Tools for Anybots that statistically monitor and compute information about fleets of robots.
I spent the summer of 2011 at an internship with Anybots, a small robotics startup in Mountain View. For a long time Anybots was mostly a research company, producing robots like Monty and Dexter (both pictured below). However, two years ago they began to design QB (pictured above), a telepresence robot that aims to act as a robotic avatar to allow users to project their presence to other people from afar.
During my short time at Anybots, I noticed a severe lack of infrastructure for monitoring the fleet of about ~130 robots out in the world. This was several million dollars’ worth of robots deployed over the country and in some cases outside of the country. A coworker of mine, Robert, was spending a tremendous amount of time every day manually checking what robots were online by SSHing into each one and reading the logs, which were often thousands of lines long. Because scalability is one of the things that can kill a startup and because I wanted to free up his time, I wrote
super_robert and Anystats to help him out by automatically gathering logs from robots and computing information and statistics about global robot fleet status.
Anybots research robots Monty (left) and Dexter (right)
Super_robert is a program that attempted to procedurally log into robots out in the field, pull down their logs and compute some information about them. This information was then passed over to Anystats, which statistically tracked, analyzed and prioritized the different events that appeared in the logs.
Depending on what state a robot was in, how long it had been that way, what the logs showed, etc., bugs and other faults could be automatically tracked and diagnosed (like sentry but for robots).
Written by Lucas Doyle