Target Vehicle Position Prediction for Frequent Lane Change Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane change assistance systems in autonomous vehicles are computationally expensive and unable to update predictions with high frequency, limiting their availability in diverse situations.
Innovation Solution
A computationally-light predictor that determines vehicle location and speed predictions in time segments, allowing for frequent updates and enabling lane change assistance even when gaps are not immediately adjacent to the host vehicle.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used for motion prediction of surrounding vehicles, then prediction accuracy is improved, but computational cost increases and update frequency decreases
Solution Approach 1:
The prediction time horizon is segmented into multiple time intervals (first time interval, second time interval, third time interval). Different prediction methods are applied to different segments: sensor data is used for the first interval, while predictions from the previous interval are used for subsequent intervals. This segmentation allows the system to achieve both accuracy and high update frequency by updating predictions at each time interval boundary without performing full machine learning computations continuously.
2Measurement precision
If computationally expensive machine learning models are used, then prediction accuracy is improved, but the system cannot update predictions with high frequency
Solution Approach 1:
The system performs prediction computations in advance at defined time intervals rather than continuously. Predictions are computed at the boundary of each time interval using sensor data and previous predictions, then these pre-computed predictions are used throughout the interval. This preliminary action approach reduces computational time by avoiding redundant calculations while maintaining prediction accuracy through periodic updates.
3Device complexity
If lane change assistance is provided only when gaps are immediately adjacent, then system simplicity is maintained, but adaptability to diverse situations is reduced
Solution Approach 1:
The system extends lane change assistance beyond the immediate adjacent spatial dimension by incorporating temporal prediction. Instead of only detecting gaps currently present next to the host vehicle, the system predicts future vehicle positions and identifies gaps that will exist during the lane change maneuver time window. This adds a temporal dimension to the spatial gap detection, enabling assistance in diverse situations where gaps are not immediately adjacent but will be available soon.
Data Source
AI summary
A system includes a vehicle location prediction module (VLPM) configured to receive sensor data associated with a set of target vehicles. For each respective target vehicle, the VLPM is configured to determine a prediction corresponding to the respective target vehicle. The prediction is based on a set of criteria. The prediction includes a plurality of prediction segments corresponding to future time intervals. The prediction indicates a position of the respective target vehicle at each of the plurality of prediction segments. A first prediction segment of the plurality of prediction segments is based on the sensor data. The first prediction segment of the plurality of prediction segments includes maintaining or adjusting a target vehicle acceleration or executing a portion of a lane change based on the outcome of the set of criteria. A second prediction segment of the plurality of prediction segments is based on the first prediction segment.