Accurate LiDAR Simulation: Ocean™'s Approach to Optical Modeling Innovation
Ocean™, developed by Eclat Digital, is a spectral ray tracer that provides high-accuracy illumination calculations. Thanks to its specific spectral features, it allows a highly physically accurate description of materials and provides physically true predictive images. Relying on a valid and strong physical approach, it guarantees a perfect match between renders and the real world. This capability makes it ideal for assessing the quantity of light received by elements of a scene, such as light passing through glazing structures or evaluating light reaching specific areas in automotive contexts.
This real-world approach is particularly useful for for automotive manufacturers and OEMs looking for a reliable software to design and test optical tools. Ocean™ can model realistic illumination conditions, making it a powerful resource for evaluating the performance of LiDAR systems in real-world environments..
This article will show how to model a LiDAR system using Ocean™. We will first present the basics of LiDAR technology, then discuss the inputs and outputs specific to LiDAR modeling, and finally explore a simulation creation methodology.
What is LiDAR?
A LiDAR tool is a device that sends wavelengths in the near-infrared light range to capture, on a sensor or receiver, rays bouncing back from targets. This process allows for analyzing the surrounding environment, including object distances and geometries. You can see a drawing of how a LiDAR functions in figure 1.
Figure 1 - LiDAR functioning explanation
First, the emitter sends a specific amount of radiant flux (or intensity, expressed in W) as thousands of beams. The light spreads in space and reaches the targets. The amount of light arriving on a surface is measured as irradiance (in W/m² – see difference between radiant flux and irradiance on figure 2). Some of the light received by the surfaces is reflected back in specific directions, depending on the surfaces’ properties. While many rays are lost in the environment, some find their way back to the LiDAR and are captured by the receiver. The receiver measures the irradiance on its pupil.
Figure 2 - Radiant flux and irradiance difference
The LiDAR will transform the received light in an electric signal to analyse the results with system functions in an integrated system.
LiDAR tools have numerous applications across industries, including automotive, agriculture, aerospace, and building materials. From an optical perspective, optical system design for automotive, such as LiDAR, focuses on two aspects: the components’ design and their integration into the environment to study performance influences.
This system-level modeling approach is particularly well-suited for Ocean™ simulations.
Simulation Goals for LiDAR in automotive applications
In automotive scenarios, the focus is often on integrating LiDAR into complex environments rather than dissecting every component of the LiDAR system. By treating the emission system and sensor as black boxes, we can concentrate on the scene’s influence on LiDAR performance. This approach allows manufacturers and OEMs to simulate a LiDAR’s integration in a specific environment, we don’t need to focus on every component of the device. Instead, we represent the emission system and sensor as black boxes with proper characteristics and focus on the scene environment. This approach allows us to study how the environment influences the LiDAR’s performance while simplifying the simulation process.
At a macro level, LiDAR integration in automotive system with Ocean™ is modelled by simulating the emission of thousands of beams, tracking their light paths through the scene (e.g., through the LiDAR’s cover or bouncing off targets), and capturing what is received by the sensor (see figure 3).
Figure 3 - Simulation's light path representation
These are the basics of what is a LiDAR system level simulation. These simulations enable post-processing of results to validate performance in specific settings.
LiDAR Simulation Inputs: Emission and Receiver Systems Explained
There are several kinds of LiDAR already on the market. Here, we will consider a flash solid state LiDAR: its head does not move, and it flashes all its light pattern at one time. We will not consider the frequency of the LiDAR’s flash and model the exact moment when the device flashes.
Emission system
The emission system of a LiDAR device determines its field of view and light distribution. Even though the emitter component is modeled as a black box, certain LiDAR characteristics must be known to replicate its performance effectively. To simulate this system in Ocean™, we require:
- Horizontal and vertical field of view
- Focal length
- Pupil diameter
- Sensor dimensions and aspect ratio
- Rays’ beam properties (number, positions, angular opening, intensity, spectrum)
A quick representation is shown in figure 4.
Figure 4 - Emission system or emitter's black box’ principle (original drawings from https://www.reddit.com/r/Optics/comments/m0whoi/zemax_telescope_design_for_compact_imager/)
Figure 5, you can see an emitting surface (with the beams pattern), modelled in a CAD software, and ready to be exported to Ocean™. More than 10.000 beams are emitted by it, along a specific field of view (the surface is curved).
The surface design represents the whole black box system. Obviously, some hypotheses have been necessary, and they depend on the modelled LiDAR’s technology.
Figure 5 - Emission system or emitter's black box CAD design
Receiver
Ocean™ simplifies receiver modeling with built-in instruments. These are adjusted directly within Ocean™ without the need for external CAD design. Parameters include:
- Field of view and focal length
- Sensor dimensions and ratio
- Wavelength range for detection
- Focus distance
- F-number
Figure 6 - Receiver instrument settings
From Light to Data: Analyzing LiDAR Simulation Outputs
Outputs Localization
LiDAR simulations with Ocean™ recreate the light path from the emitter to the receiver. This allows users to extract information at key points. Using a numerical LiDAR allows to extract irradiance information at any point, including where it would be physically difficult. We propose to focus on two specific outputs localisation (as shown in figure 7):
- Target-level irradiance: Measures light received by a target (e.g., a vehicle or pedestrian) to ensure system accuracy. (Checking what is received by the target, at mid-way for the light path, and insure everything is coherent at this step)
- Sensor-level irradiance: Evaluates the amount and distribution of light captured by the LiDAR sensor. (The most important output, because it is the image seen by the LiDAR, a sensor vision)
Figure 7 - Outputs localisation representation
Outputs Localization
At those output levels, we want to measure irradiance. Meaning we measure the amount and repartition of light reaching the instruments. For the target level, the instrument measures exclusively at the target (for instance, a traffic sign) and compute the amount of light reaching it from the emitter and the environment. For the sensor level output, we have the receiver instrument, looking at the environment and target, capturing the light bouncing back from the target and from the environment.
These two kinds of outputs localisation require different instruments, available in Ocean™.
Looking at irradiance means we can post-process three things on each output levels:
- Each beam’s location on the sensor
- Each beam’s shape when seen by the sensor
- Each beam’s irradiance and intensity level when received by the sensor
We can post-process this set of information to do some extra-analysis: seeing how many points are received on one target permits to see if it has a chance to be detected by the LiDAR, the standard deviation of the point cloud after crossing several elements (which is a part of the distorsion), …
But irradiance remains optical information only. What usually interests people working with LiDAR is specific LiDAR information: how far can the LiDAR sees, how accurately can it detect objects, can it position them precisely in space? The transformation from optical information to LiDAR performance (shown in figure 8) is made by system functions inside the LiDAR. LiDAR systems are different for each LiDAR and is a property of LiDAR suppliers.
Figure 8 - Outputs information explanation
That means those additional practical information can be extracted from any numerical LiDAR model only with the support of a LiDAR supplier.
Building Realistic LiDAR Simulations with Ocean™
Simulation Workflow
Now we want to add a complete real-life scenario. For instance, in the case of a LiDAR inside a car, create a complete scene where the LiDAR is behind the windshield, on a road, with other cars and pedestrians, as well as in real weather conditions.
Before creating a complete real-life scenario, some steps are required first, to ensure we are physically representative in a simple scene. And then begin to add complexity.
Figure 9 - Difference between 1-path and 2-path modelling
1-Path Model: Simulate the LiDAR with a single flat target, focusing on emission results.
The only thing we need at the beginning is the LiDAR and a large flat diffuse target. Therefore, as first simplifications, we will set our simulation as empty, no ground, no exterior lights, no obstacles of any kind. At first, we will look at results on the target level, and not sensor level. This will allow us to reduce the complexity of the phenomenon (we do not look at the light being sent then reflected, but only at the light being sent).
2-Path Model: Extend to include sensor-level irradiance, capturing the full light path.
In a second step, with the LiDAR still an isolated system, instead of looking at irradiance results on the target, we look at irradiance at sensor level (quantity of light received by the sensor). This allows to see the LiDAR’s full phenomenon, and to begin to compare results with physical data if necessary. The difference between 1-path and 2-path is shown on figure 9.
Adding complexity with geometries
After validating basic simulations, additional elements can be introduced:
- LiDAR’s cover and holder
- Varied target dimensions and properties
- Environmental factors like weather or additional light sources
- Complex geometries and reflective properties
These additions of objects and environmental conditions will add complexity to the phenomenon to see how the LiDAR behaves under such and such conditions, allowing for sensitivity analysis of design and environmental factors.
Navigating the challenges of optical LiDAR simulations
While Ocean™ handles complex optical scenarios effectively, certain situations, such as light-absorbing materials or retro-reflective properties, may require simplifications.
For situations where the LiDAR is weak, numerical modelling will be difficult. The case of light absorbing material (pedestrians’ dark clothes) or retro-reflector properties (traffic sign) can be tough to model because the intensity coming back is very low. That does not mean it is impossible, but there will probably be some simplifications to consider. Specular properties such as car paints are also difficult in optical simulation, although Ocean™, thanks to its bidirectional path tracer, performs quite well in these case studies. Engineers should validate each step with physical data to benefit entirely from Ocean™’s accuracy.
Conclusion - Ocean™: A Game-Changer for LiDAR Simulation in automotive industry
Ocean™ equips automotive manufacturers and OEMs with the tools to simulate LiDAR systems under authentic conditions, delivering key insights into their operational performance. It employs spectral ray tracing for vehicle systems, enabling highly accurate simulations of LiDAR performance in real-world automotive environments. With Ocean™, manufacturers in the automotive industry can accelerate development timelines, streamline LiDAR system integration, and achieve reliable performance in diverse and challenging environments.
Q&A
What are the benefits of LiDAR for automotive applications?
LiDAR enhances vehicle safety by providing precise object detection and distance measurement. It is crucial for autonomous driving and ADAS.
How does Ocean™ support LiDAR simulations?
Thanks to its bi-directional path tracing and global illumination capacities, Ocean™ enables accurate modelling of light paths, helping manufacturers evaluate LiDAR performance in complex environments without physical prototypes.
Can Ocean™ simulate environmental factors?
Yes, Ocean™ can incorporate variables like real-world weather data, complex objects, and material properties to simulate real-world conditions.
Why focus on system-level modelling?
System-level modelling simplifies complex LiDAR integration in automotive systems, enabling efficient analysis of performance within a specific environment.
Responses