The Everyday Robot — A robot for anytime, anyplace
I’m not sure about you guys, but I’m not a huge fan of chores. Especially with quarantine forcing everyone indoors, everyone is doing something and leaving a mess somewhere. Every time I clean up the area, someone ends up making a mess, and I have to wipe down every surface again, and again, and AGAIN. It gets so annoying sometimes because I can’t work in a messy area. Every day I say “Ugh, I wish there was a robot that could do my chores for me.”
But how could one make a robot that can do everything without it being so expensive and complicated to create? That’s where the Everyday Robot comes in handy.
The Everyday Robot Project is an X Company project that is creating an affordable robot that can be easily taught by anyone to do any task. You need a coffee but you are in a middle of a meeting? Everyday Robot has got your back.
But how?
Building a robot like this is not easy. Current robots are built to do specific, repeated tasks, but the Everyday Robot Project is attempting to create a robot that can do many different tasks, and constantly learn new ones, while also being exposed to different environments such as a home or an office. Plus, the robot has to be able to easily adapt and interact with the people and objects around it. They have to create a robot that is capable in adapting to the unpredictable and unstructured life we all live.
Perception
To solve this, the Everyday Robot uses various cameras and advanced machine learning to create a robot that can learn faster than most robots we have today. They do this by running simulations in the cloud, shared experiences of other robots and AI, and from human demonstrations, instead of straight on coding them. For the machine learning, the robots use a combination of simulation, reinforcement learning, and collaborative learning. Every night, thousands simulation robots are put to the test, practicing all sort of tests until they are efficiently able to complete the tasks without failure. The virtual training is then moved to the real robot to refine the new skill. The data from the real-world training is integrated back into the simulations and shared with the rest of the robots.
Manipulation
In order to manipulate objects, the robots use an arm with 5 joints to have varying degrees of motion. This way the robots have many different ways to interact with the objects and environment, allowing for experimentation and practice.
Navigation
The final component of the robot is navigation. To find the most efficient path, the robot uses varying sensors, from video cameras to microphones to detect humans and obstacles in the vicinity. It’s trained through simulations to learn how to avoid collisions in unpredictable environments, a feat rarely accomplished in the world of robotics.
Progress
After sometime in testing, the robots have reduced the amount of trash displacement from 20% to less than 4%. Now it’s not exaaactly a huge problem solver, BUT this is a great test to test the robots’ learning abilities, making it ready for more complex tasks. There is a lot of promise coming from the Everyday Robot Project, and I am stoked to see what they have in store!
Pssstt…
Hey there! If this is your first time reading my articles, my name is Emina Awan and I’m a 15 y/o innovator working on cool projects in robotics, cars, AI, and much more!
If you want to see more of my work, go join my monthly newsletter, watch my Youtube videos, follow me on Medium, and connect with me on LinkedIn!