Within the next few years self-driving cars will become a reality. There isn't a day that goes by, when you don't hear about new developments in this space. All the major technology companies (and others that are making huge strides in this space), as well as automobile manufacturers are in a race to perfect this technology. The likes of Google, Apple, Uber and now even Samsung, are taking the fight to traditional automotive companies like Ford, Audi, Mercedes, and GM to name a few.
The technology behind this rapidly changing futuristic landscape is something called LiDAR.
What is LiDAR?
Lidar is an acronym that stands for "light detection and ranging". Its operating principle is similar to that of Sonar and Radar - you "shoot" something out into the world, then track how long it takes to come back.
Sonar uses sound for navigating and ranging. Submarines, for example, send out "clicks" into the water and closely monitor the echoes. Knowing how fast sound travels in water and knowing the time it took the echo to come back, one can determine how far an object is in the water.
Radars do the same thing using radio waves for detection and ranging. The giant radar dishes at airports “fire” out radio waves and detect how long it takes for the radio waves to bounce off, say a plane for example. Since the speed at which light travels in the atmosphere is known, the time it takes for signals to bounce back helps determine how far the plane is from the airport.
LiDAR operates on a similar principle using lasers. Lasers are used to send out light pulses, just outside the visible spectrum, and the time it takes for the pulses to reflect after hitting different objects, is detected. This data is recorded into a 3D map that outlines the direction and distance of the various objects from the LiDAR unit.
Why LiDAR and not Sonar or Radar for self-driving Cars?
While the principle used for detection and ranging may be the same for all three technologies, there are certain limitations of using sound or radio waves for self-driving cars. Radio waves propagate long distances and can pass through certain objects. This make them ideal for long range detection however challenging for short range mapping. Sound on the other hand moves relatively slowly and dissipates quickly making its use limited to extremely short-range detection - on the order of a few feet away.
Using light in the visible or near infrared range, has a lot shorter wavelength than radio waves, which alleviates some of the propagation issues through certain objects. It also travels much faster than sound does, making it an ideal comprise for detection and ranging.
Image Source: ESA Science and Technology
Components Of A LiDAR Systems In Self-Driving Cars
LiDAR systems in autonomous vehicles are comprised of lasers, detectors, sensors, optics, navigation systems and 3D modelling software. What make lasers the ideal choice in LiDAR systems, are their fundamental properties. Lasers are:
- Monochromatic - i.e. one colour (one wavelength)
- Collimated - i.e. very small divergence over large distances
- Very intense - i.e. lots of energy in a small area
- Polarized - i.e. energy is aligned in one direction
The LiDAR lasers project millions of pulses per second, generally at 'eye safe' power levels and at wavelengths usually of 532 nm, 1064 nm or 1550 nm. These pulses are projected from a spinning protrusion on the rooftop of a vehicle. As the pulses hit objects and bounce back, they are detected and plotted as a detailed 3D image. This image then enables the vehicle's computer vision brain to decipher the difference between a truck, a bike, a tree, an adult, or a child. The response of the vehicle to stationary objects and the predictability of the movement of moving objects is critical in determining when to accelerate, brake, move over, turn etc.
Challenges Facing Self-Driving Cars
There are a few challenges that LiDAR makers are currently trying to solve ranging from human hand-offs to reliable detection of objects, cyber security to affordability.
There is a fair amount of skepticism around the so-called handoff of vehicle control from machine to a distracted human in rapidly moving vehicles in emergency situations. This has more to do with human nature than with LiDAR technology. Humans in automated settings are easily distracted and the response time to react in emergency situations is of real concern.
Also, subtle signals that we take for granted such as eye contact between a bicyclist and a driver, intuition we use while navigating through faded and unclear lane markings, ability to make left turns in high speed moving traffic, all need to be reliably taught to computer vision systems. This is the reason why most autonomous vehicles today are limited to operate below 25 miles per hour.
Recent news of malicious software attacks highlights another area of growing concerns for self-driving car makers. Since these cars are a collection of networked computers and sensors wirelessly connected to the outside world, cyber security ensuring the safety of passengers and pedestrians is of utmost importance.
LiDAR technology is still quite expensive however there is a concerted effort in this space to make cheaper versions of the technology by expanding the wavelengths used for detection and the development of higher performing sensors
Final Remarks
There is still a lot of work that needs to be done before a mass rollout of self-driving cars but the future is nearly here.
Click here to see some of our products that can be used in a LiDAR application
