Autonomous Anatomy: LiDAR

The first piece of the autonomous vehicle puzzle is LiDAR which has been in debated over how necessary it even is given the power of current software. We’ll discuss what LiDAR is, a simple overview of how it works and if it’s even needed anymore in autonomous vehicles.

What’s LiDAR?

LiDAR (Light Detection and Ranging) help provide the “eyes” to the self driving car using Light detection to determine the car’s surroundings. If you haven’t heard of LiDAR, you likely have heard of Sodar(Sounds Waves) or Radar(Radio Waves) which accomplish similar tasks using different means.

How Does LiDAR in autonomous vehicles work?

As we learned earlier in the article, LiDAR uses light, usually an infrared light, to measure the surroundings. Initially, a pulse of light is sent towards a targeted direction and when the pulse hits an object, the reflection is then sent back to the device’s sensor. The time is takes to return the reflection back to the censor will determines the distance between the object and the vehicle. With autonomous vehicles, there are many LiDAR pulses being sent off every second to survey the surrounding area. The returned information is stored as a data points(Cartesian coordinates) within a point cloud. The point cloud is where these data points are then plotted to perceive the surrounding area and “reconstruct” the data to recreate the surroundings. Here is where these surrounding objects can be classified as human, car, tree, driving lane etc.

Is it necessary to have LiDAR in autonomous vehicles?

Some would argue no, especially Elon Musk who said LiDAR for autonomous vehicles is “freakin’ stupid”. While developing autonomous vehicles, engineers are looking to make the vehicle drive similar to a human, except much smarter. The group that argues against LiDAR is typically in favor of using computer vision being that humans use vision to determine what an object is. LiDAR can only give us so much information attempting to classify these surrounding objects. Computer vision on the other hand will likely be much better at determining not only if there are objects in the area but what those objects are. For example, say there is some minor obstruction falling in front of the car; nothing worth stopping for. Using LiDAR may cause the vehicle to stop suddenly as it’s perceived to be dangerous to the vehicle and passengers. Computer Vision will be able to classify this object immediately and realize that the vehicle can continue driving as there is no threat. Those in favor of using LiDAR argue that while computer vision can classify objects better, LiDAR provides much better depth-perception; after all it is measure distance. There’s a lot to consider. Most autonomous vehicle companies still use LiDAR, however, the debate continues. What do you think?

Follow me on LinkedIn: https://www.linkedin.com/in/mihir-bommakanti/