Road Speed Prediction via Trained Difference Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine learning models for road speed prediction face challenges due to fluctuations in road speeds and wild variations in feature values, leading to inaccurate and inefficient predictions.
Innovation Solution
A system and method that utilize a trained prediction model based on prior speed differences to estimate future road speeds by obtaining current vehicle speeds, determining current road speeds, and predicting speed differences, which reduces the complexity of feature values and increases accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning models use historical road speed data from various road sections with a large number of features, then the model can be trained to make predictions, but the prediction accuracy deteriorates due to fluctuations of road speeds and wild variations of feature values
Solution Approach 1:
The patent extracts only the necessary feature - current road speed - from the complex historical data set. Instead of using a large number of historical features from various road sections, the method focuses on extracting and utilizing the current road speed of the target road section, which significantly simplifies the feature set while improving prediction accuracy.
Solution Approach 2:
The patent transforms the prediction target from absolute road speed to speed difference (change in speed). By predicting the change in speed rather than the absolute speed value, the model becomes more robust to fluctuations and variations, improving prediction accuracy while reducing the complexity of handling diverse feature values.
2Reliability
If traditional machine learning models include a large number of features for road speed prediction, then more information is available for training, but the model efficiency deteriorates due to the complexity of processing numerous features
Solution Approach 1:
The patent extracts and utilizes only the most relevant feature - current road speed - for training the prediction model. This extraction approach maintains prediction reliability by focusing on the key determinant of future speed while dramatically improving training efficiency by eliminating the need to process numerous redundant historical features from multiple road sections.
3Quantity of substance
If traditional methods use historical road speed data from various road sections, then more data is available for training, but the prediction accuracy deteriorates due to wild variations of feature values across different road sections
Solution Approach 1:
The patent applies local quality by focusing prediction on the specific target road section rather than aggregating data from various road sections. By using only the current road speed of the target section and predicting its speed difference, the method eliminates the negative impact of wild variations across different road sections while maintaining sufficient data quantity for reliable prediction.
Data Source
AI summary
A method for speed prediction may include obtaining current vehicle speeds associated with first vehicles that pass through a target road section in a current time interval. The method may also include determining a current road speed associated with the target road section in the current time interval based on the current vehicle speeds. The method may also include determining a predicted speed difference between the current road speed and a future road speed associated with the target road section in a future time interval based on the current road speed and a trained prediction model, which is based on prior speed differences, wherein the current time interval and the future time interval are separated by a first time period. The method may also include estimating the future road speed based on the predicted speed difference and the current road speed.