Radar-Lidar Calibration via Point Cloud Entropy Minimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional target-based radar-lidar calibration methods are prone to target detection errors, require high control and resources, and are not suitable for unstructured environments or on-road use, making them inefficient and unreliable for autonomous vehicle systems.
Innovation Solution
The use of point cloud registration and entropy minimization to align radar and lidar point clouds gathered from different vehicle poses, allowing for a more robust and accurate calibration process that can be performed in unstructured environments without special calibration targets, by generating aggregated maps and optimizing six degrees of freedom between sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional target-based radar-lidar calibration methods are used, then calibration can be performed with structured targets, but the method is prone to target detection errors and requires high control and resources
Solution Approach 1:
The patent extracts the calibration problem from the constraint of using special targets. Instead of relying on detected targets, the system uses directly observable point clouds from radar and lidar to perform calibration, removing the intermediate target detection step that introduces errors
Solution Approach 2:
The patent introduces point cloud registration as an intermediary process between raw sensor data and calibration results. By registering point clouds from multiple poses and minimizing entropy, the system creates a robust intermediate representation that enables accurate calibration without direct target dependency
2Adaptability or versatility
If traditional target-based calibration methods are used, then calibration can be performed in controlled environments, but the method is not suitable for unstructured environments or on-road use
Solution Approach 1:
The patent makes the calibration system universal by enabling it to function in multiple environments (controlled and unstructured/on-road). The point cloud registration approach with entropy minimization serves as a multi-functional solution that works regardless of environment, using the same core algorithm adapted to available data
Solution Approach 2:
The system performs calibration using data naturally collected during normal vehicle operation. Instead of requiring external calibration equipment or controlled environments, the vehicle uses its own sensor data from regular driving to automatically calibrate, making the process self-service and environment-independent
3Productivity
If traditional calibration methods are used, then calibration can be performed with simple processes, but the method requires high control and resources
Solution Approach 1:
The patent replaces the mechanical/target-based calibration system with a computational approach. Instead of physical targets and manual alignment procedures, the system uses point cloud registration algorithms and entropy minimization computations to achieve calibration, substituting mechanical complexity with information processing
Data Source
Figure 1
Figure 2A~2B
Figure 3A~3B
AI summary
Methods and systems are provided for performing radar-to-lidar calibration. In some aspects, a process can include steps for receiving, at an autonomous vehicle system, radar data from a radar of an object, receiving, at the autonomous vehicle system, lidar data from a lidar of the object, generating, by the autonomous vehicle system, a plurality of cost functions based on the radar data and the lidar data of the object, and adjusting, by the autonomous vehicle system, at least one setting based on the plurality of cost functions of the radar data and the lidar data of the object.