Travel Speed Prediction Using Machine Learning Spatial Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Navigation services face inaccuracies in travel speed predictions due to noise in locational information, natural variable speeds, and sudden traffic changes, leading to incorrect route suggestions and estimated times of arrival.
Innovation Solution
A system utilizing machine learning models, such as convolutional neural networks, to predict travel speeds by processing locational information, vehicle trajectories, weather data, incident data, and sensor data, while identifying spatial contexts and formatting features for input into models to generate accurate speed predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If heuristics and locational information are used to estimate travel speeds, then route suggestions and estimated times of arrival can be provided, but noise in locational information and natural variable speeds cause large fluctuations in predicted speeds
Solution Approach 1:
The system segments the travel network into multiple segments and processes locational information at different granularities. By dividing the road network into segments and analyzing speed patterns segment-by-segment, the system reduces noise impact while maintaining prediction capability. This segmentation allows the system to identify consistent speed patterns even when individual data points are noisy.
Solution Approach 2:
The system performs preliminary processing of locational information by computing historical average speeds and identifying speed patterns before making real-time predictions. This preliminary action includes pre-processing GPS data, computing statistical measures, and establishing baseline speed profiles for different road segments, which helps filter out noise before final speed estimation.
2Ease of operation
If historical average speeds are used when data points are sparse, then predictions can be made with low data density, but large errors occur due to reliance on historical averages under normal conditions
Solution Approach 1:
The system dynamically changes parameters based on data availability. When data points are sparse, it adjusts by incorporating historical average speeds with appropriate weighting. The system modifies its prediction parameters adaptively, switching between relying more on real-time data versus historical averages based on the density and quality of available locational information, thereby maintaining accuracy across different data conditions.
Solution Approach 2:
The system implements feedback mechanisms where prediction accuracy is continuously monitored and used to adjust future predictions. When historical averages produce large errors, the system learns from this feedback and adjusts its reliance on historical data versus real-time observations, improving its ability to handle sparse data conditions while maintaining precision.
3Reliability
If real-time locational information is processed to identify sudden traffic changes, then current traffic conditions can be reflected, but latency occurs until predicted speeds finally reflect the sudden traffic changes
Solution Approach 1:
The system performs preliminary monitoring and analysis of locational information streams to detect potential traffic changes before they fully manifest in speed measurements. By pre-processing GPS data and identifying anomalies in real-time, the system can alert to upcoming traffic conditions earlier, reducing the latency between actual traffic changes and their reflection in predicted speeds.
Solution Approach 2:
When sudden traffic changes are detected through preliminary monitoring, the system rushes through the normal processing pipeline by bypassing certain computational steps or giving priority processing to the affected segments. This allows the system to quickly update speed predictions for segments experiencing sudden traffic changes, reducing the time lag between detection and prediction update.
Data Source
AI summary
One or more techniques and/or systems are provided for travel speed prediction. A spatial context of a prediction segment of a travel network for which a speed prediction is to be made is identified. The spatial context comprises one or more segments of the travel network that are part of trajectories of objects passing through the predication segment and that have predicted likelihoods of influencing travel speed along the prediction segment above a threshold. Features of the spatial context are formatted into a format compatible for input into the model based upon a structure of the model. The features are input into the model for processing using machine learning functionality to output the speed prediction for the prediction segment.