Road User Trajectory Ranking Using Endpoint Confidence Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing trajectory prediction methods for autonomous vehicles face challenges in efficiently generating and ranking multiple feasible trajectories for road users, requiring significant computational effort and lacking reliable assessment of trajectory reliability.
Innovation Solution
A method that determines trajectory end points for road users using extracted characteristics, evaluates these end points with a classification to provide a confidence score, and utilizes a machine learning algorithm trained with generative adversarial networks to assess trajectory feasibility, reducing computational effort and improving reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If entity-based models are used to encode every road user and environment elements, then trajectory prediction can be performed, but computational complexity scales with the number of road users leading to high processing effort
Solution Approach 1:
The patent extracts only the essential characteristics needed for trajectory prediction from each road user and environment element, rather than encoding complete entity models. This selective extraction of relevant features reduces the data volume and computational complexity while maintaining prediction accuracy.
Solution Approach 2:
The patent segments the trajectory prediction task into independent evaluations for each trajectory endpoint, rather than processing complete trajectories. This allows parallel processing and reduces the computational burden by breaking down the complex prediction problem into smaller, manageable segments.
2Productivity
If grid maps with detailed environment information are used, then trajectory prediction is independent from number of road users, but post-processing and sampling are required to extract actual coordinates and dynamic properties
Solution Approach 1:
The patent performs preliminary classification and confidence score calculation during the main prediction process, rather than requiring separate post-processing steps. By integrating the evaluation of trajectory feasibility into the prediction algorithm itself, the system avoids additional time-consuming sampling and extraction operations.
Solution Approach 2:
The patent introduces confidence scores as an intermediary metric that directly links the prediction process to trajectory evaluation. This intermediary allows the system to assess trajectory feasibility without requiring complex post-processing to extract coordinates and dynamic properties from heat maps.
3Reliability
If multiple future trajectories are generated for road users, then comprehensive prediction coverage is achieved, but reliable ranking of trajectories to identify most likely paths becomes challenging
Solution Approach 1:
The patent changes the evaluation parameter from complex multi-dimensional trajectory comparison to a single confidence score metric. By transforming the evaluation into a scalar value that represents trajectory feasibility, the system can reliably rank multiple trajectories without complex evaluation procedures.
Solution Approach 2:
The patent replaces complex mechanical-like trajectory evaluation and comparison processes with a machine learning-based classification system. This substitution allows automatic, consistent ranking of trajectories based on learned patterns from training data, eliminating the need for manual or heuristic evaluation methods.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer implemented method for determining and evaluating a trajectory of a road user is provided. Input data associated with a state of movement and with an environment of the road user is received. Characteristics related to the road user are extracted from the input data. One or more trajectory end points are determined for the road user by using extracted characteristics. For each of the trajectory end points, a respective trajectory associated with one of the trajectory end points is determined by using the associated trajectory end point and the extracted characteristics, and the respective trajectory of the road user is evaluated by using a classification which relies on the extracted characteristics to provide a confidence score for each trajectory associated with one of the trajectory end points.