Continuous Wave Lidar Simulation via Ray Set Mixing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lidar sensor simulation methods using point clouds are insufficient as they fail to capture the time-varying signals and artifacts present in real lidar data, such as object shape, class, and environmental effects, which are crucial for accurate sensor development and algorithm training.
Innovation Solution
A method and device for simulating continuous wave lidar sensor data by generating a ray set with specific emission characteristics, propagating rays through a simulated scene, computing signal contributions, and mixing them with the CW signal to produce an output signal that includes phase shifts and throughput, effectively recreating the full waveform signals of real lidar sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If point cloud simulation methods are used, then the simulation process is simple and fast, but the simulation data lacks time-varying signal information and artifacts present in real lidar data
Solution Approach 1:
The patent segments the continuous wave lidar signal into multiple discrete rays, each representing a specific time interval. This segmentation allows the simulation to process individual signal components separately while maintaining the overall time-varying structure, thus preserving signal information without requiring simulation of the entire continuous waveform at once.
Solution Approach 2:
The patent adds a time dimension to traditional point cloud simulation by incorporating emission starting times and durations for each ray. This transforms the simulation from static spatial point clouds to dynamic spatiotemporal signal representations, enabling recovery of time-varying signal characteristics while maintaining computational efficiency through selective time sampling.
2Loss of information
If full waveform signal simulation is implemented, then complete signal information including artifacts is preserved, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent applies partial action by simulating only the essential components of the full waveform signal - specifically selecting key rays that represent critical signal features such as direct reflections and important multiple bounces. This partial simulation approach captures the most significant signal information and artifacts without requiring computation of every possible ray path, thus reducing processing complexity while maintaining signal fidelity.
Solution Approach 2:
The patent changes key parameters of the ray representation including emission starting time, emission duration, and ray propagation characteristics to efficiently encode signal information. By optimizing these parameters, the simulation achieves accurate signal reconstruction with reduced computational requirements compared to full waveform simulation.
3Shape
If multiple rays per beam are used to provide multiple points, then object shape information is improved, but the simulation still loses the time-varying signal characteristics
Solution Approach 1:
The patent merges the advantages of multiple rays per beam with time-varying signal representation by associating each ray with specific emission time parameters. This combination allows the simulation to simultaneously capture object shape information through multiple ray paths and preserve time-varying signal characteristics through temporal parameter assignment, resolving the trade-off between geometric detail and signal fidelity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides high-quality, complete, and accurate simulation data that includes object shapes, classes, and environmental artifacts, enhancing sensor development and algorithm training by accurately mimicking real lidar sensor outputs.
Implementation Method 1
The propagation of the ray is computed using ray-tracing up to the nearest intersection point with a surface
Implementation Method 2
only mirror-reflection is considered
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to a method for simulating sensor data of a continuous wave, CW, Light Detection and Ranging, lidar, sensor, comprising the steps: generating a ray set comprising at least one ray, based on a CW signal 51, each ray in the ray set having an emission starting time (t_0) and an emission duration (T); propagating, for each ray in the ray set, the ray through a simulated scene comprising at least one object; computing, for each ray in the ray set, a signal contribution 71 of the propagated ray at a detection location in the simulated scene; generating an output signal 73, based on mixing the CW signal 51 with the computed signal contributions of the rays in the ray set; and at least one of storing and outputting the output signal. The invention provides simulation data corresponding to full-waveform signals of CW lidar sensors. The provided sensor data also allows to determine additional information, such as object shapes and object classes, and object orientation. Moreover, the sensor data also comprises artifacts which are present in real lidar data, which originate from physical properties of the lidar sensor and the environment and are absent in point-clouds- type data, such as multiple reflections of the lidar beam. Therefore, sensor development can be improved by also taking said artifacts into account. The invention accurately simulates how a real CW lidar sensor works and therefore accurately recreates the output of a real CW lidar sensor. The provided high-quality data, which is both complete and accurate to the real world, can also be of high value for developing signal processing algorithms, training neural networks, virtual validation, and the like.