Top-Down Lidar Frame Fusion for Fast Object Behavior Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in safely navigating through environments due to the uncertainty in predicting the behavior of dynamic objects, which is exacerbated by limited processing time and computational resources.
Innovation Solution
The system employs a top-down segmentation and classification approach, aligning and reducing the data from prior lidar frames to create a multichannel top-down representation, which reduces processing requirements while providing temporal input to machine learned models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If temporal data from multiple prior lidar frames is utilized to improve prediction accuracy, then object behavior prediction accuracy is improved, but processing time and computational resources are overwhelmed
Solution Approach 1:
The patent segments the temporal data from multiple prior lidar frames by selecting only specific key frames (e.g., the most recent frame and select earlier frames) rather than processing all available frames. This segmentation reduces the temporal data input to machine learned models while maintaining predictive accuracy, thereby reducing processing time and computational resource requirements.
Solution Approach 2:
The patent extracts and removes redundant or less important temporal data from the sequence of prior lidar frames. By taking out only the essential temporal information needed for accurate prediction and discarding unnecessary data, the system maintains prediction accuracy while significantly reducing the processing burden and time consumption.
2Measurement precision
If temporal data from multiple prior lidar frames is utilized to improve prediction accuracy, then object behavior prediction accuracy is improved, but computational resources are overwhelmed
Solution Approach 1:
The patent segments the temporal data from multiple prior lidar frames by selecting only specific key frames (e.g., the most recent frame and select earlier frames) rather than processing all available frames. This segmentation reduces the temporal data input to machine learned models while maintaining predictive accuracy, thereby reducing processing time and computational resource requirements.
Solution Approach 2:
The patent applies partial action by using only a subset of available temporal data (selecting specific prior frames rather than all frames) to achieve sufficient prediction accuracy. This partial use of temporal information reduces computational resource consumption while maintaining the necessary level of accuracy for safe autonomous vehicle operation.
3Reliability
If multiple prior lidar frames are processed to reduce uncertainty in object behavior prediction, then prediction reliability is improved, but processing time increases
Solution Approach 1:
The patent segments the temporal data from multiple prior lidar frames by selecting only specific key frames (e.g., the most recent frame and select earlier frames) rather than processing all available frames. This segmentation reduces the temporal data input to machine learned models while maintaining predictive accuracy, thereby reducing processing time and computational resource requirements.
Solution Approach 2:
The patent performs preliminary selection and filtering of prior lidar frames before processing them through machine learned models. By pre-identifying and selecting only the most relevant frames that contribute to reducing prediction uncertainty, the system achieves reliable predictions more efficiently, reducing the overall processing time while maintaining prediction reliability.
Data Source
AI summary
Techniques for detecting and classifying objects using lidar data are discussed herein. In some cases, the system may be configured to utilize a predetermined number of prior frames of lidar data to assist with detecting and classifying objects. In some implementations, the system may utilize a subset of the data associated with the prior lidar frames together with the full set of data associated with a current frame to detect and classify the objects.