By Jonny Lupsha, Wondrium Staff Writer
Amazon’s first home robot puts Alexa on wheels and carries small items around your house. It also contains several features and upcoming accessories for everything from pet feeding to health care. How do robots learn to do jobs?
Astro is Amazon’s first ever home robot, releasing later this year for $999—or $1,499 after the introductory window closes. The small bot works like a smart assistant that can follow its user around, set alarms and reminders, and even carry drinks or similar small items to other members of the house. While the user is away from home, they can pull up Astro’s companion app to visually check in on their home using its 42″ retractable periscope camera.
Several companies are even working with Amazon to make accessories to broaden Astro’s use. Popular dog camera Furbo will make a tray that shoots dog treats from Astro’s small trunk while an upcoming subscription known as Alexa Together will work with Astro to help with remote care for the elderly.
How do we get a robot to perform a task? In his video series Robotics, Dr. John Long, Professor of Biology on the John Guy Vassar Chair of Natural History at Vassar College, said the robot has to understand two things: the workplace and the task itself.
A Clean Workplace Is a Healthy Workplace
A seemingly simple task for a robot is one that will take out the trash. Considering that task, Dr. Long said, a robot must recognize its environment and understand what its task really is.
“When we look at a workplace, a big deal in robotics is to think about how stable it is,” he said. “A stable workplace we call a structured workplace, and it’s one in which the world doesn’t change, or doesn’t change much. It maintains its structure: Most objects don’t move, and those that do, do so in predictable and regular ways—and people don’t come and go willy-nilly.”
A factory floor with an assembly line, he said, is a classic example of a highly stable and structured workplace.
Unstructured workplaces are, obviously, the opposite. These are busy, crowded places, especially in outdoor environments. Most workplaces—homes, schools, and offices—fall somewhere in between. Dr. Long said that knowing the level of structure in a work environment for a robot lets the programmer teach that robot what to pay attention to and what to ignore.
The Task at Hand
When it comes to taking out the trash, the idea is simple, but the execution is far from it.
“We need to turn on, somehow, our robot when the trash needs to be emptied [because] we don’t want our robot always worrying about when the trash is going to be there and using a lot of energy,” Dr. Long said. “We also need to navigate to the location: If the robot is not connected to the bin, we need to be able to navigate to the location of that small bin.
“We need to grasp the can, lift the bin, and move the bin over to the big trash bin—and once we get there, if we’re a little robot, we need to be able to lift the bin higher and dump it into our bigger bin.”
Following that, he said, the robot needs to make sure the little bin is empty and return the bin to wherever it originated. Finally, the robot must return home. That alone is about a dozen subtasks, each of which must be programmed into the robot using its various components like sensors and actuators.
“You can see in those subtasks things that include navigation, situational awareness, when the trash needs to be emptied, and so forth,” Dr. Long said.
At this point, the robot designer needs to discover if there’s already a robot built that could be programmed to do this job or if designing a new robot is necessary. That will determine the future of the project.
To that end, Amazon is desperate to prove that Astro is more than “Alexa on Wheels” in order to help drive sales over its far cheaper Echo devices. The customer will ultimately decide Astro’s fate.