We all love LiDAR: it measures precisely, it saves time, it saves money, and it produces beautiful maps. But will it pilot your first self-driving car? Some recent news indicates that the future isn’t so certain.
Hard Numbers
London-based research firm Frost & Sullivan have released a report on LiDAR manufacturing for automated vehicles. They found that 7 out of 13 of the top OEMs (original equipment manufacturers) are working on an automated vehicle that incorporates LiDAR. They also mention LiDAR-enabled cars have (somehow) already raked in $51.1 million in 2014, and those numbers are expected to hit $141 million by 2021.
Surprisingly, they say that only 29 percent of those LiDAR units are going to be used for “fully automated driving purposes.” That means more than 2/3 of all LiDAR units in automated vehicles are going to be used for “active safety functions,” like help with object avoidance and driving in low visibility.
Why is that? Frost & Sullivan explain that even though LiDAR is still the best sensor for object detection and mapping in low-visibility conditions, it costs too much. Affordable LiDAR sensors are available, but you’ll have to pony up a lot of money to get one with the proper resolution and range for self-driving applications.
While LiDAR might have developed to maturity for static scanning, mobile mapping, and airborne applications, it still has a long way to go before it’s cost-effective for automated driving. My guess is that cars including LiDAR for self-driving purposes will be prohibitively expensive for some time.
Elon Musk Speaks
A second bit of news bears this out. PR lightning rod and CEO of Tesla, Elon Musk, recently held a press conference to describe the car company’s new Autopilot feature, which uses GPS for active safety features.
In the Q&A session, a reporter asked Musk about Tesla’s plans to give their cars more automation. The big takeaway from his answer is that we may need more hardware than Teslas currently include in their sensor suite, but we almost certainly don’t need any LiDAR.
“For full autonomy,” Musk said, “you’d really want to have a more comprehensive sensor suite and computer systems that are fail proof. That said, I don’t think you need LIDAR. I think you can do this all with passive optical and then with maybe one forward RADAR… if you are driving fast into rain or snow or dust. I think that completely solves it without the use of LIDAR. I’m not a big fan of LIDAR, I don’t think it makes sense in this context.”
Musk went on to explain that he doesn’t dislike LiDAR in general—it goes in his Dragon spacecraft for docking with the space station—but it doesn’t make sense in a car context. “I think it’s unnecessary,” he said.
Why are they unnecessary? As the Frost & Sullivan report states, they’re just too expensive, and possibly even more powerful than a self-driving car needs. Why use a flamethrower when a simple match will do the work for less money and trouble?
Google’s adorable self-driving car, with LiDAR in the sensor payload on top.
Why Google IS Using LiDAR (A Few Speculations)
You might have noticed that Google is still using LiDAR. There are a number of reasons why the company would still do this, even though the sensors are expensive. Here are a few speculations.
First, if Google is treating their self-driving cars the same way they’re treating their Nexus phones, they’re not looking to develop a product for mass appeal. They’re looking for power users who are willing to pay the price for the newest technology.
Furthermore, Google is in the business of gathering and selling data. We know that LiDAR is great for gathering road asset data (and *might* be OK at gathering data on road conditions). If Google has some thousands of people out driving their LiDAR-equipped cars, that means they’re gathering lots of data on lots of roads without having to do the work themselves. Google could then turn around and sell the information to anyone who needs it.
All that being said, there may be a point where LiDAR is cheap enough to be useful for fully automated driving purposes. But it’s looking like that point won’t come for more than 5 years, well past the 2021 benchmark that Frost & Sullivan have identified.