Vehicle Initial Pose Localization Along Road Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining a vehicle's initial pose require high computing power due to multidimensional hill-climbing algorithms, especially in GNSS-based localization, which lacks accuracy and availability.
Innovation Solution
A method using a control device to receive data from GNSS and odometry sensors, extract trajectories from historical data, and perform an optimization algorithm along these trajectories to refine the initial pose, utilizing radar, LIDAR, or camera sensors to compare statistical features with a feature map, reducing the problem to a one-dimensional optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a multidimensional hill-climbing algorithm is used for feature-based localization, then the pose determination accuracy is improved, but the computing power requirement increases
Solution Approach 1:
The patent segments the localization problem into two stages: first determining an approximate pose using GNSS and odometry, then performing feature-based optimization only within a reduced search space around this approximate pose. This segmentation allows the computationally intensive hill-climbing algorithm to operate in a constrained one-dimensional space rather than full multidimensional space, thereby reducing computing power requirements while maintaining accuracy.
Solution Approach 2:
The patent transforms the multidimensional optimization problem into a one-dimensional problem by leveraging the approximate pose from GNSS/odometry to define a restricted search trajectory. Instead of searching across all possible pose dimensions simultaneously, the algorithm follows a predetermined trajectory (e.g., along the GNSS-derived path) and optimizes only the positional parameter along this trajectory, effectively reducing the dimensionality of the optimization space.
2Measurement precision
If feature-based localization is performed without an approximate initial position, then the localization accuracy is improved, but the computing power and time required increase significantly
Solution Approach 1:
The patent performs preliminary action by using GNSS and odometry sensors to determine an approximate pose before initiating the computationally intensive feature-based localization. This preliminary estimation provides a starting point that constrains the subsequent optimization search, allowing the algorithm to converge faster to the accurate pose without having to search the entire configuration space.
Solution Approach 2:
The localization process is segmented into an initial approximate positioning phase (using low-power GNSS/odometry) followed by a refinement phase (using feature-based optimization). This segmentation allows the system to spend minimal time on the粗定位 phase and then focus computational resources efficiently on the refinement phase within a constrained search space, reducing overall computation time.
3Ease of manufacture
If GNSS-based localization is used, then the implementation simplicity is improved, but the availability and accuracy are reduced
Solution Approach 1:
The patent merges GNSS-based localization with feature-based localization in a hybrid approach. The GNSS provides a simple approximate pose estimation, while the feature-based method refines this estimate using sensor data and map features. This combination maintains the implementation simplicity of GNSS while compensating for its accuracy limitations through the addition of feature matching, achieving both ease of implementation and high accuracy.
Data Source
AI summary
A method for ascertaining an initial pose of a vehicle using a control device. Measured data ascertained by a GNSS sensor system and/or an odometry sensor system are received and evaluated to ascertain an approximate pose of the vehicle with a margin of uncertainty. At least one trajectory of road users is extracted from a trajectory map for the ascertained margin of uncertainty. Test points are positioned along the extracted trajectory. An optimization algorithm is performed for each test point along the trajectory. The optimization algorithm ascertains poses having corresponding cost functions. A pose having the greatest cost function is determined as the initial pose of the vehicle from the poses ascertained by the optimization algorithm. A control device, a computer program, and a machine-readable storage medium are also provided.