Cadillac is doing something new with LiDAR. Instead of sticking a puck on its cars, it's using the sensors to map the highways of the United States and Canada and geofence its semi-autonomous Super Cruise feature, instead of letting drivers use it anywhere they want. It seems like a bold move, but in reality, it's how self-driving cars will initially enter the market.
The realization of a fully autonomous-car future rests on regulations, sensors, high-powered computers and maps. While safety is the basis for most vehicle laws, it's becoming increasingly important for there to be blanket federal regulations concerning these vehicles (instead of the current patchwork of laws that vary from state to state).
Sensors are increasing in accuracy while dropping in price and the silicon that crunches all that data is getting smaller and more powerful. In the middle of all that are the maps that are essential for actual navigation, both around town and within lanes. So while Cadillac is the first automaker to use high-definition LiDAR maps, it certainly won't be the last.
Nearly every car with some sort of semi-autonomous feature is using a map to help keep it on the road. Tesla has its community-sourced information, while Mercedes-Benz is using Here maps to help determine the appropriate speed to take an upcoming corner.
Yet, it'll be highly detailed maps like the one used by Cadillac that'll make robotaxis a reality. The LiDAR map used by Super Cruise and stored in the trunk of the CT6 is accurate within 10 centimeters. It sees 2,500 meters ahead of the car based on its trajectory and shares information with the computer that determines if the semi-autonomous feature should continue or if the driver should take over.
When Uber and Lyft finally drop their truly self-driving taxis onto the streets or cities, or unveil their fleets of autonomous buses, they'll need to be able to access a locally stored map like the CT6 does to navigate the complex roads of a metropolitan area.
The vehicles will need that data to be stored on the vehicle even if it is updated daily, hourly or even every few seconds. That's because passengers won't be happy if their ride to work suddenly stops working because it lost its network connection.
Even with all the data and sensors that'll be available on the first autonomous vehicles, they'll most certainly be geofenced. The cars travel along predetermined "routes" to reduce the chance of an incident. Maybe they'll deviate slightly to pick up passengers, but if you're way off the "grid" there's a chance you're going to need to walk to grab your first-generation robotaxi. The technology for self-driving ride-hailing services is something that's already being developed, at least at the chip level.
During the announcement of the new NVIDIA Drive Pegasus AI computer for Level 5 vehicles, the company's senior director of automotive, Danny Shapiro, said that initially the cars will be geofenced. "That's part of the evolution of these taxis." But it won't be that way for long.
In the future, thanks to techniques like Cadillac's, cars will learn as they travel along their routes. The sensors will take in more data, feed it through even more powerful computers, and the level of detail added to the maps will grow exponentially. Plus, if more manufacturers adopt this technology, all our cars will start to figure out how to make smarter decisions based on the situations they've encountered on the road and during simulations.
But for now we have the Cadillac CT6, a luxury sedan that's using LiDAR in a new way that'll help bring autonomous cars to all the roads, not just highways. Even if you'll need to walk a bit to grab a robotaxi in the first few years.