Linearization of Optical Frequency Modulation and Improvement of Distance Accuracy in FMCW LiDAR Using the Predistortion Method

Abstract

In this paper, we propose and experimentally verify a ‘software-centric temperature-adaptive predistortion linearization technology’ that actively compensates in real-time for thermal drift errors caused by light source modulation nonlinearity and self-heating, which inevitably occur in standalone open-loop frequency-modulated continuous wave （FMCW） LiDAR systems for small mobility platforms. Conventional bulky thermoelectric cooling （TEC） chambers and complex optical feedback loops have been major obstacles to drone and robot integration by limiting system lightweighting and low-power consumption. To overcome these hardware limitations, this study constructed a system capable of measuring temperature without an external auxiliary optical system by combining a thermistor with an integrated laser diode （LD） and a constant-current driving circuit, and by monitoring the analog voltage of a variable T_act pin based on Ohm's law in real-time. In the numerical analysis and linearization modeling stages, the initial flyback transient response region （0 to 40 μs） occurring during period crossing was excluded by defining it as an invalid interval to block uncertainty. Subsequently, the fitting accuracy of the least squares method was maximized by establishing a time-axis third-order polynomial curve model based only on 102,400 high-density data points extracted from the stabilized effective prime interval from 40 to 200μs, which is the actual measurement region. In addition, we developed a 'temperature function modeling' by converting a set of discrete modulation coefficients （a, b, c, d） of a third-order polynomial function extracted in 5°C increments within the self-heating saturation band of 25°C to 45°C into a second-order polynomial function for temperature T_ACT. Through this, we completed a software architecture capable of performing real-time dynamic computation on a predistortion modulation signal corresponding to an arbitrary actual operating temperature with an extremely high coefficient of determination （R^2） of over 99.3%, utilizing only 12 sets of temperature-dependent eigenparameters （α_i,β_i,γ_i） without storing a vast lookup table （LUT）.

To verify the practical measurement performance of the proposed algorithm, we constructed a homodyne FMCW lidar hardware system based on a spatial optical system that completely excludes auxiliary interferometers. We secured a basic distance resolution of 9 cm by combining linear modulation parameters with a light source modulation bandwidth of 1.65 GHz and a chirp sweep time of 200 μs. As a result of actual measurements on a one-dimensional fixed target with a retroreflective plate, it was confirmed that in a simple ramp signal environment without modulation compensation, severe frequency distortion and spectral broadening errors were present. In contrast, injecting the proposed predistortion voltage resulted in a linear increase of 55 kHz per meter across the up to tens meters, achieving the design specifications for the beat frequency, and yielded a narrow single peak in the FFT spectrum where energy leakage to surrounding bins was blocked. Furthermore, a 2-axis scanner was operated in a corridor environment with a mix of multiple objects, and by organically integrating it with the ROS （Robot Operating System） framework in an Ubuntu environment, 5,000 high-density 3D point clouds per frame were reconstructed in real time. As a result, geometric distortion and depth noise, which are common in non-linear light sources, were effectively suppressed across the entire range, demonstrating 3D spatial volumetric mapping performance with high alignment with actual geometric structures. Finally, a 1:1 simultaneous comparative measurement with a commercial Direct Time-of-Flight （Direct ToF） LiDAR was performed under high-density smoke/fog visibility conditions, a typical limiting scenario for autonomous driving sensors. While ToF systems completely lost the ability to perceive forward objects and suffered data loss due to multiple scattered light noise from haze particles, the linear FMCW lidar system of this study demonstrated excellent atmospheric penetration and environmental robustness by constantly suppressing external scattered disturbances as noise through a coherent homodyne interference mechanism and precisely restoring the contours of a target blocked 14 m behind and background structures without defects. In conclusion, the software-centric open-loop temperature-compensated predistortion linearization technology established in this study boldly eliminates large control hardware, thereby consistently ensuring dramatic weight reduction, miniaturization, and low power consumption of the FMCW lidar. At the same time, it clearly demonstrated that it can maintain the highest measurement accuracy and frequency stability even under severe thermal drift errors and outdoor disturbance environments. This technology is expected to serve as a pivotal engineering foundation for securing the practical operational reliability of next-generation core perception sensors for unmanned aerial vehicles （UAVs）, autonomous mobility, and small robot platforms.