As technology makes our lives more convenient, I am urged to ask more questions about safety. Are we living a safer life today? I have heard horrific stories about safety hazards brought about by so-called ‘smart’ features, oftentimes via automated functionality.
It is undeniable, particularly within the automotive space, that a desire for convenience is a rising trend. People are willing to tradeoff something for convenience, even if that means a higher car price, or fewer designs to choose from. As a VW Beetle collector, I know if I want something vintage-looking, I cannot expect to get Level 2+ ADAS. But should one trade off safety? Never.
Convenience is only a valid concept when it’s enjoyed without safety concerns. As I noted in my previous Driving Lidar post, if an automated driving system is not safer than a human operator, how can a driver really sit back and relax knowing they might need to take over any moment?
That makes it all the more important to provide mass-market access to solutions that combine both safety and autonomy – features that make driving less tiring while reducing traffic collisions, pedestrian harm and the financial toll of accidents.
A few companies – the most popular being Tesla –have invested their future advancements in camera-based computer vision to address common driving scenarios and enhance safety. Elon Musk did not think lidar was needed, saying in 2019 that lidar was both unnecessary and expensive.
As a lidar company CEO it is no surprise that I am constantly asked to comment on Musk's view on lidar. Instead of treating it as an attack, I see it as an opportunity to clear up the common misconceptions around lidar technology.
So – is lidar necessary? Yes.
The certainty lidar brings to vehicle perception is unparalleled. Lidar sees things in 3D, while camera does in 2D, lacking the measurement of depth. With that, lidar does more than just telling you how faraway an object is.
Let’s say there is a dog in the street. A camera will capture the dog’s color, but it can’t tell you if it’s a small dog right in front of you, or a big dog 30 ft. away. With the help of artificial intelligence, it can give you size and distance information, but it can be inaccurate due to inaccurate or inadequate machine learning. Musk shared himself that Autopilot once mistook the stop sign graphic on his T-shirt as a real stop sign. In theory, anyone standing on a sidewalk could recreate the same action.
This happens a lot with computer vision today, because algorithms are trained to focus on certain elements of a graphic, which can be easily tricked.
There are things that AI is not trained well enough to accurately recognize. They are probably less common in our lives, but are they less dangerous? Not necessarily. Missing those things can lead to critical safety loopholes.
Lidar data contains a higher level of precision and intelligence, and focuses on things that matter. Let’s talk about the dog again. Lidar will tell you how big this dog is, how far away it is and how fast it is running. It can even tell you if the dog is accelerating or slowing down. This 3D imaging capability is key to real-time vehicle perception.
Another key advantage of lidar is that it “sees” in the dark, which I’m sure you’ve heard of. But lidar addresses way more challenges associated with lighting than what some might consider to be an easy fix with headlamps and streetlamps.
Lidar is agnostic to environmental lighting, but a camera can be blinded by strong light – light from other cars, or sunlight. When the car gets into and out of a tunnel, the camera on the car needs a few seconds to adapt to the abrupt changes in lighting, much like our eyes do. For an automated system, this is problematic, leaving areas pitch dark or over exposed to the point of obscurity – which can then affect the vehicle’s decision-making processes. In a similar scenario, lidar’s performance is completely unaffected.
Lidar uses photons to “touch” everything, making the environment more palpable. It is absolutely essential to safe autonomy. It’s why I’m so passionate about bringing this to motoring at scale.
Of course, there is a new cost to vehicles in installing lidar. Mr. Musk has also commented on the cost of lidar in the past, so join me next time as I explore the question – is lidar too expensive?