Occupancy Map Filtering for Autonomous Vehicle Localization Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately localizing themselves due to erroneous measurements from sensors, especially when driving over uneven surfaces, which increases computational intensity and risks incorrect localization.
Innovation Solution
A method for filtering inputs to a localization method of an autonomous vehicle, involving a predefined occupancy map, an optimized data structure representing occupied cells, and a process to detect objects using sensor units. Each measured value is searched within a predetermined search radius for an occupied cell; if found, the value is forwarded; if not, it is discarded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high frequency of localization methods and high granularity of sensor data are used to achieve high precision localization, then localization precision is improved, but computational intensity increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-processing sensor data to generate an optimized data structure (k-d tree) from the occupancy map before localization is performed. This pre-organization of spatial data enables efficient nearest-neighbor searches during localization, reducing the computational burden of processing high-granularity sensor data at high frequencies while maintaining localization precision.
Solution Approach 2:
The patent replaces direct comparison of raw sensor data with occupancy map patterns by substituting it with a geometric search approach using k-d trees. Instead of computationally intensive pattern matching, the system uses efficient spatial partitioning and nearest-neighbor search algorithms to identify corresponding features, dramatically reducing computational intensity while preserving measurement precision.
2Reliability
If sensor units capture detailed environment data for accurate localization, then localization reliability is improved, but erroneous measurements from uneven surfaces increase computational load and may cause incorrect localization
Solution Approach 1:
The patent applies the taking out principle by extracting and filtering only the most relevant spatial features from sensor data. The system identifies and processes only occupied cells within a defined search radius around the vehicle position, discarding irrelevant data from uneven surfaces and other noise sources. This selective extraction maintains localization reliability by focusing on meaningful features while reducing data processing complexity.
Solution Approach 2:
The patent applies local quality by making the data processing approach adaptive to local conditions. The search radius and filtering criteria are adjusted based on the local occupancy map structure and vehicle position. In areas with complex terrain or uneven surfaces, the system locally adapts its processing to focus only on relevant occupied cells, thereby maintaining reliability without uniformly increasing processing complexity across all environments.
3Measurement precision
If all measured values from sensor units are processed in the localization method, then comprehensive localization accuracy is improved, but processing time increases due to high number of data points
Solution Approach 1:
The patent applies partial action by processing only a subset of measured values that are most relevant to localization. Instead of processing all sensor data points, the system selectively processes occupied cells within a predetermined search radius of the vehicle position. This partial processing approach maintains localization accuracy by focusing on critical data while significantly reducing processing time through the efficient k-d tree search structure.
Data Source
AI summary
The present invention relates to a method for filtering inputs to a localization method of an autonomous vehicle (10) in an operating environment (U), comprising the steps of: providing a predefined occupancy map which represents known objects (W) present in the operating environment (U); generating an optimized data structure which represents occupied cells in a display of the occupancy map in a coordinate system; detecting the operating environment (U) of the vehicle (10) by means of at least one sensor unit (12) which is configured to detect objects in the operating environment (U) and their distances from the vehicle (10), wherein each detected object is assigned a measured value (M1, M2) in the coordinate system; and for each measured value (M1, M2) searching within a search radius around the measured value (M1, M2) for an occupied cell in the optimized data structure; if an occupied cell is found within the search radius, forwarding the measured value (M1) to a subsequent localization method, and, if no occupied cell is found within the search radius, discarding the measured value (M2).