Dynamic Travel Time Model Using ML Positional Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional navigation systems face challenges in providing accurate and dynamic estimated travel times, as they often deviate significantly from actual times due to unexpected changes in travel and traffic behavior, and require substantial development and testing time for updates.
Innovation Solution
A method using machine learning algorithms trained with historic positional data from devices traveling within a navigable network to generate and update models for determining estimated travel times, allowing for continuous refinement and adaptation to changes in network conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional navigation systems use fixed routing algorithms and static traffic data, then the system complexity is low and ease of operation is high, but the accuracy of estimated travel times deteriorates due to inability to adapt to changing traffic conditions
Solution Approach 1:
The patent applies dynamics by transitioning from static routing algorithms to dynamic machine learning models that continuously learn from new positional data. The system dynamically adapts its estimation methodology based on actual travel patterns, traffic conditions, and route characteristics, allowing the model to improve accuracy over time while responding to changing network conditions.
Solution Approach 2:
The patent implements feedback mechanisms by using actual travel time data from positional traces to train and refine the machine learning model. The system compares predicted travel times with actual recorded times, uses the discrepancies to update model parameters, and continuously improves its estimation accuracy through this closed-loop feedback process.
2Adaptability or versatility
If the navigation system continuously updates its models with new positional data, then the accuracy and adaptability improve, but the processing time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-processing and storing positional data in a structured format before it is needed for model training. The system maintains a database of historical positional traces and extracts relevant features in advance, so when new data arrives, it can be quickly integrated without requiring complete reprocessing of existing data.
Solution Approach 2:
The patent uses partial action by selectively training the machine learning model only on the most relevant and recent positional data rather than reprocessing all historical data. The system identifies and processes only the portions of data that are most indicative of current network conditions, reducing computational overhead while maintaining adaptability.
3Measurement precision
If the system uses machine learning algorithms to determine estimated travel times, then the accuracy improves through continuous learning, but the device complexity and difficulty of detecting changes increase
Solution Approach 1:
The patent implements feedback by continuously monitoring model performance against actual travel data and using this information to detect when pattern-breaking events occur. When the model's predictions systematically deviate from actual travel times beyond expected variance, the feedback mechanism triggers an update or alert, making change detection automated and continuous.
Solution Approach 2:
The patent replaces traditional mechanical detection methods with machine learning-based anomaly detection. Instead of using fixed thresholds or manual analysis to detect pattern-breaking events, the system uses trained neural networks and statistical models to automatically identify deviations from normal travel patterns, improving detection capability while managing complexity through automation.
Data Source
AI summary
Disclosed herein is a method for generating a model that can be used by a routing module to determine estimated travel times for routes within a navigable network. The model is generated by a process of machine learning with training data that is obtained from historic positional data recorded by devices travelling within the navigable network. The model can be updated over time as new positional data is obtained. The newly obtained positional data is also used for evaluating the performance of the model and for detecting so-called ‘pattern-breaking’ events where the underlying conditions in the navigable network suddenly change such that the previous version of the model may no longer give accurate estimated travel times.