Autonomous Driving Sensor Simulation Using Reflection Rate Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving simulation systems face significant computational complexity and reduced simulation speed due to the need to accurately simulate every polygon and light reflection of complex objects, such as rotating bicycle wheels, which increases processing demands.
Innovation Solution
A dynamic reflection rate table is used to simplify the simulation of target objects by defining their shape, size, and reflection information, allowing for a pre-determined mapping of attributes like the number of spokes, wheel speed, and angle to a reflection rate, which is then used to generate sensor data, such as LiDAR point clouds, reducing the need for detailed polygon recreation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed polygon recreation and full light reflection simulation are used for complex objects like bicycle wheels, then measurement precision and reliability of sensor simulation are improved, but computational complexity increases and simulation speed decreases
Solution Approach 1:
The patent uses simplified geometric primitives (cylinders, spheres, planes) to copy the essential reflective properties of complex objects like bicycle wheels without recreating every polygon. The reflection rate table stores pre-calculated reflection characteristics that represent the optical behavior of complex surfaces, allowing accurate sensor simulation at reduced computational cost.
Solution Approach 2:
The patent transforms the complex problem of simulating every surface polygon into a parameter-based approach by defining objects with key attributes (number of spokes, wheel radius, material properties) and using reflection rate tables that map these parameters to reflection probabilities. This parameterization allows accurate simulation without detailed geometric reconstruction.
2Reliability
If every polygon and light reflection is simulated accurately, then reliability of sensor simulation is improved, but device complexity and processing demand increase
Solution Approach 1:
The patent extracts only the essential reflective properties of complex objects and stores them in reflection rate tables, separating the complex geometric details from the essential optical behavior. This extraction allows the simulation to use simple geometric primitives while maintaining accurate reflection characteristics through pre-stored parameters.
Solution Approach 2:
The patent performs preliminary calculations of reflection rates for various object configurations and stores them in lookup tables before runtime simulation. This pre-computation phase captures the complex light reflection physics in advance, allowing the actual simulation to simply query pre-calculated values rather than performing complex ray-tracing calculations in real-time.
3Productivity
If simple object definitions with reflection rate tables are used, then simulation speed and productivity are improved, but manufacturing precision of sensor data may be reduced
Solution Approach 1:
The patent replaces the mechanical/geometric approach of detailed polygon modeling with a statistical/probabilistic approach using reflection rate tables. Instead of tracking individual light rays through complex geometries, the system uses probability distributions based on pre-calculated reflection rates, achieving comparable accuracy with much lower computational complexity.
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 significantly reduces computational complexity and enhances simulation speed by using pre-determined reflection rate tables based on real-world testing data, enabling faster and more efficient simulation of sensor data without the need to recreate exact object shapes.
Implementation Method 1
Active sensors, such as a LiDAR sensor, implemented in autonomous driving systems generate sensor data by transmitting light or waves (e.g., a laser beam) and receiving a reflection of that transmitted light that has bounced off of a surface of a target object
Data Source
AI summary
A method of simulating a sensor in an autonomous driving simulation includes obtaining values for a plurality of attributes of a target object to be sensed by a sensor simulator in the autonomous driving simulation. The sensor simulator may simulate an active sensor that outputs rays to an object and receives reflections of the rays from the object. The method also includes inputting the obtained values for the plurality of attributes to a predetermined reflection rate table to obtain a reflection rate mapped to the obtained values; and generating sensor data corresponding to the target object based on the obtained reflection rate.