Rolling in Comfort
Dave Niewinski has built an AI-controlled La-Z-Boy tank chair that you can drive to pick up a snack without ever having to stand up.
Just about everyone that has ever kicked back in a recliner after a long day has dreamed about being able to drive their chair around so that they do not need to get up to go grab a snack. Picture it: a mobile recliner that transforms your living room into a lazy person's paradise on wheels. No more inconveniences, no more debates about who has to make the snack run. Your recliner becomes the ultimate mobile command center, allowing you to navigate the perilous journey from the TV to the kitchen and back without ever leaving the soothing embrace of your beloved chair.
YouTuber (and suspected couch potato) Dave Niewinski of Dave's Armoury clearly does not like to leave the comfort of his living room, so he decided to bring his living room with him wherever he goes. He created a tank, topped with a La-Z-Boy recliner, end table, and lamp, that can go where no chair has gone before. And to make things as simple and convenient as possible for the — umm, driver… no, passenger… no, I have it now, chairman — Niewinski decided to use AI to control the tank chair. With no safety features of course, because when a tank chair is your ride, you naturally like to live on the edge.
The chair is built on top of an Indro Robotics tracked robot, known as Frank the Tank, that is quite capable of carrying a person, and their La-Z-Boy, up even steep hills and in tough off-road conditions. An NVIDIA Jetson AGX Orin Developer Kit was included to run the AI algorithms with its powerful 2,048-core NVIDIA Ampere architecture GPU that enables the system to perform 275 trillion operations per second. The Jetson controls the tracked robot via a CAN interface.
Niewinski's first experimented with voice control by leveraging NVIDIA’s Riva — a multilingual speech and translation AI software development kit that can provide highly accurate transcriptions of speech. In that way, simple voice commands, like “right,” or “fast” could be transcribed and used to control the actions of the robot. As another potential option, a convolutional neural network, trained with Roboflow, was tested to see how well hand gestures would work as a control mechanism.
These control methods were tested out in a backyard obstacle course built for this purpose. It was found that voice control was quite accurate, and it also proved to get the chair around the course nearly as fast as when using traditional hand controls. Hand gesture recognition did not work nearly as well. There were many cases where the tank chair went far off course, and it was also slow in completing the circuit. No self-respecting chairman would want to be seen going out of control with the hand gesture system, so voice control won the day.
With all of the hard work out of the way, Niewinski kicked back in his recliner and took a spin around town. He even picked up a snack at a local fast food restaurant’s drive-through window without ever having to get up. The workers seemed surprisingly unfazed by the event. Maybe tank chairs are not as uncommon as they seem? Or perhaps they are just accustomed to the antics of Niewinski.
The source code has been released for anyone that is interested in any components of this system.
R&D, creativity, and building the next big thing you never knew you wanted are my specialties.