Road Boundary Recognition Using Abnormal 3D Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving vehicles face challenges in accurately recognizing road boundaries at long distances due to the limitations of Lidar sensors, leading to recognition failures or misrecognition errors, which affects the stability of driving behavior and user experience.
Innovation Solution
A method that utilizes abnormal detection data from a laser scanning-based three-dimensional recognition sensor by fusing it with data from other sensors, such as cameras and radar sensors, to recognize road boundary objects. This method involves filtering out abnormal data associated with internal static and dynamic objects and using distribution information of normal boundary data to validate candidate detection data as road boundary data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Lidar sensor is used to recognize road boundary at long distance, then measurement precision is improved, but reliability deteriorates due to recognition failures and misrecognition errors
Solution Approach 1:
The patent combines multiple sensor types (Lidar, camera, radar) to process detection data collectively. By merging the strengths of different sensors, the system achieves both high measurement precision for long-distance road boundary detection and maintains reliability through cross-validation of detection results from multiple sources.
Solution Approach 2:
The patent introduces an abnormal detection data processing mechanism that acts as an intermediary between raw Lidar data and final recognition results. This intermediary process identifies and corrects abnormal detections, allowing the system to utilize long-distance Lidar data while filtering out unreliable measurements.
2Reliability
If only short-distance Lidar data is used for vehicle location estimation, then reliability is improved, but productivity deteriorates due to limited perception range
Solution Approach 1:
The patent merges short-distance and long-distance detection data from Lidar with data from other sensors. This combination allows the system to maintain high reliability by weighting short-distance data appropriately while incorporating long-distance data to expand the effective perception range and improve overall processing efficiency.
Solution Approach 2:
The patent applies partial action by selectively processing Lidar data based on distance thresholds and detection quality. Rather than discarding all long-distance data, the system processes only those portions that meet reliability criteria, thereby improving productivity without compromising the reliability provided by short-distance data.
3Measurement precision
If abnormal detection data is discarded, then measurement precision is improved by eliminating errors, but loss of information increases due to potentially valid long-distance data being lost
Solution Approach 1:
The patent introduces an abnormal detection analysis mechanism as an intermediary that examines each abnormal detection case before discarding it. This intermediary process distinguishes between truly erroneous data and valid long-distance detections that appear abnormal, thereby maintaining measurement precision while minimizing information loss.
Solution Approach 2:
The system implements feedback by continuously monitoring detection quality metrics and adjusting the threshold for discarding abnormal data. When long-distance detections consistently show valid patterns, the system adapts to retain more of this data, thereby reducing information loss while maintaining precision through ongoing validation.
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 enhances the utilization of detection data from three-dimensional recognition sensors, improves the accuracy of road boundary recognition, and ensures safer and more stable autonomous driving, thereby enhancing user experience.
Implementation Method 1
A Lidar sensor provides a precise three-dimensional point cloud for a neighbor object by using ambient scanning based on a multi-channel laser
Implementation Method 2
a Lidar sensor for obtaining a three-dimensional shape of a neighbor object
Data Source
AI summary
A method for recognizing a boundary object of a road includes sensing an environment around a vehicle by an environment perception sensor that includes a laser scanning-based three-dimensional recognition sensor and a sensor of a different type. The method also includes determining candidate detection data from abnormal detection data that is perceived to be abnormal by the three-dimensional recognition sensor and is unassociated with an internal static object of a road and an internal dynamic object of the road, by referring to a boundary region. The method also includes adopting estimated boundary data that is considered as a road boundary object in the candidate detection data, based on distribution information of normal boundary data that is normally perceived in relation to the road boundary object by the three-dimensional recognition sensor. The method also includes employing the normal boundary data and the estimated boundary data as boundary data.