Occupancy Grid Mapping for Shadowed Drivable Space Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing occupancy mapping systems for autonomous vehicles fail to accurately account for sensor limitations and environmental conditions, leading to errors in detecting drivable space due to shadowing effects and sensor noise, which can impact the vehicle's ability to navigate safely.
Innovation Solution
A device within the vehicle learns grid locations of stationary objects in shadowed areas, adjusting occupancy probabilities based on sensor data to improve accuracy and reduce latency by preventing premature reduction of occupancy probabilities in shadowed cells, thereby enhancing the detection of drivable space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If occupancy probability is reduced over time for cells not detected by sensor, then system responds to changes faster, but detection accuracy decreases in shadowed areas
Solution Approach 1:
The patent applies different occupancy probability reduction rates based on local characteristics of each cell. Cells identified as shadowed areas (where objects may occlude sensor views) maintain higher occupancy probabilities longer, while non-shadowed cells use standard reduction rates. This localized differentiation resolves the contradiction by preserving detection accuracy in shadowed regions while maintaining responsive updates in visible regions.
Solution Approach 2:
The system preliminarily identifies shadowed areas using map data and object detection before applying occupancy probability updates. By pre-characterizing which cells are likely to be shadowed based on environmental models and detected objects, the system can apply appropriate reduction rates in advance, ensuring both rapid response and accurate detection are maintained where needed.
2Device complexity
If standard occupancy probability reduction is applied to all cells, then computational processing is simplified, but navigation safety decreases due to false negatives in shadowed areas
Solution Approach 1:
The patent segments the occupancy grid into different types of cells: shadowed cells and non-shadowed cells. This segmentation allows the system to apply different update rules to different segments, improving navigation safety in shadowed areas without excessively complicating the overall computational process. The segmentation is based on map data and object detection results, making it computationally feasible.
Solution Approach 2:
The system changes the occupancy probability reduction parameter dynamically based on cell type. Shadowed cells use a lower reduction rate (preserving occupancy longer), while non-shadowed cells use a higher reduction rate (faster updates). This parameter differentiation improves navigation safety without requiring complete redesign of the computational framework, thus balancing reliability and complexity.
3Measurement precision
If occupancy probability is maintained high in shadowed cells, then detection accuracy improves, but system responsiveness to real changes decreases
Solution Approach 1:
The patent implements local quality by applying different temporal decay characteristics to different spatial regions. Shadowed cells maintain higher occupancy probabilities for longer durations to account for potential occluded objects, while visible cells update more rapidly. This localized temporal differentiation resolves the contradiction between detection accuracy and responsiveness by matching update behavior to local environmental characteristics.
Solution Approach 2:
The system dynamically adjusts occupancy probability reduction rates based on real-time object detection and shadow identification. When objects are detected that may create shadowed areas, the system dynamically lowers reduction rates for affected cells. This dynamic adaptation allows the system to maintain accuracy when needed while remaining responsive when conditions change, reducing unnecessary latency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In some aspects, a device may receive point data associated with a cell of an occupancy grid for controlling a vehicle. The device may determine, based on the point data, a characteristic of the cell that is associated with an occupancy probability of the cell, wherein the occupancy probability is determined according to a first technique based on the point data. The device may configure, based on the characteristic, the occupancy probability for the cell, within the occupancy grid, according to a second technique. Numerous other aspects are described.