Vehicle Location Detection Using Map-Aware Kalman Filter Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle systems face challenges in achieving high-precision vehicle location detection, particularly in environments where GNSS signals are disrupted, such as tunnels or urban canyons, due to the dynamic weighting of input variables within the Kalman filter not adequately accounting for predictive scenarios.
Innovation Solution
The method involves using map data to determine predictive weighting factors for input variables, such as GNSS data and inertial sensor data, within a Kalman filter to adjust the influence of GNSS data based on expected surroundings, such as street canyons or tunnels, thereby improving location detection precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dynamic weighting of input variables is used in the Kalman filter, then location calculation precision is improved in general conditions, but reliability deteriorates in environments where GNSS signals are disrupted (tunnels, urban canyons)
Solution Approach 1:
The system performs preliminary analysis of map data to predict future GNSS signal availability before the vehicle actually enters problematic areas. By anticipating upcoming tunnels or urban canyons using pre-stored map information about building locations and tunnel entrances, the system proactively adjusts weighting factors in advance, ensuring reliable location calculation continues without interruption when GNSS signals are disrupted.
Solution Approach 2:
The weighting factors for different input variables are made dynamically adjustable based on predicted environmental conditions. The system continuously monitors the vehicle's trajectory against map data and automatically transitions between different weighting configurations - using GNSS-heavy weighting when signals are available and inertial-sensor-heavy weighting when signals are predicted to be disrupted, thereby maintaining both precision and reliability across varying conditions.
2Measurement precision
If GNSS data are given high weighting, then location precision is improved in open areas, but reliability deteriorates in street canyons and tunnels where signals are blocked
Solution Approach 1:
The system uses pre-stored map data containing information about tunnels, urban canyons, and other signal-blocking structures to predict future GNSS signal availability. Before the vehicle enters such areas, the system proactively reduces the weighting of GNSS data and increases reliance on inertial sensors, ensuring continuous reliable location calculation without waiting for signal loss to occur.
Solution Approach 2:
The weighting parameters for different input variables are dynamically changed based on the predicted environmental context. When map data indicates the vehicle is approaching areas with poor GNSS reception, the system automatically adjusts the weighting factors to reduce GNSS data influence and increase inertial sensor influence, thereby maintaining reliability across different operational environments.
3Reliability
If inertial sensor data are given high weighting, then reliability is improved in signal-disrupted environments, but measurement precision deteriorates in open areas with good GNSS coverage
Solution Approach 1:
The system dynamically adjusts the weighting of inertial sensor data based on the operational environment detected through map data analysis. In open areas with good GNSS coverage, the weighting of inertial sensors is reduced to avoid degrading overall precision. In signal-disrupted environments, the weighting is increased to maintain reliability, achieving optimal performance across different conditions through adaptive parameter adjustment.
Data Source
AI summary
A method for satellite-based detection of vehicle location uses a motion and location sensor. GNSS data is received as an input variable, at least one further input variable is also received. Weighting factors for the input variable and the at least one further input variable are determined. The input variable and the at least one further input variable are weighted by the weighing factors. The vehicle location is detected via the weighted input variable and the weighted at least one further input variable.