Fixed-Route Bus Speed Prediction With Dwell and Turn Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting bus running times fail to accurately account for bus stop durations and unique factors affecting bus speeds, leading to inefficiencies in schedule generation and budgeting for bus routes.
Innovation Solution
A system utilizing machine learning models to predict bus speeds and trip durations by filtering historical location data, distinguishing between bus dwelling and turning, and incorporating features like edge characteristics, bus stops, and intersections, enabling real-time route adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods (querying car-driving durations from online map tools) are used to estimate bus running times, then the process is simple and quick, but the accuracy is poor because it fails to account for bus stop durations and unique factors affecting bus speeds
Solution Approach 1:
The patent segments the bus route into multiple edges (road segments) and processes each edge separately through the machine learning model. This allows the system to account for different speed factors on different segments while maintaining manageable computational complexity. The route is divided into discrete segments that can be individually analyzed and summed to get the total predicted running time.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the raw route data and the final running time prediction. This intermediary component processes the route features (edges, bus stops, intersections) and transforms them into accurate speed predictions, resolving the contradiction between simplicity and accuracy by adding a specialized processing layer.
2Measurement precision
If machine learning models with multiple features (edges, bus stops, intersections) are used to predict bus speeds, then the prediction accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by pre-processing the route data into edges and features before feeding them to the machine learning model. The route is segmented and characterized in advance, allowing the model to make predictions more efficiently during actual computation. This pre-preparation reduces the computational burden during the prediction phase.
Solution Approach 2:
By segmenting the route into discrete edges and processing them individually through the machine learning model, the system reduces the computational complexity of predicting the entire route. Each edge can be processed independently and in parallel, significantly reducing total processing time while maintaining high accuracy through comprehensive feature analysis.
3Measurement precision
If historical location data is filtered to exclude locations indicating bus dwelling or turning, then the speed prediction accuracy improves, but the data processing complexity increases
Solution Approach 1:
The patent applies the extraction principle by identifying and removing (taking out) locations from the historical data that indicate bus dwelling or turning movements. This filtering process extracts only the relevant speed data points that correspond to actual travel segments, improving prediction accuracy by eliminating noisy or irrelevant data while maintaining manageable processing complexity through systematic filtering criteria.
Data Source
AI summary
A system and method for predicting fixed route travel time (e.g., bus speeds along bus routes) is provided. The system and method include a first machine learning model trained to predict speed along the fixed route without turning and dwell times. The speed from the first machine learning model, along with dwell time and turn time can be used with a second machine learning model to determine the overall route time.