Dual-Pipeline Perception Updates for Low-Latency Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional perception systems in autonomous vehicles suffer from significant processing latency and outdated sensor data, particularly with objects that change trajectory rapidly, compromising safety.
Innovation Solution
Implementing an out of sequence perception system that includes a full object tracking update pipeline and a fast object tracking update pipeline, utilizing recent radar data to reduce latency and improve the recency of sensor data for navigation decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional perception systems process sensor data through complete detection, classification, and prediction pipelines, then comprehensive object information is obtained, but processing latency increases significantly and data becomes outdated
Solution Approach 1:
The perception system is divided into two independent pipelines: a full object tracking update pipeline that performs complete detection, classification, and prediction at lower frequency, and a fast object tracking update pipeline that uses radar data for rapid updates at higher frequency. This segmentation allows the system to maintain both comprehensive accuracy and low latency by updating different aspects of object tracking at appropriate speeds.
Solution Approach 2:
The system performs preliminary detection and classification using the full pipeline, then uses this pre-processed information as a foundation for rapid updates via the fast pipeline. By preparing object tracks in advance with complete processing and then updating them incrementally with radar data, the system reduces the computational burden during critical update phases while maintaining accuracy.
2Speed
If sensors capture data at high frequency to track rapidly changing objects, then recency of data improves, but processing complexity and computational load increase
Solution Approach 1:
The fast pipeline extracts only the essential radar data processing steps needed for rapid updates, separating them from the complete detection and classification procedures. By taking out only the necessary fast-update operations and using pre-computed information from the full pipeline, the system achieves high update rates without proportionally increasing overall processing complexity.
Solution Approach 2:
The fast pipeline performs partial processing on radar data, focusing only on the critical updates needed for rapidly changing objects rather than complete processing. This partial action approach maintains high update rates while avoiding the excessive computational load that would result from applying full processing pipelines at every update cycle.
3Reliability
If complete object tracking pipelines are used for all sensor data, then accurate object classification is achieved, but processing time becomes too long for safety-critical applications
Solution Approach 1:
The system dynamically adjusts the processing depth and frequency based on object characteristics and environmental conditions. The dual-pipeline architecture allows the system to switch between comprehensive processing and rapid updates as needed, optimizing the balance between reliability and response time for different operational scenarios.
Solution Approach 2:
The fast pipeline provides continuous feedback on object position and velocity changes, allowing the system to detect when rapid changes occur and trigger appropriate responses. This feedback mechanism ensures safety-critical updates are captured in real-time while maintaining overall system reliability through the complementary full pipeline.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides more accurate and timely object perception tracks, allowing autonomous vehicles to react earlier and generate safer trajectories, reducing the risk of collisions.
Implementation Method 1
receiving first radar data captured by a second sensor during the first period of time
Implementation Method 2
receiving first light detection and ranging (lidar) data captured by a first sensor during a first period of time
Data Source
AI summary
Techniques for updating data operations in a perception system are discussed herein. A vehicle may use a perception system to capture data about an environment proximate to the vehicle. The perception system may receive image data, lidar data, and/or radar data to determine information about an object in the environment. As different sensors may be associated with different time periods for capturing and/or processing operations, the techniques include updating object data with data from sensors associated with a shorter time period to generate intermediate object data. Such intermediate object data may reduce a delay in updating a position of an object in an environment.