Autonomous cars need to drive well and look good
Searching for Sleek
Once while laying out a particularly messy test circuit one of my coworkers told me, “it doesn’t matter how it looks as long as it works.” In fewer words, form follows function. Car manufacturers, though, may not agree. Most people don’t want to be seen driving around an “uncool” car. While we may disagree, the average person would categorize a large LIDAR array on top of their car as “uncool.” Shrinking LIDAR technology can give autonomous vehicles a sleeker look while increasing the sensor’s potential.
Most LIDAR industry leaders are now pushing to shrink their technology. A research group at MIT is working on developing a LIDAR microchip, with no moving parts. This chip is being designed for a 300 mm wafer, now that’s small. With sensors that small, you could mount them on any part of the car. Using Multiple small sensors would be especially useful for autonomous freight vehicles. It’s difficult to use a single sensor on an eighteen wheeler because of its large size. With sensors this small, you could mount them around the cab and all down the trailer. Small multi-sensor arrays could revolutionize the shipping industry.
Velodyne, who also manufacturers LIDAR sensors, is also in the race. They are attempting to shrink their technology down to the size of a hockey puck. A single point sensor array like theirs is better suited for personal autonomous vehicles. In fact, one of their LIDAR arrays is currently in use on Google’s Waymo cars. While I would not describe Waymo as “sleek” it is cool that there are autonomous vehicles being produced right now. Downsizing that dome on the top to the size of a hockey puck would certainly up the cool factor.
The Fast and the Futuristic
Just like most people want a “cool” car, many also want a fast car. Most wouldn’t stop to think about how fast their LIDAR array is, though. You know that current LIDAR arrays have to spin 360 degrees to get a full picture of their surroundings. Some LIDAR arrays are already spinning faster than helicopter blades. As the maximum mechanical limit for RPM is fast, several developers are investing in faster, next gen scanning methods.
The next best thing after mechanical scanning is electrical scanning. Several industry leaders are now venturing into this area. The MIT research group, Velodyne, and Quanergy, to name a few, are all producing “solid state” sensors. These electronic scanning sensors can scan their environment hundreds of times faster than a sensor using a mechanical method. Faster scanning means you get more data points to work with.
One thing autonomous vehicles have been chasing for some time is a low price point to go along with sleek and fast scanning methods. The required sensors exist, but also cost tens of thousands of dollars. If your LIDAR array alone costs $50,000, you’re going to end up with a car that no one can purchase. These same researchers that are trying to make LIDAR smaller and faster, are also trying to make it less expensive.
Remember that MIT group trying to make LIDAR sensors on microchips? One of the biggest advantages of LIDAR on chips would be their low cost. As you know, microchips are made in foundries that pump out millions of them. If LIDAR sensors could be made in microchip form, they would be produced in high volume, and sold for next to nothing. That dream of mounting multiple LIDAR sensors to large vehicles would be a lot closer.
The other groups mentioned are also trying to lower the price for their sensors. Quanergy is supposedly aiming for a $250 price point. The future of driving is fully autonomous. Every car will need a full 360 degree LIDAR array, and they will need them at a tiny fraction of their current cost. The sooner LIDAR manufacturers can pump out low cost LIDAR sensors, the sooner we’ll all be riding around feeling like Tom Cruise.
No matter what kind of LIDAR array you’re using, you’ll need to write software for it that can process its data quickly and accurately. Libraries, like LAPACK (available from TASKING), contain the complex linear algebra functions you’ll need to process signals from LIDAR arrays. Utilizing LAPACK’s functions will allow you to focus on what really matters, writing excellent software that will make those sleek cars in Minority Report a reality.
Have more questions about the future of autonomous vehicle sensors? Call an expert at TASKING.