Vehicle Map Positioning Using Conditional Particle Resampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining a vehicle's precise position on a map, such as those using particle filters, are unreliable when road markings are not visible and can lead to dangerous errors if they select a single incorrect particle, as they heavily depend on exteroceptive sensors and visibility conditions.
Innovation Solution
A method that distributes particles on a map based on the vehicle's geolocated position, calculates their likelihood using map data independently of exteroceptive sensor data, selects a limited set of particles, and resamples using a low-variance technique only when necessary, to ensure reliable positioning without relying on external conditions or depleting particles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a particle filter method using exteroceptive sensors is used to determine vehicle position, then positioning precision can be improved when road markings are visible, but reliability deteriorates when road markings are not visible or when a single incorrect particle is selected
Solution Approach 1:
The patent introduces map data as an intermediary element to mediate between the vehicle's geolocated position and the particle filter algorithm. Instead of relying solely on exteroceptive sensors to directly determine particle likelihood, the system uses map data (road geometry, lane markings, infrastructure positions) as a reliable reference framework. This intermediary allows the system to verify and correct particle positions without being affected by poor visibility conditions, thereby maintaining both precision and reliability.
Solution Approach 2:
The patent implements a feedback mechanism where the calculated likelihood of each particle based on map data is used to iteratively refine the particle distribution. The system continuously compares the vehicle's observed position with the map-based particle predictions, calculates likelihood ratios, and updates particle weights accordingly. This feedback loop enables the system to correct positioning errors and maintain reliability even when exteroceptive sensors provide insufficient information.
2Stability of the object's composition
If automatic resampling is performed continuously in the particle filter, then particle distribution can be maintained, but particle depletion occurs which may eliminate correct particles
Solution Approach 1:
The patent replaces continuous automatic resampling with periodic, conditional resampling triggered only when specific criteria are met. Instead of resampling at every time step, the system monitors particle distribution quality metrics and performs resampling only when degradation is detected or when the vehicle enters new map areas requiring updated particle distributions. This periodic approach maintains stability while preventing premature particle depletion.
Solution Approach 2:
The patent dynamically adjusts particle filter parameters including resampling thresholds, particle counts, and likelihood calculation weights based on vehicle speed, map data availability, and sensor confidence levels. By changing these parameters adaptively rather than using fixed automatic resampling, the system maintains particle distribution stability while preserving correct particles under varying operational conditions.
Data Source
AI summary
A method for estimating a precise position of a vehicle on a map includes acquiring at least one geolocated position of the vehicle by way of a geolocation system, pre-positioning the vehicle on the map, and particle filtering in which possible positions of the vehicle, called particles are processed. The particle filtering includes distributing particles on the map, and then updating the particles on the map, calculating the likelihood of each particle, selecting a limited set of particles, and, if an indicator relating to the likelihood of the selected particles and to the number of selected particles drops below a threshold, resampling particles on the map.