Frequency-modulated CW LiDAR enables 3D imaging and multi-parameter sensing

In a development that could reshape how autonomous vehicles perceive and protect themselves at the same time, a research team in China has unveiled a frequency-modulated continuous-wave LiDAR system that does far more than map the world in three dimensions. The new architecture, described by scientists led by Professor Yongkang Dong of the National Key Laboratory of Laser Spatial Information at Harbin Institute of Technology and the Zhengzhou Research Institute of Harbin Institute of Technology, performs high-precision ranging and multi-parameter environmental sensing simultaneously through a single optical platform and a single demodulator. Published in the journal Light: Advanced Manufacturing, the work demonstrates for the first time that the laser beam pattern typically reserved for building point clouds of surrounding scenery can also interrogate optical fibers, fiber Bragg gratings, Fabry–Pérot cavities, and gas cells to extract temperature, liquid density, and trace gas concentrations with remarkable precision.

The motivation behind the system is rooted in one of the most pressing safety challenges of the electric vehicle era: battery thermal runaway. When a lithium-ion battery enters an uncontrolled exothermic reaction, the consequences can be catastrophic, and early warning currently depends on the coordinated monitoring of several distinct physical and chemical indicators—rising temperature, changing electrolyte density, and the release of characteristic gases such as acetylene, carbon dioxide, and methane. In today’s vehicles, these functions are handled by entirely separate subsystems. Cameras and scanning LiDAR units handle perception of the road; discrete thermal sensors, gas detectors, and electrochemical probes handle battery health. Each additional subsystem adds cost, wiring complexity, calibration burden, and potential points of failure. The Harbin team’s central insight was that a properly designed coherent laser ranging system could carry both workloads at once, because the physics underlying distance measurement and the physics underlying spectroscopic sensing share a common foundation in the frequency domain.

At the heart of the technique is frequency-modulated continuous-wave, or FMCW, LiDAR. Unlike pulsed time-of-flight systems that emit short bursts of light and time their return, an FMCW LiDAR sweeps the frequency of a continuous laser beam in a linear fashion, typically across a wavelength or frequency ramp. When the emitted light, after traveling to a target and back, is mixed with a sample of the original laser light—known as the local oscillator—the two beams interfere to produce a beat signal whose frequency is directly proportional to the optical path difference between them. Because the frequency of the beat note encodes distance with extraordinary sensitivity, FMCW LiDAR systems routinely achieve millimeter-level or better ranging precision, and their coherent detection architecture provides strong immunity to ambient light interference, a persistent headache for pulsed automotive LiDAR operating in bright sunlight.

The new system extends this principle in a clever way. Rather than illuminating only free-space targets, the architecture also directs light into optical fibers, where it interacts with a series of specialized sensing elements. Fiber Bragg gratings, which reflect light at a narrow wavelength that shifts with temperature and strain, serve as thermometer elements. A Fabry–Pérot interferometer, whose resonance spectrum changes with the refractive index of the surrounding medium, provides a route to measuring liquid density—in this case, the sulfuric acid electrolyte used as a stand-in for battery fluid. A multi-pass gas cell, in which the laser light bounces repeatedly through a small volume, amplifies the weak absorption signatures of target gas molecules so that their concentrations can be retrieved from the reflection spectrum. Each of these elements, like the free-space target, produces a reflection at a distinct optical path length, and therefore at a distinct beat frequency in the detected signal.

The demodulation process unfolds in two elegant mathematical steps. First, a Fourier transform is applied to the recorded beat signal in the time domain, which separates the reflections in the spatial domain—each reflecting surface or sensing element appears as a peak at a position corresponding to its optical distance from the instrument. Because the fiber sensing elements are placed at known, distinct locations along the fiber, their reflections occupy well-separated frequency bins and do not interfere with one another or with the echo from the free-space imaging target. Second, an inverse Fourier transform is applied to each isolated peak after mapping it into the wavelength domain, reconstructing the full reflection spectrum of that individual element. It is within those reconstructed spectra that the physical information hides: the center wavelength of a Bragg grating peak reveals temperature, the fringe spacing of the Fabry–Pérot spectrum reveals electrolyte density, and the depths of absorption lines in the gas cell spectrum reveal the concentrations of specific molecules.

The experimental results are striking for a proof-of-concept demonstration. The team imaged a plastic plate bearing the “HIT” symbol placed 30 meters away, achieving an adjustable spatial resolution spanning 0.3 centimeters to 1.2 centimeters. This tunability—somewhat unusual in conventional LiDAR designs—arises from the flexible control of the frequency sweep parameters, allowing the system to trade off resolution against other performance metrics depending on the task at hand. On the sensing side, the system measured the temperature of the electrolyte solution with an accuracy of 0.5 degrees Celsius and its density with an accuracy of 3 × 10⁻⁵ grams per milliliter, a level of precision that rivals dedicated laboratory instrumentation. For gas sensing, the critical markers of battery thermal runaway were detected with minimum detectable concentrations of 0.07 parts per million for acetylene, 48 parts per million for carbon dioxide, and 0.56 parts per million for methane.

These gas detection limits deserve particular attention in the context of battery safety. Acetylene, in particular, is widely regarded as an early chemical fingerprint of electrolyte decomposition, appearing at very low concentrations in the initial stages of thermal runaway well before smoke, heat, or swelling become externally observable. A detection limit of 70 parts per billion means the LiDAR-based system could, in principle, flag a failing cell at a stage when intervention is still possible. Carbon dioxide and methane, released in later stages of the degradation cascade, provide corroborating evidence and information about the severity of the event. By integrating this chemical early-warning capability into the same optical head that performs routine 3D imaging, the system collapses what would otherwise be a multi-instrument safety architecture into a single coherent channel.

The conceptual bridge that makes all of this possible is the close kinship between FMCW LiDAR and optical frequency domain reflectometry, or OFDR. When the FMCW approach is applied inside optical fibers rather than through free space, it becomes OFDR, a technique long valued in the fiber sensing community for its exceptional spatial resolution and large dynamic range. Both methods rely on a linearly frequency-modulated continuous light source and an identical positioning principle: converting optical path differences into beat frequencies. The Harbin team’s contribution is to recognize that this shared foundation means a single laser, a single interferometric receiver, and a single signal processing chain can serve both the free-space ranging task and the distributed fiber sensing task without duplication. The reflections from the imaging target and the sensing elements are simply interleaved along the frequency axis and sorted out in software.

The broader implications extend beyond the automotive sector. The researchers point to spacecraft as another application domain where the combination of precision imaging and multi-parameter sensing would be valuable, noting that mass, power, and integration constraints make multifunctional payloads especially attractive in aerospace. In new energy vehicles, the unified architecture promises to support both the perception functions demanded by automatic driving—obstacle detection, range measurement, and 3D environmental mapping—and the battery management functions demanded by safety regulators, using only one demodulator and one optical source. This kind of functional consolidation addresses a genuine engineering tension: as vehicles grow more autonomous and their batteries grow more energy-dense, the sensor burden multiplies, and any technology that reduces the number of independent subsystems while expanding capability carries significant economic and reliability benefits.

Looking forward, the team forecasts that the technique could provide a new integrated solution for improving the safety of new energy vehicles, essentially embedding a distributed nervous system into the vehicle’s structure. Optical fibers threaded through battery packs could monitor temperature and electrolyte condition at multiple points, gas cells positioned near cells at risk could sniff for chemical precursors of failure, and the same beam pattern would continue to chart the road ahead—all processed by one coherent receiver. While scaling from laboratory demonstration to production-grade automotive hardware will require further work on packaging, environmental robustness, laser sweep linearity, and real-time signal processing throughput, the demonstration establishes a compelling proof of principle. It suggests that the LiDAR of the future may not merely watch the world go by, but also listen—continuously and with laboratory-grade sensitivity—to the subtle physical and chemical signals that precede danger within the vehicle itself.