“As the core sensor of many smart devices, lidar has been widely used. Nowadays, we can see LiDAR frequently in driverless cars, service robots, AGV forklifts, intelligent road administration and transportation and automated production lines, which is enough to show its indispensable position in the artificial intelligence industry chain.
As the core sensor of many smart devices, lidar has been widely used. Nowadays, we can see LiDAR frequently in driverless cars, service robots, AGV forklifts, intelligent road administration and transportation and automated production lines, which is enough to show its indispensable position in the artificial intelligence industry chain.
As far as the mainstream lidar products on the market are concerned, the radars used for environmental detection and map construction can be roughly divided into two categories according to the technical route. One is TOF (Time of Flight, time-of-flight) radar, and the other is It is a triangular ranging radar. I believe many people are familiar with these two terms, but to talk about which of these two schemes are better or worse in terms of principle, performance, cost, application, and what is the reason behind, maybe everyone still has more or less Doubtful. Today, the editor will throw a few bricks and attract jade, and make an analysis on these issues.
The principle of trigonometry is shown in the figure below. The laser emits laser light. After irradiating the object, the reflected light is received by the linear CCD. Since the laser and the detector are separated by a certain distance, according to the optical path, objects with different distances will be imaged on the CCD. on different locations. According to the trigonometric formula, the distance of the measured object can be deduced.
Just looking at the principle, doesn’t it feel very simple.
Figure 1. Principle of triangulation ranging
However, the principle of TOF is simpler. As shown in Figure 2, the laser emits a laser pulse, and the timer records the outgoing time, the returning light is received by the receiver, and the timer records the returning time. The “time of flight” of light is obtained by subtracting the two times, and the speed of light is constant, so it is easy to calculate the distance once the speed and time are known.
Figure 2. TOF ranging principle
Sadly, the world would be a great place if everything was as easy to do as it was thought. These two solutions will have their own challenges in specific implementation, but in comparison, TOF has obviously many more difficulties to overcome.
The implementation difficulties of TOF radar mainly lie in:
1. The first is timing. In the TOF scheme, the distance measurement is dependent on the time measurement. But the speed of light is too fast, so to obtain accurate distances, the requirements of the timing system become very high. One data is that the lidar needs to measure a distance of 1cm, and the corresponding time span is about 65ps. Students who are a little familiar with electrical characteristics should know what this means for the circuit system.
2. The second is the processing of the pulse signal. There are two parts here:
a) One is laser: There is almost no requirement for the laser drive in the triangular radar, because the measurement depends on the position of the laser echo, so only one continuous light output is required. However, TOF is not enough. Not only does it need pulsed lasers, but also the quality should not be too bad. At present, the pulse width of the outgoing light of TOF radars is about a few nanoseconds, and the rising edge is required to be as fast as possible. Therefore, each product has a laser drive solution. There are also highs and lows.
b) The other is the receiver’s. Generally speaking, the time identification of the echo is actually the time identification of the rising edge. Therefore, when processing the echo signal, it is necessary to ensure that the signal is not distorted as much as possible. In addition, even if the signal is not distorted, since the echo signal cannot be an ideal square wave, the measurement of different objects at the same distance will lead to the fluctuation of the leading edge. For example, for the measurement of white paper and black paper at the same position, two echo signals may be obtained as shown in the figure below, and the time measurement system must measure that the two front edges are at the same time (because the distance is the same distance), which requires special treatment.
Figure 3. Differences in echo signals with different reflectivity
In addition, the receiving end also faces problems such as signal saturation and noise floor processing, which can be said to be difficult.
2. Performance PK, know what it is and why it is?
Having said so much, in fact, from the perspective of downstream users, it does not care whether it is easy or difficult for you to implement. Users are most concerned about two points: performance and price. Let’s talk about performance first. Most people who know this industry know that TOF radar is better than triangular radar in terms of performance. But what are the specific aspects, and what are the reasons behind it?
In principle, TOF radar can measure longer distances. In fact, in some occasions that require distance measurement, such as driverless car applications, it is almost always TOF radar. There are several reasons for the fact that the triangular radar does not measure far. First, it is limited in principle. In fact, it is not difficult to observe Fig. 1 carefully. The farther the distance of the object measured by the triangular radar is, the smaller the position difference on the CCD is. Because beyond a certain distance, the CCD is almost indistinguishable. Second, the triangular radar cannot obtain a higher signal-to-noise ratio than the TOF radar. TOF uses pulsed laser sampling, and also tightly controls the field of view to reduce the effects of ambient light. These are all prerequisites for long-distance measurements.
Of course, the distance does not mean absolute good or bad, it depends on the specific usage scenario.
When lidar depicts the environment, the output is a point cloud image. The number of point cloud measurements that can be completed per second is the sampling rate. In the case of a certain rotation speed, the sampling rate determines the number of point clouds in each frame of images and the angular resolution of the point clouds. The higher the angular resolution and the larger the number of point clouds, the more detailed the image depicts the surrounding environment.
As far as the products on the market are concerned, the sampling rate of the triangulation radar is generally below 20k, and the TOF radar can be higher (for example, the satellite-second TOF radar PAVO can have a sampling rate of up to 100k). The reason is that TOF only needs one light pulse to complete a measurement, and real-time time analysis can also respond quickly. However, the calculation process required by the triangular radar takes longer.
Figure 4. Imaging effects of different sampling rates for objects at the same position
(A): Low sampling rate point cloud pattern; (B): High sampling rate point cloud pattern (PAVO)
Lidar is essentially a ranging device, so the accuracy of distance measurement is an unquestionable core indicator. At this point, the accuracy of trigonometry at close range is high, but the accuracy of its measurement will become worse as the distance increases. This is because the measurement of trigonometry is related to angle, and as distance Increase, the angle difference will be smaller and smaller. Therefore, the triangular radar is often marked with a percentage (commonly such as 1%) when marking the accuracy, so the maximum error is 20cm at a distance of 20m. The TOF radar is dependent on the flight time, and the time measurement accuracy does not change significantly with the increase of length. Therefore, most TOF radars can maintain the accuracy of several centimeters within the measurement range of tens of meters.
2. RPM (frame rate)
In mechanical radar, the image frame rate is determined by the speed of the motor. As far as the two-dimensional lidars currently on the market are concerned, the maximum speed of the triangular radar is usually below 20Hz, and the TOF radar can achieve about 30Hz-50Hz. Usually, the triangular radar usually adopts the structure of upper and lower parts, that is, the upper part is responsible for laser emission, reception and acquisition, and the lower part is responsible for motor drive and power supply. The TOF radar usually adopts an integrated semi-solid structure, and the motor only needs to drive the mirror, so the power consumption of the motor is very small, and the speed that can be supported is also higher.
Of course, the difference in speed mentioned here is just an objective analysis of existing products. In fact, there is no essential connection between the rotational speed and whether the radar adopts TOF or trigonometry. The mainstream multi-line TOF radars also use the upper and lower split structures. After all, the optical design of the coaxial structure is subject to many restrictions. The speed of multi-line TOF radar is generally below 20Hz.
However, high rotation speed (or high frame rate) is very meaningful for point cloud imaging. A high frame rate is better for capturing fast-moving objects, such as vehicles on the highway. In addition, when mapping itself, the moving radar mapping will be distorted (for example, if a stationary radar scans a circle, then when the radar moves in a straight line, the scanned image becomes an ellipse ). Obviously, high speed can better reduce the effect of this distortion.
If you only look at the performance comparison, it seems that the performance of the TOF radar completely outperforms the triangular radar. However, the competition of products is not only the competition of performance parameters, users also care about price, stability and service.
At least in terms of cost, the current cost of triangular radar is lower than that of TOF radar, and the cost of short-range triangular radar is already at the level of 100 yuan. At present, the price of imported TOF radars is often more than 10,000 yuan. It can be said that the high price is an important factor restricting the further expansion of TOF lidar applications.
However, with the rise of domestic TOF radar manufacturers in recent years, the cost of TOF radar has been greatly reduced, and the price of domestic TOF radar products is quite competitive compared to imported brands. In the future, with the improvement of the production process and the further increase of the shipment volume, it is believed that the cost of TOF radar will be further compressed, and it is not impossible to reduce it to a level similar to that of the triangular radar.
4. Application scenarios
The scene of triangular radar is mainly used in indoor short-distance applications, and the most typical scene is the sweeping robot. In scenes with a large detection range (such as shopping malls, airports or stations), and outdoor scenes, TOF is more widely used. It is also worth mentioning that the exposed rotation scheme of the triangular radar makes the product very fragile in terms of dust and water resistance. In some special scenarios, such as the workshop where the AVG car works, there is often a lot of dust. In the environment, the motor of the triangular radar is very easy to be damaged. In contrast, the semi-solid design adopted by TOF radar can have better protection effect and longer working life.
Figure 5. Star-second TOF lidar PAVO
At present, domestic TOF radar is developing rapidly. The 2D TOF lidar PAVO launched by SIMINICS can reach a measurement distance of 20m, a point cloud rate of 100kHz, a maximum angular resolution of 0.036°, and an IP65 protection level. , its application has been involved in many fields such as unmanned driving, robot, AGV, security, road administration, etc. It is an excellent representative of domestic TOF radar.