Multi-modal Traffic Detection Using Cluster Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current algorithms fail to accurately handle tri-modal traffic scenarios, particularly in noisy data environments, such as roads with multiple lanes or mixed transportation modes, leading to unreliable lane-level traffic information.
Innovation Solution
A method involving cluster analysis of multi-dimensional probe data to identify and separate modes within multi-modal traffic scenarios, using a gap value and bucket-based approach to associate probe samples with clusters representing different traffic modes, and providing real-time traffic information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple speed averaging is used to derive average speed of traffic participants, then the processing is simple and fast, but the traffic information becomes wrong in multi-modal scenarios
Solution Approach 1:
The patent segments the traffic data into multiple modes using cluster analysis. Instead of treating all traffic participants as a single group, the algorithm divides them into distinct clusters (modes) based on their speed characteristics. This segmentation allows accurate representation of multi-modal traffic scenarios where different groups of vehicles have different speed profiles, resolving the contradiction between simple processing and accurate information.
2Reliability
If special algorithms are used to analyze divergent traffic speeds before splitting junctions, then bi-modal traffic distribution problems are solved, but tri-modal and more complex scenarios cannot be handled
Solution Approach 1:
The patent implements a universal cluster analysis algorithm that can handle any number of modes (bi-modal, tri-modal, or more complex distributions). The algorithm uses a gap statistic approach that automatically determines the optimal number of clusters based on the data, making it adaptable to various traffic scenarios without requiring scenario-specific configuration. This universal approach resolves the contradiction by providing both reliable bi-modal detection and extended versatility for complex multi-modal scenarios.
3Measurement precision
If lane-level separation of divergent traffic speeds is attempted, then accurate lane-level traffic information can be obtained, but the processing becomes unreliable due to noisy positioning data
Solution Approach 1:
The patent employs a gap statistic feedback mechanism that evaluates the quality of cluster separation. The gap statistic compares the within-cluster variation to the between-cluster variation, providing feedback on whether the observed modes are statistically significant or merely artifacts of noisy data. This feedback mechanism allows the algorithm to distinguish between genuine multi-modal traffic patterns and noise-induced variations, resolving the contradiction between achieving lane-level precision and maintaining reliability in noisy conditions.
4Quantity of substance
If data from multiple transportation modes (bikes, cars, pedestrians) using the same road is collected, then comprehensive traffic information is obtained, but current algorithms cannot reliably handle such multi-modal scenarios
Solution Approach 1:
The patent changes the analysis parameters by using speed distribution characteristics as the basis for mode identification rather than relying on predefined transportation mode categories. The cluster analysis automatically detects distinct speed patterns in the data, which correspond to different transportation modes (pedestrians, cyclists, vehicles). This parameter change allows the algorithm to reliably handle mixed transportation mode data without requiring prior knowledge of the specific modes present, resolving the contradiction between comprehensive data coverage and algorithm reliability.
Data Source
AI summary
A method is provided, performed by at least one apparatus, the method including: obtaining probe data including a plurality of probe samples of a multi-dimensional probe sample space, the probe data being representative of a potentially multi-modal traffic scenario; performing a cluster analysis for at least a part of the probe samples of the probe data, the cluster analysis including: associating at least a part of the probe samples with respective clusters, each cluster being representative of a mode of the potentially multi-modal traffic scenario.


