Modern autonomous driving and robotics technologies rely on the ability of a machine to “see” the world around it in the same way as a human does, but with much greater accuracy. The key element of this system is lidar - an optical rangefinder that creates a detailed three-dimensional map of the area in real time. It is thanks to this device that self-driving cars can confidently move along complex city routes, avoiding obstacles and reading road markings.
The operating principle of the device is based on the emission of laser pulses and analysis of the time during which the light returns back to the sensor. This method, known as Time-of-Flight, allows you to determine the distance to objects with millimeter accuracy. Unlike cameras, which rely on light, laser scanner works effectively in complete darkness, and modern models cope even in conditions of light fog or rain.
However, blind trust in sensor readings can lead to critical errors in navigation. ⚠️ Attention: Even the most advanced lidar can malfunction when the signal is reflected from mirror surfaces or light is absorbed by black objects, so systems are always duplicated by cameras and radars.
The implementation of such systems in cars requires a deep understanding of the physics of the process. Optical radar emits thousands of pulses per second, forming a so-called “point cloud”. Each point has its own coordinates in space, which allows the on-board computer to build a three-dimensional model of the environment. This is the foundation for algorithms that make decisions about braking or changing lanes.
How laser scanning works
The basis of the technology is the measurement of the transit time of a light beam. The laser diode emits a short pulse, which is reflected from the object and returned to the photodetector. Knowing the speed of light, the processor calculates the distance using a formula where delay time is a key parameter. This process is repeated millions of times per second, creating a dense array of data.
To create a three-dimensional image, a rotating mechanism or solid-state matrix is used that directs the rays in different directions. Viewing angle modern devices can reach 360 degrees horizontally and up to 90 degrees vertically. This provides complete coverage around the vehicle without blind spots.
When installing additional lidar equipment, it is important to consider the cleanliness of the protective glass - even a thin film of dirt can scatter the laser beam and create “ghost” objects on the map.
The measurement accuracy directly depends on the rotation speed of the scanning element and the power of the emitter. High speed models use technology Flash LiDAR, which illuminates the entire scene with one powerful pulse, allowing you to instantly obtain a full frame without mechanical rotation. This significantly increases system reliability in dynamic driving conditions.
Types of lidars for navigation
There are several main types of sensors on the market, each of which has its own design features and scope of application. Mechanical models mounted on the roof of many test drones physically rotate as they scan the space. They provide the highest detail, but have moving parts that are subject to wear.
Solid state (Solid-state) devices have no moving parts, which makes them more compact and durable. They are often built directly into the car body or bumper. Hybrid solutions combine elements of both approaches, offering a balance between cost, reliability and quality of the resulting picture.
⚠️ Attention: sensor characteristics declared by manufacturers may differ in real operating conditions due to thermal expansion and vibration, therefore system calibration is a mandatory procedure after installation.The choice of scanner type depends on the required range and viewing angle. For parking at low speeds, short distance models are sufficient, while for highway driving, long-range versions with a narrow beam are required.
Comparison of characteristics of different models
When choosing equipment to create a navigation system or modernize an existing one, it is important to take into account the technical parameters. Different models are optimized for different tasks: some provide a high density of point clouds at close range, others are able to “pierce” long distances.
Parameter Mechanical 360° Solid-state Flash LiDAR Range up to 200-250 m up to 150 m up to 50-80 m Viewing angle 360° horizontal 120° - 150° wide, fixed Refresh rate 10-20 Hz up to 100 Hz high (personnel) Reliability Medium (mechanical) High Very high As can be seen from the table, mechanical scanners They gain in range and viewing angle, but lose in reliability and size. Solid-state solutions are becoming the standard for mass production of automobiles due to their compactness.
For urban environments, where objects are close and movement is chaotic, solid-state lidars with a high update rate are often more effective than long-range mechanical counterparts.
Application in the automotive industry
In the automotive industry lidar navigation has become the standard for autonomous driving systems at L3 level and above. Sensors allow the car not only to see obstacles, but also to classify them: distinguish a pedestrian from a pole, determine the dimensions of a truck and predict the trajectory of a cyclist.
In addition, the technology is used to build maps (HD-maps). Specialized mapping vehicles are equipped with arrays of scanners that create digital twins of roads with centimeter accuracy. These maps are then used by regular cars to position themselves with precision not possible with GPS.
Why haven’t lidars been installed in mass-produced cars for a long time?
For a long time, the main obstacle was the high cost of production, reaching tens of thousands of dollars per unit, and the difficulty of miniaturizing mechanical parts for a mass production line.
Reducing component costs and switching to semiconductor technologies open the way for installing sensors even in budget models. This allows the introduction of advanced emergency braking and adaptive cruise control systems that work in all lighting conditions.
Software data processing
Raw data from the sensor is a cloud of points, which in itself is not very informative for decision making. A complex software stack comes into play that filters out noise, combines points into objects and tracks their movement through time. This process is called segmentation and classification.
Machine learning algorithms are trained on millions of examples to recognize patterns. For example, the system must understand that a group of dots moving at a certain speed and height is a person, and a static rectangular object near the road is a speed limit sign. Modern neural networks are capable of restoring the shape of an object even if it is partially blocked (occluded) by other obstacles.
Data processing requires significant computing resources. Therefore, cars are equipped with powerful on-board computers that can process terabytes of data per hour. The delay between receiving a signal and the system's response must be minimal to ensure safety.
☑️ Checking the operation of the scanning system
Done: 0 / 4Challenges and limitations of technology
Despite progress, the technology remains weak. Weather conditions such as heavy snow, dense fog or rainfall can scatter the laser beam, creating "noise" or reducing range. In such cases, the system should automatically switch to backup channels such as radar.
Another problem is the mutual influence of sensors. If another vehicle with a similar system is driving in the opposite direction, their lasers may interfere, creating false readings. Manufacturers solve this by encoding pulses and synchronizing frequencies so that each optical radar recognized only his own signals.
⚠️ Attention: When installing external sensors on a vehicle yourself, you must take into account the legal restrictions on the use of laser equipment of a certain power in your country.It is also worth considering the influence of vibration and temperature. Long-term use can lead to system decalibration, when the virtual map no longer matches reality. Regular testing and software adjustments are necessary to maintain accuracy.
Prospects for the development of navigation systems
The future is sensor fusion—the deep fusion of data from cameras, radars, and lidars. No technology is perfect, but a combination of them can compensate for the shortcomings of each. Artificial Intelligence will play an increasingly important role in interpreting this data, allowing cars to anticipate dangerous situations.
The emergence of chips that combine an emitter, detector and processor in one chip is expected. This will dramatically reduce the cost and size of devices. Quantum lidars, currently under development, promise even greater accuracy and noise immunity.
The development of smart city infrastructure will also be tied to this technology. Traffic lights and road signs equipped with scanners will be able to transmit data directly to cars, warning of hidden threats around the corner. This will make orientation in space absolute.
Can lidar work through a windshield?
Regular car glass may distort the laser beam or create double reflections, causing errors. To work through glass, special optical windows or moving the sensor outside the glass are required.
Is a laser scanner harmful to the eyes?
Certified automotive lidars belong to laser safety class 1, which means complete safety for human eyes even when directly hit by the beam. The radiation power is strictly regulated.
How often is calibration needed?
Software self-calibration occurs continuously while driving. However, after replacing body parts, impacts or removing the bumper, professional calibration is required on a special stand.