Trends in lidar and sensors as seen by Lee Seon-yeong, a researcher at the Electronics and Telecommunications Research Institute
“Radar and lidar will be used interchangeably, taking advantage of their respective characteristics.”

Safety is a given in autonomous vehicles. Regarding the direction of radar and lidar to achieve this, researcher Lee Seon-yeong of the Electronics and Telecommunications Research Institute said, “Radar should continue to consider MIMO-based precision improvement and multi-path improvement, while lidar should continue to consider expensive equipment and low resolution.”

Domestic and foreign companies are aiming to commercialize self-driving cars by 2020. According to a market research firm, demand for self-driving cars is expected to increase between 2020 and 2025. Accordingly, while companies are continuously researching and developing, many people and car owners are beginning to feel anxious and skeptical about Tesla’s autopilot and full commercialization of self-driving cars due to the Tesla crash last month and the recent fatal accident involving Uber.

Autopilot is Tesla's autonomous driving system that uses radar, lidar, cameras, and sensors to continuously monitor the vehicle's surroundings and calculate the driving behavior to avoid accidents or collisions. Autopilot is divided into four levels: Level 1, which drives without the assistance of sensors; Level 2, which uses sensors to assist; Level 3, which enables autonomous driving in certain sections; and Level 4, which enables fully autonomous driving.

Whether or not the core technologies that make up autonomous vehicles, such as radar and lidar, can function and detect in any environment is a determining factor in the commercialization of safe autonomous vehicles that are directly related to life and industry.



Radar: MIMO-based precision enhancement and multi-path improvement required
The trend of leading autonomous vehicle companies proves the importance of radar and lidar. Radar is divided into near, medium, and long range according to detection distance. Long range lenses can see far, but have a narrow field of view. As you get closer, the distance you can see decreases, but the field of view widens. If you look at the near range, if the vehicle is traveling at high speed, there is no braking distance, which can create a dangerous situation.

The transmission signal method uses FMCW (Frequency Modulation Continuous Wave radar). It is a sensor that can detect both distance and speed, and it is a method that can determine location information and distance depending on where the received signal is shot by changing the frequency phase. The higher the frequency, the wider the bandwidth, which is advantageous for identifying objects. This increases accuracy and is why we are moving from 24GHz to 77 and 79GHz.

The radar's signal processing component provides information to the user through two signal processing methods: a transmitter that sends radio waves, and an internal signal processing and data tracking of radio signals that bounce off objects and return, to determine the location and speed of the object.

In the past, antennas had multiple receiving and transmitting ports to view a wide band, but the actual function was done on a single channel. Recently, MIMO technology, which is widely used in 4G/5G and Wi-Fi, has been applied to transmit and receive signals, and through signal processing on multiple channels, objects can be received in three dimensions to obtain a lot of information.

Lee Seon-yeong, a researcher at the Electronics and Telecommunications Research Institute, said, “Most companies will develop radars in the direction of increasing precision formation based on MIMO, and the 79GHz frequency band that has not been opened yet will also be opened as high bandwidth is required.” “In addition, it is necessary to stabilize the unstable multi-path situation where signals bounce around guardrails and tunnels, and radar acquires a lot of information, but there is a limit to judging the location of lanes and the distinction between objects, so this needs to be resolved,” he added.



Lidar: When expensive equipment and low resolution continue to be a concern
There are no commercial products for LiDAR because of its high price and it is not applied to vehicles. The signal processing unit is no different between radar and LiDAR, but while radar transmits radio waves, LiDAR transmits 900nm light. This method uses a power amplifier mounted on the front to emit light, convert the received analog signal into a data signal, and then process it into information by computing.

Lidar consists of a cylinder that recognizes objects by shooting light in a radius around it, a channel that shoots a light source and enables the representation of objects as three-dimensional information, a light source that shoots light, a processor such as a power management unit, a DSP, and an MCU, and a sensor that actually processes the data and creates pulses and transmits signals.

It is moving from the 'point 2D scan lidar' that was initially widely used to determine whether there is an object in front or not to '3D scan lidar'. Experts predicted that the demand for vehicle lidar will expand from 2D scan lidar-centered (2.9 million units) in 2016 to 3D-centered (13.2 million units) in 2021.

Basically, Lidar can measure distance and speed, and it is reported that it can distinguish objects by shooting multiple light sources. However, the light source is being developed to shoot light that is not harmful to the eyes because of problems with the optic nerve. Due to price issues, Lidar currently adopts a method of limiting the horizontal field of view to 120-150 degrees instead of 360-degree rotation and reducing the price.

In addition, the lidar manufacturing method is changing from a motor-based mechanical state method to a solid-state method using MEMS technology. The motor-based method has a high probability of failure due to obstacles, terrain, and objects due to shaking. On the other hand, the MEMS method uses semiconductors to make the size smaller and is stable for transmitting and receiving signals because it is not a rotation method. In order to increase the field of view, it is being developed considering mixed use with the solid method and scanning lidar.

Autonomous vehicles are based on ADAS. It is important to miniaturize the system and obtain a fixed cost with clear distance and angle shapes. Existing sensor cameras are weak in daytime, nighttime, rain, and snow. Radar is relatively strong, but has limitations in recognizing objects such as colors, people, and lanes. Lidar emits light, so nighttime is not a problem, but it is weak in rain and snow.

Accordingly, the researcher explained, “Lidar can generate noise, such as when splashed by raindrops. In other words, radar and Lidar each have their own characteristics, so they are likely to be used interchangeably.” He continued, “Startups will have to continuously worry about Lidar’s high price and low resolution.”