Factor Graph Vehicle Positioning via Sliding Window Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle positioning methods in automatic driving rely heavily on single-frame data, which limits positioning accuracy and computational efficiency, especially when integrating data from various sensors like GNSS, IMU, and wheel speed meters.
Innovation Solution
A vehicle positioning method utilizing a sliding window to collect sensor data from multiple frames, creating a target function for a factor graph model that optimizes pose and position determination through sensor fusion, including GNSS, lane line, wheel speed meter, and IMU data, with constraints like pre-integration and marginalization to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-frame data is used for vehicle positioning, then computational efficiency is maintained, but positioning accuracy is limited
Solution Approach 1:
The patent divides the positioning problem into multiple frames (N frames) and processes them segment by segment using a sliding window approach. Each frame is processed independently through the factor graph model, allowing the system to accumulate positioning information from multiple segments while maintaining manageable computational complexity through incremental optimization.
Solution Approach 2:
The patent merges data from multiple frames (N frames) into a unified factor graph model that integrates sensor data, lane line data, and positioning information across time. This combining of multiple data sources and time points resolves the contradiction by achieving higher positioning accuracy through data fusion while the factor graph's optimization structure maintains computational efficiency.
2Measurement precision
If multiple frames are processed through factor graph optimization, then positioning accuracy is improved, but computational burden increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing sensor data and lane line data before inserting them into the factor graph model. Constraints such as pre-integration and marginalization are established in advance, which simplifies the optimization process and reduces the computational burden when processing multiple frames, while still achieving improved positioning accuracy.
Solution Approach 2:
The patent employs a dynamic sliding window that adapts the number of frames (N) and adjusts the factor graph model dynamically. This allows the system to process multiple frames for higher accuracy when needed, while reducing computational burden by adjusting the window size or skipping frames in less critical situations, thus resolving the contradiction between accuracy and complexity.
3Measurement precision
If sensor data from multiple sources is integrated, then positioning precision is enhanced, but system complexity increases
Solution Approach 1:
The patent creates a universal factor graph model that can handle multiple data sources (sensor data, lane line data, positioning data) through a unified framework. This multi-functional model integrates diverse inputs using consistent mathematical constraints, enhancing positioning precision while avoiding the need for separate processing systems for each data type, thus managing system complexity.
Solution Approach 2:
The factor graph model acts as an intermediary that mediates between multiple data sources and the final positioning result. It provides a standardized interface for integrating sensor data, lane line data, and other inputs through defined constraints and optimization processes, thereby enhancing positioning precision while simplifying the overall system architecture by centralizing the integration logic.
Data Source
AI summary
The present disclosure provides a vehicle positioning method implemented by an electronic device, including: obtaining sensor data about N frames of a vehicle through a sliding window, N being a positive integer; creating a target function for a factor graph model in accordance with the sensor data about the N frames, and performing optimization solution on the target function; and determining a pose and a position of the vehicle in accordance with a target value obtained through the optimization solution on the target function.