Autonomous Vehicle Scenario Filtering for Pedestrian Intent Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle systems fail to accurately predict the motion of non-stationary objects like pedestrians and bicyclists, leading to unnatural vehicle movements due to insufficient information about their intentions and behaviors.
Innovation Solution
The use of machine learning-based models to predict hidden context attributes associated with traffic entities, such as pedestrians and bicyclists, by training models with video frames that represent various traffic scenarios, allowing the system to classify and filter data to identify subsets of scenarios for proper training and validation, and generating navigation action tables based on ground truth analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional motion prediction techniques are used for non-stationary objects, then the system is simple to operate, but the prediction accuracy is insufficient
Solution Approach 1:
The patent introduces an intermediary machine learning model that processes sensor data to predict hidden context attributes of traffic entities. This intermediary layer between raw sensor data and motion prediction enables accurate prediction of pedestrian and bicyclist behavior by inferring intentions, goals, and state of mind, thereby resolving the contradiction between prediction accuracy and system complexity.
Solution Approach 2:
The patent changes the parameters being analyzed from simple motion characteristics to hidden context attributes including state of mind, awareness, and intentions. By transforming the prediction task from detecting visible motion patterns to inferring hidden psychological states, the system achieves superior prediction accuracy for non-stationary objects while managing complexity through structured attribute classification.
2Reliability
If the autonomous vehicle waits for every detected pedestrian to cross, then collision safety is improved, but the vehicle experiences loss of time
Solution Approach 1:
The patent implements feedback through the machine learning model that continuously predicts whether a detected pedestrian will actually cross the street based on hidden context attributes. This feedback mechanism allows the autonomous vehicle to dynamically adjust its waiting decision - only waiting when the prediction indicates high probability of crossing, and proceeding when the prediction suggests the pedestrian will remain stationary, thereby balancing safety with time efficiency.
Solution Approach 2:
The patent applies partial action by selectively waiting for pedestrians only when the hidden context prediction indicates they are likely to cross. Instead of universally waiting for all detected pedestrians, the system applies the waiting action only in specific cases where the predicted probability of crossing exceeds a threshold, thus avoiding unnecessary time loss while maintaining collision safety.
3Reliability
If the autonomous vehicle suddenly stops to avoid potential pedestrians, then collision prevention is improved, but the vehicle movement becomes unnatural
Solution Approach 1:
The patent applies preliminary action by using the machine learning model to predict pedestrian crossing intentions before the pedestrian actually moves into the vehicle's path. By detecting hidden context attributes such as state of mind and awareness in advance, the system can prepare appropriate responses - smoothly decelerating or maintaining course - rather than making sudden stops, thereby preventing collisions while maintaining natural movement patterns.
Data Source
AI summary
A system uses a machine learning based model to determine attributes describing states of mind and behavior of traffic entities in video frames captured by an autonomous vehicle. The system classifies video frames according to traffic scenarios depicted, where each scenario is associated with a filter based on vehicle attributes, traffic attributes, and road attributes. The system identifies a set of video frames associated with ground truth scenarios for validating the accuracy of the machine learning based model and predicts attributes of traffic entities in the video frames. The system analyzes video frames captured after the set of video frames to determine actual attributes of the traffic entities. Based on a comparison of the predicted attributes and actual attributes, the system determines a likelihood of the machine learning based model making accurate predictions and uses the likelihood to generate a navigation action table for controlling the autonomous vehicle.