S5, E1: Tell it like it is | Autonomous Vehicles
This season we focus on stories of innovation and kick off with an exciting conversation with Saurabh Chandra, co-founder of Ati Motors, one of the coolest startups in India that makes autonomous cargo vehicles based in Bangalore. Cool for many reasons, not the least for an unusual team of co-founders, the youngest of whom was barely a teen when he joined. Club that with using tech like AI/ML, power electronics, control systems, mechanical engineering, system software and electronic hardware, and it makes the whole enterprise unique.
As you will see over our conversation, a strong purpose, and relentless focus on solving for problems a.k.a performance in the Indian context are two key themes that emerge.
Some excerpts from our conversation:
MK: Right now, all of us are extremely familiar with home robots, at least the cleaning variety, right. The technology that they typically use is that they scan your entire premises, they do a couple of rounds and then over a period of time, they understand where there is a slight dip or where they need to dock or where there is some sort of obstruction. So, do your vehicles work in a similar fashion? How different are the vehicles that you develop to home robots?
SC: So, in terms of technologies that are there in the vehicle, they are remarkably similar. Ok?
So, every autonomous bot must ask three questions — Where am I? What is around me? And what is my next step? You ask these three questions in a loop, you’ll get to your destination.
And this is it, ok?
At a very high level, this is the algorithm of autonomy. Now it doesn’t matter if it is Google making Veemo, or it would be a Roomba bot, or it’d be an Ati Sherpa, all of them at some level are asking these questions. And the technical term for these would be that you have a localization algorithm, perception, obstacle detection, path planning blah blah blah…whatever but in broad terms these are the questions that all these bots or these cars etc are trying to ask. And they will do this with different kinds of sensors. Right, so you may figure out where you are based on variety of different inputs. And a home robot may have different sensors compared to us even outdoors or in industrial environments that are not only the shopfloor, we can even do building to building. That is something that differentiates us, I mean, we are unique in our category where we can go from the inside to the outside of the building without skipping a beat. So, the vacuum cleaner was basically a fantastic category which first went into this autonomy in home automation. Because with vacuum cleaners what happens is you can be 80% as good and you have still done something useful. And for a long time if you see, this was the only category. There are actually not many use cases of this type. Ok? Very late actually Yankee Robotics tried to make a toy because a toy is also similar, it did not really succeed in their mission, but they started on this premise — what can autonomy do which is good enough even if it is 80% good. And that becomes a very different ball game for eg., our case we must be right 100% of the time. You cannot say that I will deliver stuff 80% of the time or even 99% of the time. You have to be accurate all the time so you have much higher precision requirement. We have precision requirement from word — go. You do not have any step for exploration. So, home robotics is different in that sense because they can explore and its consumer grade, right so, you allow it time to explore and after that it cleans your house…I mean, these days they have become very good. I mean, when they started they were some 70–80%, I would say, accurate. Now I would say, they are 90–95% accurate. And that is great actually. For a home vacuum cleaner, that is a good enough percentage of accuracy.
Listen to the full episode:
Read a transcript of this episode here:
