Inference Graph Hybrid Traffic Flow Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current micro-traffic behavior modeling methods are inadequate for simulating hybrid traffic flow scenarios, particularly at intersections, as they fail to accurately represent complex interactions among multiple subjects on a continuous plane, leading to low precision and unreliability in modeling and simulation, and cannot provide a highly realistic virtual test environment for intelligent vehicles.
Innovation Solution
A hybrid traffic flow motion behavior modeling method based on an inference graph, which constructs an interaction graph to represent traffic participants and their relationships, generates possible interaction situations, judges their rationality using empirical decision-making criteria, and iteratively refines the scenarios to determine a final execution trajectory, incorporating physical and logical interaction relationships and utilizing Bessel curves and intelligent driver models for trajectory estimation and speed planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If lane-based structural modeling is used for traffic simulation, then the modeling structure is simplified and easy to implement, but the simulation precision and reliability are reduced when modeling complex hybrid traffic flow interactions at intersections
Solution Approach 1:
The patent segments the continuous traffic flow interaction space into discrete interaction regions based on potential conflict points between different traffic participants. Each interaction region is modeled independently with specific interaction rules, allowing complex multi-subject interactions to be decomposed into manageable segments while maintaining overall simulation accuracy
Solution Approach 2:
The patent transitions from traditional 2D lane-based modeling to a 3D interaction space model that incorporates temporal dimensions and interaction priority layers. This dimensional expansion allows the model to represent complex spatial-temporal interactions among multiple traffic participants simultaneously, resolving the contradiction between structural simplicity and interaction complexity
2Ease of operation
If priority rules and gap selection models are applied at intersections, then traffic flow control is simplified, but the ability to accurately represent complex multi-subject interactions and non-strict priority obeying is lost
Solution Approach 1:
The patent implements dynamic priority assignment where interaction priorities are not fixed but determined in real-time based on the specific interaction situation, traffic participant types, and environmental conditions. The model can dynamically adjust priority relationships during simulation, allowing non-strict priority obeying and flexible decision-making that adapts to complex multi-subject interactions
Solution Approach 2:
The patent changes the fundamental parameters of interaction modeling by introducing interaction intensity weights, priority coefficients, and situational factors that modulate the behavior of traffic participants. These parameter changes allow the model to represent nuanced interaction decisions where priority rules are applied flexibly rather than rigidly, capturing real-world driving behavior more accurately
3Stability of the object's composition
If virtual lanes are constructed for shared spaces without lane division, then the road network structure is standardized, but the realistic representation of continuous plane interactions is compromised
Solution Approach 1:
The patent introduces interaction regions as intermediary elements between virtual lanes and actual continuous space interactions. These interaction regions serve as mediators that capture the essence of continuous plane interactions while maintaining compatibility with the standardized virtual lane structure, allowing the model to represent both structured and unstructured spaces accurately
Data Source
AI summary
The disclosed relates to a hybrid traffic flow motion behavior modeling method based on an inference graph, wherein the method comprises: obtaining scene information, representing all traffic participants in the scene as vertices, and using directed edges to represent interaction relationships among traffic participants, so as to obtain the interaction graph; obtaining all possible interaction situations according to the interaction graph; based on each possible interaction situation, estimating the trajectory of each traffic participant in the interaction situation, and judging whether the trajectory conforms to a preset empirical decision-making criteria, so as to judge rationality of the interaction situation; and judging the rationality of all possible interaction situations obtained in the interaction situation generation step in turn until an interaction situation satisfying the rationality is found, and taking a trajectory of each traffic participant corresponding to the interaction situation as a final execution trajectory.