10.3.23-Showcase-Osaka-iSee

YIBIAO ZHAO: Thank you. Hi, everyone. This is Yibiao. I'm a CEO co-founder at ISEE. This is my first time in Japan, so I'm very honored and very excited to be here. ISEE is an autonomous driving company. We're focusing on automating logistic yards with autonomous trucks.

As you may know that there are a lot of autonomous driving applications. And then there's a lot of people have different opinions. Some of them have experienced autonomous driving technology in the limited level two, like highway driving application. That's already a reality. But the real application of true level four, meaning that the trucks and cars drive by itself without human, is still quite limited in terms of real-world application.

And why is the reason? So before we start a company, me, I'm working at MIT with my Professor Josh Tenenbaum and my co-founder Chris Baker. So we are trying to find out that answer from the study of humans, meaning the babies, so how a human can learn so quickly from so little data. So here is one example. So here the baby, 18 months old, sitting, standing in the corner without anyone tell him anything.

So he is observing an adult trying to accomplish something. But adult cannot finish his task. The baby is observing there, thinking about something. And later he approached the cabinet, open it for the adult. And look at his eyes, so am I helpful?

So this is an insane capability of humans that we're naturally born to be collaborative, right? So with other humans in the environment. And that is one thing that is missing in today's robots, because robots, they are built to follow programs, right? Even today's machine learning AI, they build on top of hundreds of thousands of data.

But they don't have those basic common sense to really understand human and interact with human in a humanistic way. So at ISEE our goal is to build an advanced humanistic AI to modernize the global supply chain. And we start 2017. We're funded by MIT Engine, which is a venture capital fund initiated by MIT.

And when we started in 2017, we were looking at where the application for autonomous driving could be a reality and fast to the market. So and then we find this application through talking with our customers. The logistic yards are a very unique niche. It's a private road. There's no public road regulation.

There's no insurance, and all that challenges in public road that are out of our control. It's also a fenced area. There's no random people, no kids, right? It's a very well-defined environment. But there's a great shortage of qualified drivers, because in the yard there's a lot of traffic, very sometimes even congested. So it's not an easy problem.

So this is a video, actually, when I first time visited our customer site, I took this video. Here comes the intersection. There's multiple trucks come try to go through. And then that is not an easy problem to solve, right? Who should go first, and apparently the multiple drivers, they can solve this puzzle, just like the 18-months old, right?

So when we come to some scenarios they figure out how to interact with each other. But that's not trivial for robots. And also, we can notice there's no traffic light. There's no lane lines to regularize the traffic. It's a completely unstructured environment.

So at ISEE start from day one, we try to build AI that can learn to negotiate with other traffic participants, without rules. So on the left, you can see that's a little animation, but that's how our system sees the world. So you can see the truck and trailers.

Those are the objects moving around the environment. And then those orange yellow things are the trailer, park the trailers. So each vehicle has its own goal in trying to navigate in this space. It's a very interesting scenario interaction here that if I play-- so this car, this little car, is trying to approach intersection. And there's another big guy coming.

The system has learned to make the space for the big guy, just like the 18-months old, and then to proceed, right? All those things have no rules. It's automatically learned through a lot of learning and interaction with the other agents in the environment, fully autonomously.

And when we apply it to the real world, that's the next video. You can see when the two ISEE truck come to the intersection, one aisle is wider versus another aisle is narrower. The system learned to negotiate appropriately, which the one vehicle makes some room for another one.

And those interactions are critical. Otherwise, the robot can get stuck, right? In those narrow areas. So and that's just the tip of iceberg, that a lot of innovation that we're doing at ISEE. And some of our work has published in the top AI Computer Vision robotics revenues.

And when we put everything together, this is how it works. So here is our autonomous trucks that are operating in one customer site. So what do you find interesting and unique about this video, is we ask the driver to step out of the cab.

So it's truly driverless. There's no one in the cab. And the vehicle is not just driving, it's also accomplishing all the tasks needed to make a fully autonomous end-to-end job. For example, right, it's trying to find a trailer, raise the trailer boom, connect the air lines between tractor and trailer so that it can release the trailer airbrake, so that it can pull the trailer out of the spot, navigate to the destination, through the aisles into the intersections.

There are also other human drivers in the environment. So our customer has compared our system with other players. And apparently our system demonstrated superior efficiency, safety, as well as intelligence to interact with other human traffic participants.

And when it comes to the dock door, and our system is able to very accurately, precisely back into the dock door, in one shot, without any back-and-forth adjustment needed. So that's often better than what a human driver can do, because human driver usually have to pull out and make a few adjustments. And that save the time and also reduce the safety incidents.

After we park we decoupled air lines and we finished the move. So this is not just demos. We have operated in a customer site for more than 10,000 moves. We have multiple trucks in customer location, multiple location, operating eight hours, full shift.

And today we have worked with quite some of our customers-- trying to go back-- and we have customers who are Asian-based, the Japanese OEMs, and they have large manufacturing centers in US. And we help them to build autonomous driving solution within their manufacturing plants. So 50 autonomous vehicles moving loads between their different facilities in their manufacturing plant.

And we also have customers span all the different categories, including warehousing, distribution centers, logistic, freight and port, and transportation. So basically any Fortune 100 company or 500 company, or any company need to move a lot of things, a lot of goods, and we help them to modernize today their supply chain.

So that's what we do and we are looking for partners who we can work with to bring their operation to the next level. That's it. Thank you.