Object Detection Neural Network Sub-period Processing Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in generating timely and accurate object detection data due to limited computational resources and high operational latency, especially when sensors require significant time to complete rotational sensing of the environment.
Innovation Solution
The system partitions time into sub-periods, processes current sensor data using an object detection neural network that includes a sequential processing subnetwork and an object detection subnetwork, allowing for timely and accurate object detection outputs by leveraging a featurization subnetwork to generate intermediate representations, even when computational resources are limited.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system waits for complete rotational sensing data before processing, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent divides the complete rotational sensing period into multiple sub-periods, allowing the system to process and generate object detection outputs for partial scenes incrementally. This segmentation enables timely detections at intermediate time points while maintaining the option to refine with complete data later, thus reducing operational latency without completely sacrificing detection accuracy.
Solution Approach 2:
The system performs preliminary object detection processing on partial sensor data during each sub-period before the complete rotational sensing is finished. This preliminary action generates early object detection outputs that enable timely trajectory planning decisions, while the system can later update these detections with complete scene data for improved accuracy.
2Loss of time
If the system processes data in smaller sub-periods, then loss of time is reduced, but device complexity increases
Solution Approach 1:
The computational system is segmented into modular components including a sensor system, an object detection system with neural networks, and a trajectory planning system. This modular segmentation allows each component to operate independently on sub-period data, reducing overall system complexity while enabling timely processing.
Solution Approach 2:
The system performs partial processing on sub-period data rather than waiting for complete data sets. By processing available partial information incrementally and updating detections as more data becomes available, the system reduces computational complexity at each time step while maintaining acceptable detection accuracy for real-time decisions.
3Measurement precision
If the system uses complete rotational sensing data, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The detection process is segmented into multiple outputs generated at different time points during the rotational sensing period. Instead of producing a single output after complete sensing, the system generates intermediate detection outputs at each sub-period boundary, increasing the detection output rate while using progressively more complete scene data.
Solution Approach 2:
The object detection system continuously processes incoming sensor data throughout the rotational sensing period, generating useful detection outputs at each sub-period rather than waiting idle for complete data. This continuous processing maintains high productivity by constantly producing valuable detection results that can inform trajectory planning.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing data generated by a sensing system that rotationally senses an environment. In one aspect, a method comprises partitioning a predetermined period of time into a plurality of sub-periods, wherein the predetermined period of time is a period of time for which data generated by the sensing system constitutes a complete rotational sensing of the environment; for each sub-period: receiving current data generated by the sensing system during the sub-period and characterizing a respective partial scene of the environment; processing the current data using an object detection neural network to generate a current object detection output that is specific to the respective partial scene of the environment.