Dynamic Lane Grid Resolution for Curved-Road ADAS Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional grid systems used in advanced driver assistance systems (ADAS) for monitoring the environment in front of a host vehicle are inefficient due to a fixed number of cells, leading to unnecessary data processing and memory usage, especially in scenarios with lane curvature, where the resolution needs to be dynamically adjusted based on the environment and driving situation.
Innovation Solution
A dynamic grid generation method that detects the lane course using sensors like visual, radar, or Lidar systems, and adjusts the grid resolution by splitting or merging cells based on qualifying data such as obstacle distances and classifications, focusing on areas of interest like the lane and its surroundings, thereby reducing unnecessary data and improving processor performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional square or rectangular grid with the same number of cells in both lateral and longitudinal directions is used to guarantee collision-free space, then safety is improved, but memory usage and processing load increase significantly
Solution Approach 1:
The grid is segmented into different regions with different cell densities. High-resolution regions are created near the host vehicle and lane boundaries where collision risk is higher, while low-resolution regions are used in areas farther away where fewer cells are needed to maintain safety
Solution Approach 2:
Different areas of the grid are assigned different qualities (cell sizes and resolutions) based on their importance. Areas closer to the vehicle and lane boundaries have smaller cells for higher resolution, while distant areas have larger cells for lower resolution, optimizing the balance between safety and computational efficiency
2Reliability
If the number of grid cells is increased in lateral direction to handle strong lane curvature, then collision-free space monitoring is improved, but memory usage and processing time increase
Solution Approach 1:
The grid structure is made dynamic and adaptive to the lane geometry. When strong curvature is detected, the grid automatically adjusts by increasing cell density in the affected lateral regions rather than uniformly increasing all cells, reducing processing time while maintaining monitoring capability
3Device complexity
If a conventional grid includes the same number of cells in both directions, then grid simplicity is maintained, but resolution adaptability to different driving situations is reduced
Solution Approach 1:
The uniform grid is segmented into zones with different resolutions. The grid can be divided into high-resolution zones near the vehicle and low-resolution zones farther away, allowing adaptability to different driving situations without overwhelming complexity
Solution Approach 2:
Grid parameters such as cell size and density are changed dynamically based on driving conditions. The system can adjust resolution parameters in real-time to match the current driving situation, whether urban or highway environments
4Measurement precision
If grid resolution is increased in certain areas to improve assistance system performance, then detection precision is improved, but memory usage increases
Solution Approach 1:
High measurement precision is applied locally only where needed for object detection and collision avoidance, rather than uniformly across the entire grid. This concentrates computational resources on critical areas while reducing memory usage in less critical regions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method is provided for generating a dynamic grid. An indicator is detected for the course of a lane of a road, and a base area is determined based on the detected indicator, wherein the base area is restricted to a region of interest. A plurality of cells is defined by dividing the base area in order to form the dynamic grid. For each of the plurality of cells, qualifying data is received, and based on the qualifying data a valuation is determined for each cell. It is decided for each cell based on the valuation whether the respective cell is to be split or to be merged with an adjacent cell, and based on the decision, either the respective cell is modified by splitting the cell or by merging the cell with at least one adjacent cell, or the respective cell is left unmodified.