Wheel Direction Classification for Trajectory Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately predicting the trajectories of wheeled vehicles, especially when they are stopped or moving slowly, leading to conservative decision-making due to uncertainty, which can impact safety and efficiency.
Innovation Solution
The use of a machine-learned model trained to classify or regress the direction of wheels from a single image, combining this information with other data to estimate the trajectory of the wheeled vehicle, allowing for quicker and more accurate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional trajectory prediction methods are used for stopped or slowly moving wheeled vehicles, then processing time is reduced, but measurement precision and reliability of trajectory prediction deteriorate
Solution Approach 1:
The system performs preliminary classification of wheel direction (straight, left, right) before full trajectory prediction. This preliminary action provides immediate directional information for stopped or slow-moving vehicles, improving prediction accuracy without requiring complete trajectory analysis, thus resolving the time-precision tradeoff.
Solution Approach 2:
The trajectory prediction process is segmented into discrete wheel direction classifications (straight, left, right) that can be quickly determined from image data. This segmentation allows the system to provide accurate directional predictions for stopped vehicles without performing computationally intensive full trajectory analysis, balancing speed and precision.
2Reliability
If comprehensive sensor data processing is used to predict trajectories, then reliability improves, but use of energy and processing resources increases
Solution Approach 1:
The system extracts only the essential wheel direction information from image data using a trained neural network, rather than processing all sensor data comprehensively. This extraction approach maintains prediction reliability by focusing on the most informative feature (wheel orientation) while significantly reducing energy consumption compared to full data processing.
Solution Approach 2:
The system uses a pre-trained neural network model that has already learned the relationship between wheel orientation and trajectory. This copied knowledge from training data allows reliable predictions without reprocessing raw sensor data, reducing energy consumption while maintaining accuracy.
Data Source
AI summary
A trajectory estimate of a wheeled vehicle can be determined based at least in part on determining a wheel angle associated with the vehicle. In some examples, at least a portion of the image associated with the wheeled vehicle may be input into a machine-learned model that is trained to classify and/or regress wheel directions of wheeled vehicles. The machine-learned model may output a predicted wheel direction. The wheel direction and/or additional or historical sensor data may be used to estimate a trajectory of the wheeled vehicle. The predicted trajectory of the object can then be used to generate and refine an autonomous vehicle's trajectory as the autonomous vehicle proceeds through the environment.