Self-driving cars are the future of transportation, and it is something most companies like Tesla, Toyota, Waymo, Uber are working on diligently. Almost all of them believe that building a self-driving car is not possible without LiDAR. However, Tesla and its CEO Elon Musk care to disagree. As a matter of fact, Musk feels as if such technology has no future in this niche, which he stated at Tesla's Autonomy Day. Discover more about both ideas and mainly why has Tesla decided to go down the path of the non-LiDAR approach to autonomous vehicles.
LiDAR Basics
The majority of companies interested in building autonomous vehicles believe that LiDAR is a crucial technology for succeeding. LiDAR is a measuring technology that could map out the surroundings of a car, measure objects, and the distance, and turn that into a 3D representation. Using a laser helps the system measure the distance precisely by calculating it via the speed of light. GM Cruise, Ford, Waymo, as well as many others are supporters of this kind of object detection technology in autonomous cars. This might lead to innumerous rideshare options such as rideshare delivery or even split rideshare in self-driving cars, similar to Waymo.
Tesla’s Computer Vision
Tesla decided to go with a different approach and incorporate computer vision into its projects. In other words, the company opted for AI technology that tries to figure out the visual world via a set of stereo cameras. But that's not all that Tesla will rely on when object detection is concerned. Apart from cameras used for depth perception, they will use GPS, radar, sensors as well as maps.
Reasons Behind Tesla's Non-LiDAR Approach
The autonomous vehicles manufacturing industry is definitely a hot topic these days. Will LiDAR be the future of it? Why is Tesla, of all vehicle manufacturing companies out there, the one against it? And is the pro-LiDAR team right for trusting the system only due to its incredible accuracy? Let’s take a closer look at it.!
1. LiDAR Is Too Expensive
One of the reasons behind Tesla’s non-LiDAR concept is definitely the cost. Namely, a LiDAR system price per unit could reach up to $10,000! This automatically puts such a self-driving vehicle out of reach for, well, the majority of people. Surely, manufacturing a larger number of vehicles might mean lower costs, but the price would still be ridiculously high. One of the main goals of Tesla is to make its vehicles affordable to as many people as possible.
2. Dangers on the Roads
Another reason is also the inability to apply LiDAR to all the situations in traffic. For example, this system has an impeccable ability to measure the distance, and possibly help a vehicle with navigation. However, not all the objects on the road are stationary objects, and LiDAR is having difficulty detecting moving objects. To be precise, it is not able to detect how the objects are moving, the degree of danger they pose, etc. Thus, such a system won't make a big difference between a real human being crossing the road or a plastic bag, which might cause serious issues.
3. Other Systems Can Be as Precise
One of the main benefits of LiDAR is its accuracy. However, researchers at Cornell University claim that using less expensive equipment such as two cameras, has proven quite effective. As a matter of fact, the 3D mapping done via two cameras positioned on two windshields is as precise as LiDAR. What they discovered is that if they switched the perspective from a frontal view to a bird’s-eye view the system was more accurate. Once again, the price of such a system would be way lower. This is precisely why Elon Musk stated that such a system is not something that will stick around for long.
4. The Ability to Adapt
According to Musk’s commentary concerning the anti-LiDAR approach at Autonomy Day, Tesla’s system is adaptable. Supposedly, their convolutional neural network helps the system take advantage of all the data to decide “rationally”. Vehicle's using LiDAR won’t be able to adapt in such a way. The majority of these vehicles were never tested in real-life situations and on real roads. Moreover, their navigation as well as decision making relies on high-accuracy maps.
The problem arises in situations where driving conditions become unpredictable. Most of these vehicles are tested in perfect conditions and with highly efficient maps - not during rainy weather, storms, poor lighting, etc. Even if Waymo with its LiDAR-based autonomous driving technology is driving in cities, it mainly stays on public roads using precise maps. So, when the vehicle suddenly goes onto a smaller road without precise information, there might be some setbacks.
5. Experience
Tesla cars have been on the market for quite some time, and people all around the world had the opportunity to use them and drive for billions of miles. Vehicles such as Waymo aren’t as nearly tested as Tesla is - this means more road data input, more unpredictable situations, more experience. Such gained data is priceless and it is the reason behind its ability to adapt and upgrade. In addition, all the accumulated information will definitely be useful when designing a completely autonomous car able to perform perfectly.
In a Nutshell
The self-driving vehicle industry is something that gained a lot of attention recently. As it appears, Tesla is aiming for the more affordable option for everybody without LiDAR tech in its plans for the future. Unless things take a different turn, vehicles such as Uber Car or Waymo will be out of the ordinary person’s reach. To sum up, with all the new systems and technologies being invented, we might say that we could see them on a regular basis soon. The question that remains is who will do it better? Will there be several different systems for navigating autonomous cars, or could it all come down to the same principle?
