ALLPCB
Introduction
Historical experience shows that convergence of technical routes often signals large-scale adoption of new technologies. Viewed this way, autonomous driving still faces fragmentation: multiple competing approaches coexist and debate continues across many dimensions.
Major debates include single-vehicle intelligence versus vehicle-infrastructure cooperation, pure-vision versus multi-sensor fusion, reliance on high-definition maps versus map-free approaches. At a finer level there are disputes over LiDAR placement, integrated parking-and-driving solutions versus separate systems, and so on.
Among these debates, the most active and interesting is the one around LiDAR technology. Disputes include:
- distance measurement method: time-of-flight (ToF) versus frequency modulated continuous wave (FMCW);
- laser types: edge-emitting laser (EEL) versus vertical-cavity surface-emitting laser (VCSEL);
- laser wavelength: 905 nm versus 1550 nm;
- detectors: avalanche photodiode (APD), single-photon avalanche diode (SPAD), and silicon photomultipliers (SiPMs);
- scanning modules: mechanical, hybrid solid-state, and solid-state. Hybrid solid-state includes prism-based, rotating-mirror, and micro-electro-mechanical system (MEMS) micro-mirror approaches; solid-state includes optical phased array (OPA) and flash solutions.
In theory these choices could form hundreds of combinations, implying many LiDAR types. However, not every combination is practical. This article series aims to explain which routes are impractical, which have been phased out, which are widely used today, and which are likely future directions. The first installment reviews the turbulent development history of automotive LiDAR.
01 A Turbulent History
Lasers were invented in the 1960s. The first widely recognized laser was a ruby laser operating at 694.3 nm, developed by Maiman and Lamb. Early LiDAR systems grew from laser technology and were initially used for scientific applications such as meteorology and terrain mapping.
In the 1980s LiDAR introduced scanning structures and larger fields of view, which enabled some commercial uses. Scanning 2D LiDARs from companies like Sick and Hokuyo became well known in industrial measurement.
After 2000, LiDAR evolved from single-line to multi-line scanning and began to show advantages in high-precision 3D environment reconstruction.
In 2004 LiDAR entered the autonomous vehicle spotlight. The first DARPA Grand Challenge attracted teams including Velodyne, a company founded by David Hall. Velodyne’s early participation led Hall to recognize the potential of rotating multi-beam LiDARs.
In 2005 Velodyne introduced a 360° multi-beam rotating LiDAR; the device was large and heavy but demonstrated the value of high-density point clouds. Velodyne continued to iterate and maintain contacts with autonomous vehicle teams.
By the 2007 DARPA Urban Challenge, several successful teams used Velodyne’s 64-beam HDL-64, bringing wide recognition to rotating multi-beam LiDAR.
Velodyne patented its high-definition LiDAR system (US7969558, often called the "558 patent"). In 2011 the U.S. Patent Trial and Appeal Board affirmed the patent’s standing, making it a foundational patent for mechanical rotating LiDARs and a point of contention in later disputes.
Another early LiDAR pioneer was Ibeo, founded in 1998 and briefly acquired by Sick. Ibeo supplied single-line scanning LiDARs to some DARPA teams. For example, Stanford used multiple Ibeo sensors in its 2005 entry.
In 2009 both Velodyne and Ibeo gained momentum. Ibeo regained independent operation and later partnered with automotive suppliers to develop production LiDARs, while Velodyne introduced the HDL-64E, which became widely used despite high cost.
Ibeo focused on production-ready automotive LiDAR. After years of development it delivered SCALA, a four-line automotive-grade LiDAR used as a front-mounted sensor on the Audi A8. However, production of A8 models with Level 3 features was later suspended by the automaker.
Ibeo then invested in solid-state LiDAR development. In 2019 it introduced ibeonEXT, a family of short-, medium-, and long-range solid-state LiDARs. Ibeo announced a strategic cooperation with a Chinese automaker, Great Wall Motor, to equip certain models with ibeonEXT, but the collaboration faced delays and later Ibeo entered insolvency in 2022.
Velodyne, meanwhile, dominated the market for a decade. High demand and limited supply placed Velodyne at the center of the LiDAR supply chain. In 2016 Ford and Baidu invested in Velodyne to secure priority supply. As low-cost domestic competitors emerged, competitive pressure increased.
In 2019 Velodyne sued several companies over the "558 patent". Later it reduced its direct presence in China and shifted to distributor models. By 2020 some disputes were settled and licensing agreements reached. However, the market for mechanical LiDAR shrank, and Velodyne faced internal and external difficulties. In 2023 Velodyne merged with Ouster and the combined company now operates under the Ouster name.
This history highlights many contributors to LiDAR’s development. The contrasting paths of Velodyne and Ibeo illustrate how technical choices, patent strategy, manufacturing scale, and customer relationships shaped the industry.
02 Laser Types and Principles
Lasers are central to LiDAR. This section summarizes laser fundamentals, types used in automotive LiDAR, and key considerations such as wavelength and eye safety.
1. How Lasers Are Generated
Laser is an acronym for Light Amplification by Stimulated Emission of Radiation. The physical principle rests on atomic energy levels and the stimulated emission process described by quantum models of the atom.
Electrons occupy discrete energy levels. Spontaneous emission occurs when an electron transitions from a higher to a lower energy level and emits a photon. Stimulated absorption occurs when an incoming photon with energy matching a level difference is absorbed, lifting an electron to a higher level. Stimulated emission occurs when an incoming photon induces an excited electron to drop to a lower energy level and emit a photon with identical frequency, phase, polarization, and direction as the incoming photon.
To generate net stimulated emission, a population inversion is required, where more electrons occupy the excited state than the ground state. This requires a gain medium and an energy pump source. To favor stimulated emission over spontaneous emission, many lasers employ an optical resonator (two mirrors forming a cavity) that causes photons to pass through the gain medium repeatedly, amplifying coherent light. When the amplified light exceeds the transmission threshold of the output coupler, a laser beam is emitted.
2. Laser Types
Lasers consist of a gain medium, a pump source, and an optical resonator, plus power and control systems. Gain media classify lasers into gas, liquid, and solid categories. Solid-state lasers include fiber and semiconductor types. Automotive LiDAR typically uses semiconductor lasers and fiber lasers; the following summarizes these two classes.
Semiconductor Lasers
Semiconductor lasers use a PN junction in a semiconductor material as the gain medium. Under forward bias, electron-hole recombination in the junction can produce photons. With a resonant cavity the device operates as a laser diode. Efficient light emission requires direct-bandgap semiconductors such as gallium arsenide (GaAs) or indium phosphide (InP).
Semiconductor lasers are compact, efficient, and well suited to mass production. They are commonly classified by cavity structure into edge-emitting lasers (EEL) and vertical-cavity surface-emitting lasers (VCSEL).
Fiber Lasers
Fiber lasers use an optical fiber doped with rare-earth elements as the gain medium and are typically optically pumped by semiconductor lasers. Their gain medium is the fiber itself, and cavity designs vary. Fiber lasers are known for high beam quality and power scalability.
3. Wavelength
Common LiDAR wavelengths are 905 nm and 1550 nm. Wavelength choice affects eye safety, atmospheric propagation, detector selection, and achievable range.
Eye safety standards such as IEC 60825 use Maximum Permissible Exposure (MPE) to classify lasers. For many pulse widths, MPE values for 1550 nm are higher than for 905 nm, meaning higher emitted power can be tolerated at 1550 nm. The reason is that 1550 nm radiation is absorbed more strongly by water, and the eye’s aqueous layers absorb much of the energy before it reaches the retina.
Higher allowable emission power at 1550 nm enables longer detection ranges. For example, a 1550 nm system may reach 250 m at 10% reflectivity, while a 905 nm system might reach 150 m under similar conditions. However, 1550 nm is more strongly attenuated by thin water films: a thin water layer can absorb a large fraction of 1550 nm energy, degrading performance in rain. Surface water films on measured objects can form even in light rain, so rain performance must be carefully evaluated when choosing wavelength.
03 Scanning Methods
Scanning method is one of the most debated aspects of LiDAR design. Scanning choices strongly influence product reliability, manufacturability, cost, and whether a solution can meet automotive requirements.
By scanning method, LiDARs classify into mechanical, hybrid solid-state, and solid-state systems.
Mechanical LiDARs rotate transmit and receive modules to achieve 360° scans. Hybrid solid-state includes prism-based, rotating-mirror, and MEMS micro-mirror approaches, where the transmitter and receiver are fixed and one or more mirrors are moved to steer beams. MEMS micro-mirrors eliminate motors and rely on micro-actuated mirrors. Solid-state LiDARs have no moving mechanical parts and include optical phased array (OPA) and flash approaches.
1. Hybrid Solid-State
Prism-Based
Livox developed a notable prism-based solution using a rotating double-prism (rotating wedge) to create non-repetitive scan patterns. By fixing a small number of transmit/receive modules and rotating optical elements, Livox aimed to obtain point cloud coverage comparable to high-line-count rotating LiDARs with fewer modules, reducing cost.
Achieving equivalent point-cloud quality with fewer modules requires very high rotation speeds of the prisms. For some use cases this means mirror rotation speeds much higher than typical automotive motor speeds, imposing stringent durability and bearing requirements. Livox’s consumer-oriented Horizon product gained attention for its cost-performance ratio, but later the company reduced focus on automotive LiDAR and shifted toward industrial and low-speed robot markets. The prism route can face challenges in meeting automotive durability and reliability requirements.
Rotating-Mirror
Rotating-mirror approaches have been used in automotive-grade products. Valeo and Ibeo developed the Scala LiDAR using a rotating-mirror design. Scala 1 and later Scala 2 reached automotive-grade production and were used as front-facing sensors on production vehicles.
Rotating-mirror designs can use one-dimensional mirrors (single-axis) or two-dimensional mirror assemblies (one axis for azimuth, one for elevation). One-dimensional rotating-mirror systems keep the number of scan lines equal to the number of transmit/receive modules; two-dimensional mirrors can increase effective line count while reducing module count. Rotating-mirror designs reduce motor speed requirements compared to prism-based solutions, but still must address thermal durability, field-of-view limits, and signal-to-noise challenges.
MEMS Micro-Mirrors
MEMS micro-mirrors are silicon-based devices with a tiny reflective surface supported by suspending beams. Two axes of motion (fast and slow) produce two-dimensional scanning. The slow axis (5–30 Hz typical) is often long and slender to achieve larger scan angles; the fast axis (>1 kHz) is shorter and operates at higher frequency for vertical scanning.
MEMS LiDARs offer fewer moving parts, potential cost advantages from semiconductor-scale production, and high scan rates. They can be implemented in coaxial or non-coaxial optical layouts. In a coaxial design the transmit and receive paths are aligned and often share optical elements, improving sunlight rejection and simplifying receiver design. However, in such designs the MEMS mirror size directly affects receive aperture and therefore range. Non-coaxial designs place receive optics off-axis, allowing smaller MEMS mirrors and improved robustness, but requiring higher-performance detector arrays and increased system complexity.
MEMS devices face thermal challenges. MEMS mirrors are sealed inside housings and can run significantly hotter than ambient, so automotive temperature ranges (-40 to 105 C for automotive-grade) present engineering hurdles for reliability and packaging.
2. Solid-State
Solid-state LiDARs have no mechanical scanning parts. The two main approaches are optical phased arrays (OPA) and flash LiDAR.
Optical Phased Array (OPA)
OPA leverages coherent beamforming. Multiple coherent emitters arranged in an array produce constructive interference in a desired direction by engineering relative phase shifts across elements, steering a narrow beam without moving parts. One-dimensional OPAs steer in one axis; two-dimensional OPAs steer in two axes. Implementing OPA requires precise phase control across many elements and is an active area of research.
Flash
Flash LiDAR emits broad, short laser pulses that illuminate the entire field of view at once and captures the return with an array detector, analogous to an active imaging camera. Flash systems typically use VCSEL arrays because VCSELs offer larger emitting areas and can better support wide fields of view. Edge-emitting lasers can be used but require more complex optics to expand and homogenize the beam.
Flash LiDARs typically have limited range due to VCSEL power density constraints. As VCSEL technology improves (for example, multi-junction VCSELs), longer-range flash systems become more feasible. Flash systems often pair VCSELs with high-sensitivity detectors such as SPAD arrays to maximize performance at low return power.
Flash LiDAR optics differ from scanning systems. Flash transmit optics aim to evenly illuminate the entire FOV, while receive optics require large relative apertures and uniform irradiance across the detector array.
Manufacturers that have pursued flash LiDAR include Continental, Ibeo, Ouster, and newer entrants. Early flash products had modest specifications, but improvements have made flash a viable approach for short- to medium-range applications, and ongoing developments aim to extend range and integration for automotive use.
The next installments of this series will analyze detectors, measurement methods, wavelength trade-offs, and practical combinations of technologies to identify which LiDAR routes are most viable for automotive deployment.
