Kernel Machine Estimator for User-Reported Location Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The accuracy of locating objects, such as speed enforcement devices, is hindered by low resolution geographical coordinates, inaccurate user reports, and false locations in systems that rely on user-provided input from mobile devices.

Innovation Solution

A method and system that utilize machine learning to estimate object locations by collecting and processing user-reported geographical coordinates and object types at a remote server, employing a kernel machine estimator trained with data from known object locations and false locations to filter and score user reports, thereby improving the detection of object locations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If user-provided input from mobile devices is used to locate objects, then the system can collect location data from multiple sources, but the accuracy is hindered by low resolution geographical coordinates, inaccurate reports, and false locations

Engineering Contradiction:
Improvenumber of user reportsVSAvoidlocation accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

A kernel machine estimator is introduced as an intermediary between raw user reports and final location determination. This statistical model processes multiple user reports, filtering out inaccurate and false locations while identifying the true object location through pattern recognition and probability analysis.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system uses feedback from multiple user reports to continuously refine location estimates. By analyzing the distribution and consistency of reports from different users, the system adjusts its determination of object locations, giving more weight to consistent reports and filtering out outliers.

Inventive Principle:
Principle #23Feedback

2Productivity

If geographical coordinates from mobile devices are collected, then location data can be gathered from users in the vicinity, but the low resolution of coordinates limits detection precision

Engineering Contradiction:
Improvedata collection efficiencyVSAvoidcoordinate resolution
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

Multiple low-resolution coordinate reports from different users are merged and analyzed collectively. The kernel machine estimator combines these distributed, low-precision data points to infer the precise location of objects, effectively synthesizing collective imprecise data into accurate location information.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system transitions from analyzing individual 2D coordinate points to examining the distribution pattern across multiple dimensions (spatial distribution, temporal patterns, user density). This dimensional expansion allows the system to extract precise location information from otherwise low-resolution inputs.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Adaptability or versatility

If a database of object locations is maintained using user reports, then the database can be dynamically updated, but false locations and inaccurate reports reduce reliability

Engineering Contradiction:
Improvedatabase update capabilityVSAvoidlocation data trustworthiness
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The kernel machine estimator performs preliminary statistical analysis on user reports before committing location data to the database. By pre-filtering and validating reports through probability modeling, the system prevents false and inaccurate locations from being stored, ensuring only reliable data enters the database.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements continuous feedback loops where new user reports are compared against existing database entries and statistical models. This feedback mechanism identifies and corrects false locations while maintaining accurate ones, progressively improving database reliability over time through iterative validation.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2867877B1Location estimation method and system
Publication Date: 2023.03.29 TOMTOM TRAFFIC
  • EP2867877B1 patent drawingFigure 1~2
  • EP2867877B1 patent drawingFigure 3
  • EP2867877B1 patent drawingFigure 4~5

AI summary

A method of estimating the geographical location of an object from location information reported by one or more mobile devices having location determining means comprises collecting location data from the one or more mobile devices, preparing the data by selecting features of the data, training an estimator by inputting selected features of the data, and applying the estimator to a map grid to estimate the location of the object. During training of the estimator a kernel machine operates on histograms of distances that are computed with respect to a considered location and the user reports nearby. During application of the estimator, if the estimator assigns a score above a certain threshold a location is marked as the true location of the object. In case multiple locations are predicted, a subsequent clustering merges multiple predictions into one by using the weighted mean.