DRIVR Part 5: Aslantis

A lot of you have been asking me questions about the wisdom in relying so heavily on a wireless network for offloading the vehicle computation. No, I'm kidding, nobody is asking questions. There are only like 7 people who read these, and 6 of them are friends doing it as a personal favor (WHICH I APPRECIATE). But it's a reasonable contention to raise. If we have bet the farm on ubiquitous 5G/6G (which currently doesn't exist), then aren't we stuck with nothing in the meantime? First of all, speaking of farms, how good is cheese?

Okay, back to business. The big change we have made is to split the onboard computation into two (or more) pieces, allowing for the hardware to live in multiple places. The goal is that the majority of the computation occurs off the vehicle, which would require a high-speed data connection. However, we can achieve the same basic architecture with all the computers located on the vehicle, connected via Ethernet, or with the heavy-lifting occurring off the vehicle but on the same high-speed local network. This would be our best-case scenario in terms of performance. In other words, if it doesn't work like that, there's no need to a pursue a wireless solution. We will carry this step-by-step philosophy throughout our development, progressing slowly from one technical milestone to the next, making it the second greatest step-by-step-related creation of all time. Absolute legend.

Anyway, back to the whole purpose of this post. Let's start talking about the advantages of pulling the computation off the vehicle. If our big brain lives in one place, then all of our vehicles get to share that single brain. You know how your partner has memorized all the athletes whose names contain animals, so you don't have to? (I'm looking at you, Kevin Duckworth) Well it's like that, but for useful stuff, such as the location of other vehicles and pedestrians. Because as much as we'd love our cars to pick themselves up by their bootstraps, maybe if they aggregated their resources (in math we call that a union), they could achieve their goals more effectively. After all, why should we make every vehicle solve the same problems time after time (cars just want to have fun). With our unified consciousness we can create a singular model of the world (let's call it a mirrorworld), into which we can place every piece of the real world that our cars perceive (or that we know ahead of time, like where the roads are). So we'll have a single world, in which everything exists in a virtual fashion...wait, that's like a video game! Not just any video game, but an MMO that finally forces you to leave the house.

Of course, perception is hard, so we won't start with that. We can populate the mirrorworld with purely virtual objects and develop our decision making algorithms

based on them. Remember, everything ends up in the mirrorworld (let's just call it Carnia from now on), so it doesn't matter how it gets there. According to our big brain, there is no reality, only ̶Z̶u̶u̶l̶ Carnia.

Okay, we'll leave it here for now. There are more advantages to our big brain approach, but those will have to wait for next time. Don't forget, I'm looking for collaborators! If you like developing games, autonomy, or social awkwardness, find me here!

-Greg

The story continues with Part 6: The future ain't what it used to be