Automated Vehicle Map Alignment Using Multi-Object Position Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated vehicles face challenges in aligning map data from multiple sources, which are often collected in different reference frames, leading to potential misalignment relative to world GPS coordinates, and existing methods may not effectively handle rotational differences or errors in object positioning.
Innovation Solution
A navigation system that uses on-board sensors and perception sensors to determine the location of a host-vehicle on multiple digital maps by identifying common objects, calculating coordinates with statistical errors, and aligning maps only when differences are within a specified error threshold, using multiple objects for accurate alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If map data from multiple sources are aligned using a single object reference, then the alignment process is simplified, but the accuracy and reliability of the alignment deteriorates due to insufficient constraint conditions
Solution Approach 1:
The alignment process is segmented into multiple stages: first aligning maps using a first object (e.g., lane marker), then verifying and refining the alignment using a second object (e.g., street sign). This segmentation allows the system to maintain simplicity while improving accuracy through multi-object constraint satisfaction.
Solution Approach 2:
The system merges multiple reference objects (lane markers, street signs, barriers) into a unified alignment framework. By combining constraints from multiple objects rather than relying on a single object, the system achieves higher alignment accuracy while maintaining a systematic approach to the alignment process.
2Device complexity
If map data are aligned offline to create a single source failure component, then system complexity is reduced, but the adaptability to handle rotational differences and positioning errors deteriorates
Solution Approach 1:
The system transitions from static offline alignment to dynamic online alignment verification. While the initial alignment can be performed offline, the system dynamically adjusts and verifies alignment using multiple objects detected during vehicle operation, allowing it to adapt to rotational differences and positioning errors that were not present in the offline alignment process.
Solution Approach 2:
The system implements feedback mechanisms where alignment accuracy is continuously monitored using multiple reference objects. The statistical analysis of positioning errors provides feedback that allows the system to detect and correct misalignments, enhancing adaptability while maintaining a relatively simple single-source architecture.
3Loss of time
If the system aligns maps using only one object, then processing time is reduced, but the reliability of alignment deteriorates due to potential errors in single-object positioning
Solution Approach 1:
The system performs partial alignment using the first object to achieve quick initial positioning, then applies excessive action by introducing a second object for verification and refinement. This approach balances processing time and reliability by not requiring full multi-object alignment for every adjustment, while still using multiple objects to ensure accuracy.
Solution Approach 2:
The second object acts as an intermediary verification mechanism. Rather than directly relying on single-object alignment, the system uses the second object as a mediator to verify and correct the alignment established by the first object, thereby improving reliability without significantly increasing processing time.
4Measurement precision
If the system uses multiple objects for map alignment, then alignment accuracy is improved, but the complexity of object detection and coordinate calculation increases
Solution Approach 1:
The system applies local quality by focusing detection and calculation efforts on specific local features of multiple objects rather than analyzing entire scenes. By identifying and using only the critical geometric features of each object (e.g., lane marker positions, street sign locations), the system achieves high alignment precision while keeping the detection and calculation complexity manageable.
Data Source
Figure 1
Figure 2
AI summary
A navigation system (10) for an automated vehicle includes an object-detector (18), a first-map (24), a second-map (30), and a controller (36). The object-detector (18) indicates relative-positions (20) of a plurality of objects (22) proximate to the host-vehicle (12). The first-map (24) indicates a first-object (26) and a second-object (28) detected by the object-detector (18). The second-map (30) is different from the first-map (24). The second-map (30) indicates the first-object (26) and the second-object (28). The controller (36) is in communication with the object-detector (18), the first-map (24), and the second-map (30). The controller (36) is configured to determine a first-coordinate (40) of the host-vehicle (12) on the first-map (24) based on the relative-positions (20) of the first-object (26) and the second-object (28), determine a second-coordinate (50) of the host-vehicle (12) on the second-map (30) based on the relative-positions (20) of the first-object (26) and the second-object (28), and align the first-map (24) and the second-map (30) based on the first-coordinate (40), the second-coordinate (50), and the relative-positions (20) of the first-object (26) and the second-object (28).