Occupancy Grid Reference Mapping With Confidence-Based Cell States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automotive localization systems face challenges in generating and maintaining accurate reference maps for large geographical areas due to processing demands and sensor limitations, leading to inefficient data storage and noise in sensor data.
Innovation Solution
A method for building a reference map by accumulating perception sensor data, generating an occupancy grid with confidence characteristics, and recording reference map cells based on these characteristics to filter out noise and minimize data redundancy, suitable for radar data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If occupancy grid mapping is used to generate localization maps from sensor data, then localization accuracy is improved, but processing demands and memory requirements increase significantly
Solution Approach 1:
The patent extracts only the most relevant information from the occupancy grid based on confidence characteristics. Instead of processing and storing the complete occupancy grid data, the system identifies and extracts high-confidence cell states that represent significant environmental features, thereby reducing processing demands while maintaining localization accuracy.
Solution Approach 2:
The patent changes the parameter representation from complete occupancy probabilities to discrete confidence-based cell states. By transforming continuous sensor data into discrete high-confidence state representations, the system reduces computational complexity and memory requirements while preserving essential localization information.
2Area of stationary object
If reference maps cover large geographical areas, then the coverage area increases, but data size and storage requirements increase
Solution Approach 1:
The patent extracts only high-confidence cell states from the occupancy grid to build reference maps. By selecting and recording only those cells that meet confidence criteria, the system maintains comprehensive geographical coverage while minimizing data size by excluding low-confidence or redundant information.
Solution Approach 2:
The patent applies partial action by recording only a subset of occupancy grid cells—specifically those with high confidence characteristics—rather than recording all cells. This selective approach ensures adequate coverage for localization while reducing overall data volume.
3Loss of information
If all sensor data is recorded in the reference map, then measurement completeness is improved, but noise and data redundancy increase
Solution Approach 1:
The patent extracts only high-confidence cell states from the sensor data before recording them in the reference map. This extraction process naturally filters out noise and low-confidence measurements, preserving essential information while eliminating harmful data quality issues.
Solution Approach 2:
The patent converts the uncertainty and noise inherent in sensor data into a beneficial filtering mechanism. By using confidence characteristics derived from uncertain sensor measurements, the system identifies and records only reliable data points, transforming data quality problems into a selective recording advantage.
Data Source
AI summary
Disclosed is a method for building a reference map. The method includes accumulating perception sensor data for a vehicle moving through an environment. The method includes generating an occupancy grid from the accumulated sensor data as the vehicle moves. The cells in the occupancy grid have a status and a confidence characteristic associated with the respective cell status determined based on the accumulated data. The method includes building a reference map for the environment by recording reference map cell states based on the confidence characteristics of respective cell statuses in the occupancy grid. In response to the recording reference map cell states, each reference map cell is assigned a single state based on the confidence characteristics associated with the respective cell. The step of building the reference map further includes recording a weight associated with each assigned reference map cell state.