Autonomous Object Association and Tracking With Classification Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately associating and tracking dynamic objects in their environment, leading to potential safety and operational inefficiencies due to incorrect object classification and tracking over time.
Innovation Solution
A computer-implemented method using machine-learned models, such as LSTM, to generate association data and determine object classification probability distributions, enabling accurate association and tracking of objects like pedestrians, cyclists, and vehicles by processing input data from sensors like LiDAR and cameras, and controlling vehicle operations based on this data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection and tracking methods are used, then the system can detect objects in the environment, but association errors occur leading to incorrect object classification and tracking over time
Solution Approach 1:
The system employs feedback mechanisms where detection results from previous time steps are fed back into the current detection and association process. This iterative feedback loop allows the system to refine object associations over time, correct classification errors, and maintain consistent tracking by continuously comparing new detections with existing tracks and adjusting associations based on accumulated evidence.
Solution Approach 2:
The system performs preliminary object detection and classification before final association and tracking. By pre-processing detections, generating candidate associations, and preparing track data in advance, the system establishes a foundation for more accurate final classification and reduces errors in the main association process.
2Measurement precision
If the system tracks objects over multiple time intervals to improve tracking accuracy, then object classification improves, but computational complexity and processing time increase
Solution Approach 1:
The system segments the object tracking process into distinct modular components: detection module, association module, classification module, and tracking module. Each module handles specific tasks independently, allowing for optimized processing at each stage and reducing overall computational complexity while maintaining tracking accuracy across multiple time intervals.
Solution Approach 2:
The system dynamically adjusts processing based on object characteristics, track age, and detection confidence. Rather than applying uniform complex processing to all objects, the system adapts its computational effort dynamically, applying more rigorous analysis only when necessary to maintain tracking accuracy.
3Measurement precision
If the system processes sensor data from multiple sources to improve detection accuracy, then object classification improves, but data processing time and computational load increase
Solution Approach 1:
The system merges sensor data from multiple sources (LiDAR, cameras, radar) into a unified detection framework. By integrating data at the detection stage rather than processing each sensor separately through the entire pipeline, the system achieves improved detection accuracy while reducing overall processing time through consolidated data handling.
Solution Approach 2:
The system applies partial processing to certain data streams based on their reliability and relevance. Not all sensor data undergoes the same level of processing - the system selectively applies full processing only to critical or ambiguous detections, while using faster processing for high-confidence, clear detections, thus balancing accuracy with processing speed.
Data Source
AI summary
Systems, methods, tangible non-transitory computer-readable media, and devices associated with object association and tracking are provided. Input data can be obtained. The input data can be indicative of a detected object within a surrounding environment of an autonomous vehicle and an initial object classification of the detected object at an initial time interval and object tracks at time intervals preceding the initial time interval. Association data can be generated based on the input data and a machine-learned model. The association data can indicate whether the detected object is associated with at least one of the object tracks. An object classification probability distribution can be determined based on the association data. The object classification probability distribution can indicate a probability that the detected object is associated with each respective object classification. The association data and the object classification probability distribution for the detected object can be outputted.