Autonomous Driving Object Recognition on Gradient Roads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems face challenges in accurately determining altitude information and recognizing objects on roads with gradients, particularly in complex environments like city scenarios, leading to frequent object misrecognition and non-recognition due to occlusions and irregular road surfaces.
Innovation Solution
An autonomous driving control apparatus that utilizes sensor fusion with high-definition maps to determine altitude information and valid contour regions based on vehicle gradients, employing a valid contour verification algorithm to filter objects within a valid height range, ensuring accurate object recognition and minimization of misrecognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor fusion with high definition map is used to determine altitude information, then object recognition accuracy on flat roads is improved, but object misrecognition occurs on roads with gradients due to limited altitude correction capability
Solution Approach 1:
The patent divides the road surface into multiple lane links with individual altitude information. By segmenting the road into discrete sections (lane links) and determining altitude for each segment separately, the system can accurately track altitude changes on gradient roads. This segmentation allows the system to handle complex road geometries by processing each lane link independently, thereby resolving the contradiction between measurement precision and reliability on gradient roads.
Solution Approach 2:
The patent introduces a vertical dimension (altitude) to the traditional 2D lane link representation by creating 3D lane links with altitude information. This dimensional enhancement allows the system to accurately represent and process road surfaces with gradients, enabling reliable object recognition by considering the third dimension (z-axis altitude) in addition to the horizontal plane coordinates.
2Ease of manufacture
If conventional road surface detection methods are used, then detection process is simple, but detection performance deteriorates on slope road surfaces which are irregular in longitudinal and lateral directions
Solution Approach 1:
The patent employs dynamic road surface modeling by continuously updating lane link information including altitude data as the vehicle moves. Instead of using static detection thresholds, the system dynamically adjusts the road surface model based on accumulated sensor data and high definition map information, enabling accurate detection on irregular slope road surfaces while maintaining computational efficiency through incremental updates.
Solution Approach 2:
The patent introduces an intermediary processing layer that fuses sensor data with high definition map information to create a refined road surface model. This intermediary representation (lane link with altitude information) acts as a mediator between raw sensor data and object detection, enabling accurate road surface detection on gradient roads by providing a corrected reference framework that accounts for longitudinal and lateral irregularities.
3Measurement precision
If altitude correction is applied in sensor fusion step, then altitude accuracy is improved, but correction is limited and object misrecognition still occurs in hill sections and tunnel entry and exit
Solution Approach 1:
The patent performs preliminary altitude correction by pre-processing lane link information to include accurate altitude data before object detection occurs. By establishing the corrected 3D road surface model in advance using high definition map information and sensor fusion, the system prepares the accurate geometric framework beforehand, enabling reliable object recognition in challenging gradient areas such as hill sections and tunnel entry and exit where altitude changes are significant.
Solution Approach 2:
The patent replaces the traditional mechanical sensor-based altitude measurement with an information-processing approach that uses high definition map data and computational geometry. Instead of relying solely on physical sensors for altitude correction, the system substitutes a computational model that calculates and applies altitude corrections based on lane link information, achieving more accurate and reliable altitude compensation in gradient areas.
4Adaptability or versatility
If multiple detection methods are used to handle various road surfaces, then detection coverage is improved, but system complexity increases and underfitting problems occur
Solution Approach 1:
The patent creates a universal lane link-based detection framework that can handle various road surface types (flat, gradient, irregular) through a single unified approach. By representing all road surfaces as 3D lane links with altitude information, the system achieves multi-functionality where the same core algorithm can process diverse road geometries without requiring separate specialized methods, thereby reducing system complexity while maintaining broad adaptability.
Data Source
AI summary
An autonomous driving control apparatus for determining an object using high definition map and sensor fusion on a road with a gradient and a method thereof are provided. The autonomous driving control apparatus includes a sensor that obtain information about an object around a vehicle, a communication device that receives information about a high definition map around the vehicle, and a processor electrically connected with the sensor and the communication device The processor navigates a vehicle lane link and a target lane link based on the information about the object and the information about the high definition map, determines a valid contour region by applying a valid contour verification algorithm to the vehicle lane link and the target lane link, and recognizes an object in the valid contour region.