Lidar for autonomous driving - How science fiction becomes science reality

08/04/2021 Know-How

It is the dream of many to be brought from A to B without any effort on their part whatsoever – a dream that is embodied in Knight Rider’s K.I.T.T. or the vehicles in “The Fifth Element”. Modern cars are increasingly approaching these visions of the future, and various sensor technologies play one of the most important roles here. One of the most promising of these is lidar.

Like radar, lidar is a method for detection and ranging (DAR). Both use the same method of echolocation as bats. They send out ultrasonic waves and detect where objects or prey are located based on how they are bounced back. While radar uses radio waves, lidar uses light waves.

Lidar sensors use a pulsed laser diode that sends out a light pulse. If it is reflected by an obstacle, the sensor will detect it. The time of flight (ToF), which is the time between the sending and receipt of the light waves, allows the distance between the sensor and obstacle to be calculated.

Highly sensitive detectors

The light waves are dispersed in many directions depending on the distance and form of the reflective object or living being. This is why the broader the detector range, the more precise the image of the environment it produces - because it can detect even more reflections. Current lidar sensors use avalanche photodiode (APD) arrays of 8, 12 or 16 diodes. Each diode represents a pixel of the overall image. This means that in addition to the size of the array, the distance between the diodes (=pixels) is also a factor in the sensor's resolution. The APD sensitivity also plays a role. Ideally, they should detect as few traces of the reflected beam as possible.

The ideal light beam

The length of the light pulses plays a critical role in the sensor resolution, which is why lidar manufacturers put a lot of effort into developing pulse lengths that are as short as possible. Presently, they measure 5ns on average up to a maximum of 10ns. Another factor is the size of the light beam. Because the laser diode sends out an extremely focused light beam, it can only measure the distance of a point that is of the same size. This is nowhere near sufficient for use in driver assistance systems and certainly not in autonomous vehicles. There are various solutions for enlarging the field of view (FOV). The challenge here is to detect even the smallest faces within a large FOV.

Eye and skin safety

One limiting factor in lidar development is eye safety. Given that laser beams may shine into the retinas of people, especially when used for road traffic, it is important to ensure that it does not cause eye damage. Human skin can also be attacked by laser beams. The standard EN 60825-1 defines various classes based on their risk to eyes and skin - both wavelength and pulse length play a role here. Three examples for laser classes: Class 1 applies to laser radiation that is non-hazardous or is used in an enclosed housing. Class 2 applies to laser radiation in the visible spectral range of between 400 and 700nm. With a short exposure time of max. 0.25s, it poses no danger to the eye. Class 4 lasers are highly dangerous for the eyes and skin, even if diffused.

Flash Lidar - Scattered light

One method of widening the FOV is based on scattering the light beam so that it covers a large FOV with a broad beam angle. Known as "flash lidar", the light in this case is diffused, however, and much weaker than a focused light beam. Even so, to achieve a long range and high resolution, laser diodes with a very high output of 1-2kW are used.

For applications where objects only need to be detected at short distances, vertical-cavity surface-emitting laser diodes (VCSELs) with a wavelength of between 850 and 940nm are ideal. They can be used to develop powerful 2D arrays. For detectors, highly sensitive sensors that can even detect individual photons - known as single-photon avalanche diodes (SPADs) can be beneficial. In order to increase range and also for conditions with intense sunlight exposure, the Fraunhofer Institute has developed CMOS SPAD detectors for microelectronic circuits and systems. SPADs are integrated into a CMOS process that is certified for the automotive industry and has been optimized for optoelectronic applications. This produces highly sensitive, very fast avalanche photodiodes with a momentary amplification of up to 108, high pulse rates and low noise.

Laser Components offers a flash lidar sensor with CMOS SPADs - the SPAD2L192 is a 192×2-pixel solid-state CMOS sensor for flash lidar applications. It measures distances based on the first-photon, direct ToF principle. Single-photon detectors offer very high sensitivity and high temporal resolution. The pixel-internal time-to-digital converter with a temporal resolution of 312.5 ps and an end-of-scale value of 1.28μs enables a nominal range of 192m and a resolution at a distance of 4.7cm.

Scanning Lidar - Mobile mirrors

To preserve the intensity of the light while covering a wider FOV, scanning lidar technology employs the principle of "scanning" the field with a beam of light. Using moving micro-mirrors, the light is directed over the FOV to be scanned. Scanning lidar sensors usually use between 1 and 16 laser diodes. Edge-emitting lasers with a wavelength of 905nm generate the best results here, while high-power laser diodes with over 100W achieve ranges of up to 150m.

As only a few diodes with relatively low power are sufficient, scanning lidar sensors offer good thermal properties. This enables very high pulse rates, allowing for eye safety even at a wavelength of 905nm.

The FOV is normally 145° on the y-axis and 3.2° on the z-axis. Theoretically, a 360° panoramic scan should be possible with this technology, but in practice, there do tend to be "blind spots" - the light beam is unable to scan the immediate vicinity - but additional radar and camera solutions provide a workaround for this flaw. However, due to their size and poor robustness, scanning lidar sensors are not suitable for use in vehicles. They measure around 10.5cm x 6cm x 10cm, making them too large to be used in the housing of a spotlight, for example. The moving mirrors are also sensitive to vibrations, impact, dust and extreme temperatures of the types that cannot be avoided in vehicles.

Appropriate diodes are available from Laser Components - the 905DxxUA series includes pulse laser diodes with single and multi-junction designs with up to 110W laser output and a wavelength of 905nm. The components are absolutely reliable, offer superb thermal stability and very precise chip alignment in a hermetically sealed package. This makes them suitable for distance measurement and obstacle detection, surveying equipment, laser radars and many medical applications. The AEC-Q101-qualified models can also be used in automotive applications.

Si-APD or Si-APD arrays are recommended for detectors. The SAHA series Si-APDs from Laser Components are optimized for wavelengths between 850 and 905nm. The semiconductor material is especially efficient here, and the pulsed laser diodes also emit at these wavelengths. In a miniature SMD package, Si-APDs offer high quantum efficiency and thus high sensitivity and low noise. The SAH1Lxx array series with 8, 12 or 16 high-sensitivity Si-APDs in an LCC44 package with a protective window offers the same characteristics. They offer especially low spacing of 40µm. An array with 12 APDs is also available in a 14-pin DIL package. In addition to the standard array, there are also customized arrays available with specific numbers and sizes of APDs.

Solid-State Lidar - Semiconductors instead of mechanical components

Solid-state lidar sensors are a smaller, more robust alternative. These rely on semiconductor instead of mechanical components to direct the light beam. There are two versions of these: those with MEMS-based mirrors and those with OPAs (optical phased arrays).

Lidars with MEMS-based mirrors use a matrix of micromirrors, with each mirror having an edge length of just a few micrometers. They switch back and forth between two positions several thousand times a second, moved by electrostatic fields. These kinds of lidars are used in applications such as scanner checkouts or DLP projectors (digital light processing), so they are a proven, tested technology with relatively low production costs.

However, for automotive applications, the sensors need to satisfy much more stringent requirements. For example, they require a wider FOV compared to POS checkout systems or projectors. With a scan frequency of over 100Hz, current solutions offer angles of 40°; MEMS systems with broader angles are currently under development.

With lidars using OPAs, the phase of the emitted light from each laser diode is modified by a modulator to enable a pulse to cover a larger area. The technology is still currently in the research stage. A variant of this uses a silicon circuit measuring just several square millimeters as a replacement for the rotating emitter and detector unit. For higher outputs and a broad FOV, tests are underway using wavelengths that reach further into the infrared range than the 905nm currently conventionally used. A wavelength of 1550nm, for example, is not harmful to the eyes, but could be adversely affected by snow or rain. Other detectors are also required here.

Many technologies turn science fiction into science reality

We're still a few years away from achieving autonomous driving as science fiction films depict it, but each assistance system - be it adaptive cruise control (ACC); emergency brake assist (EBA) or lane departure warning (LDW) - is a step towards achieving that goal. For many of these, lidar is an essential component, one that absolutely should be combined with other technologies such as ultrasonic sensors, cameras and radar solutions - because every technology has its strengths and weaknesses.

 

Find components at www.rutronik24.com.

Subscribe to our newsletter and stay updated.