Road Object Localization Using Iterative Trace Recalibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing high-definition mapping methods using vehicle sensors suffer from inaccuracies due to sensor imprecision and high vehicle speeds, limiting data correction to only when reference points are detected, which are costly and have limited lifespan.
Innovation Solution
A method and device for recalibrating vehicle data using a list of calibration road objects with known locations, calculating recalibration parameters, and applying them to detected objects to improve location accuracy, even when no reference points are detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reference points with known absolute locations are used to correct sensor discrepancies, then location accuracy is improved, but the cost and complexity increase due to manual creation and limited coverage
Solution Approach 1:
The patent creates virtual reference points by detecting road objects (signs, traffic lights, etc.) and assigning them temporary reference status. These virtual references copy the function of physical reference points without requiring manual placement or permanent physical markers. The system extracts features from images of road objects and uses them as temporary calibration anchors, eliminating the need for costly manual reference point creation while maintaining location correction capability.
Solution Approach 2:
The patent employs temporary reference objects that can be easily detected and discarded. Instead of permanent physical reference points, the system uses detectable road objects that serve as references only during the calibration process. These temporary references are automatically detected by vehicle sensors, used for calibration, and then discarded, eliminating the need for expensive manual reference point installation and maintenance.
2Measurement precision
If manually created reference points are used for calibration, then location correction is possible, but coverage is limited to areas where reference points have been manually installed
Solution Approach 1:
The system enables vehicles to autonomously detect and utilize road objects as reference points without human intervention. The vehicle's sensor system automatically identifies suitable road objects (signs, traffic lights, distinctive features) and employs them for calibration purposes. This self-service approach allows any vehicle to create its own reference framework based on detected objects, dramatically expanding coverage across the entire road network without requiring manual reference point installation in every area.
Solution Approach 2:
The patent makes any detectable road object potentially serve as a reference point. Instead of requiring specific physical reference markers, the system allows any identifiable road object (signs, traffic lights, distinctive architectural features) to function as a calibration reference. This universality enables reference points to be generated from any road object detection, expanding coverage to all areas with detectable road infrastructure.
3Measurement precision
If reference points are used to correct location data, then calibration is possible, but the reference points may be moved or removed reducing their lifespan
Solution Approach 1:
The patent transitions from static physical reference points to dynamic virtual references based on detected road objects. The reference framework is continuously updated based on current detections, allowing the system to adapt to object movements, removals, or changes. When a physical reference point is moved or removed, the system can detect this through changes in detected object positions or characteristics and automatically recalibrate, maintaining location accuracy without being constrained by the physical stability of fixed reference markers.
Data Source
Figure 1a~1b
Figure 1c~1d
Figure 2
AI summary
The invention relates to a method and a device for locating road objects from a plurality of traces transmitted by at least one vehicle traveling on a road network, a trace transmitted by a vehicle comprising a plurality of successive locations of the vehicle acquired during a driving session, and at least one particular road object detected during the driving session by a sensor of the vehicle, associated with a location of the vehicle at the time of its detection. The method is noteworthy in that it comprises a step of selecting traces comprising at least one calibration road object contained in a list of calibration road objects in which a particular road object is associated with a location which is initialized with at least one road object whose real location is known, and in that, for each selected trace, at least one calibration parameter representative of an offset between a location associated with an object in the trace and a location associated with said object in the calibration list is calculated, the calculated calibration parameter is applied to the objects contained in the trace in order to obtain a calibrated trace, and the list of calibration objects is updated with the locations of road objects contained in the calibrated trace, the calculating, parameter application and update steps being repeated for as long as at least one calculated calibration parameter is greater than a particular threshold.