Vehicle Path Prediction Using Future Sensor Data Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current path prediction systems for intelligent driving monitoring systems (IDMS) and autonomous driving systems face challenges in accurately predicting paths of travel, especially in complex driving scenarios with unclear or occluded lane markings, and require large amounts of labeled training data, which is cumbersome and expensive to obtain.
Innovation Solution
The system employs a 'Back-to-the-Future' approach that incorporates information from later time points to determine and map the path of travel, using a combination of camera, GPS, Lidar, and radar data, allowing for faster and higher-resolution processing, and reduces the need for human-labeled data by using future data as machine-labeled inputs for training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional path prediction systems use current-time data only, then processing is simpler and faster, but prediction accuracy deteriorates in complex driving scenarios
Solution Approach 1:
The system performs preliminary actions by collecting and storing sensor data (camera, GPS, Lidar, radar) at multiple future time points before the actual prediction is needed. This pre-collected data from later time points is then used to improve path prediction accuracy without adding complex real-time processing requirements during critical decision-making moments.
Solution Approach 2:
Instead of predicting the future path using only current data, the system inverts the approach by using data from future time points to determine and map the path of travel. This reverse chronological approach allows the system to leverage information that would naturally become available later to improve current prediction accuracy.
2Measurement precision
If large amounts of human-labeled training data are used, then model accuracy improves, but data collection time and cost increase
Solution Approach 1:
The system performs self-service by automatically generating its own training data through the 'Back-to-the-Future' approach. The sensor data collected from future time points serves as machine-labeled inputs for training, eliminating the need for time-consuming and expensive manual human labeling while still achieving high model accuracy.
Solution Approach 2:
The system creates copies of real-world driving scenarios by collecting and storing sensor data at multiple future time points. These copied scenarios serve as training data, allowing the model to learn from realistic driving situations without requiring physical human annotation of each scenario.
3Reliability
If sensor data from later time points is incorporated, then path prediction robustness improves, but processing latency increases
Solution Approach 1:
The system performs preliminary data collection and processing during non-critical periods, storing sensor data from multiple future time points before predictions are needed. This pre-processing approach allows the system to use robust multi-timepoint data without adding latency to real-time decision-making, as the heavy lifting is done in advance.
Data Source
AI summary
The present disclosure provides systems and methods for mapping a determined path of travel. The path of travel may be mapped to a camera view of a camera affixed to a vehicle. In some embodiments, the path of travel may be mapped to another view that is based on a camera, such as a bird's eye view anchored to the camera's position at a given time. These systems and methods may determine the path of travel by incorporating data from later points in time.