DaxOS
How It Works
DaxOS runs on various chips which cover different “motivations” or functions. Traditionally, robots run on top-down operating systems that rely on complex algorithms to complete definable tasks. This system allows them to prioritize and complete tasks in uncontrolled environments.
Imagine that we task Dax with a delivery: his pathing chip proposes a motion that Dax goes on the delivery—and all of Dax’s chips vote on it. Since this motion finds no resistance, Dax’s pathing chip will send him off in the right direction at the right speed.
Now, imagine that Dax is rolling down the street. Someone jumps out in front of him: His collision avoidance chip submits a motion that he stops (a vote that it will win), overriding the pathing chip that wants to continue driving.
Dax’s Cortex (a “brain” that tallies votes) declares the vote in favor of stopping, and Dax comes to a halt.
“votes are weighted by the importance of the chip voting and balanced by the confidence that that chip has in its vote”
Votes are weighted by the importance of the chip voting and balanced by the chip's confidence in its stance. Some members (like collision avoidance) have more important votes than others.
At this point in our scenario, an operator might take over and drive around the human (or animal) in Dax's path or wait for the person to move.
We recently gave Dax an interactive chip, which operates when he's not on delivery. In the future, he'll have to complete tasks while interacting with people.
When Dax has a directive to make people happy while completing a task, he will have to choose between the two.
If someone greets him, his interaction chip might propose that he stop, tilt his head, and make a happy noise, but if he's running late, his pathing chip might say he needs to keep going. The pathing chip will win in this case, and Dax might nod instead of stopping.
Suppose Dax has extra time for his task, and someone tries to interact with him. Because he has spare time, his pathing chip is less adamant that he can’t stop. His interaction chip wins the vote, causing Dax to pause and interact for a minute before going on.
There are many directions that we could go with DaxOS: Imagine, for instance, creating a video game where the NHPs (non-human players) ran on DaxOS with chips representing curiosity, affection, and loneliness, and each chip was weighted just a little differently for each NHP— this would create “real” characters, which someone could even set to filling a personal agenda.
The Future
As exciting as it is to imagine all the new avenues we could explore with DaxOS, we would like to see DaxOS used for absolutely everything robotics-related.
We don’t yet know all of the uses that DaxOS could have. Jean-Baptiste Joseph Fourier invented the Fourier Transform to model heat transfers. Now, we use it for audio sampling and digital signal processing. While it will always make sense for simple tasks to run on top-down algorithms, providing new tools to the tech community is what we’re passionate about.
While protecting the interests of shareholders, we would love to provide tools for more than just Dax—tools that revolutionize what we currently think of as possible and build into the synergy of the creative community.