters to extend their communication range. This strategy for extending wireless connectivity has been used by a number of teams in the Systems Competition and now it is available in simulation as well. You can find the full tutorial here.
SubT Hello World Tutorials: