Autonomous Object Association Tracking With Dynamic Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately associating and tracking dynamic objects in their environment due to changes in object states and types over time, leading to inefficiencies and potential safety issues.
Innovation Solution
A computer-implemented method using machine-learned models 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 such as LiDAR and cameras, and leveraging LSTM models for improved performance.
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 increase when objects change state or type over time
Solution Approach 1:
The system dynamically adapts object classifications by continuously updating probability distributions based on temporal patterns and historical data. The machine learning model adjusts object associations in real-time as objects change state, allowing the tracking system to maintain accuracy despite dynamic environmental conditions and object transformations.
Solution Approach 2:
The system employs feedback mechanisms by using detected object associations and classification probabilities to continuously refine future detections. The temporal pattern recognition uses historical tracking data to inform current associations, creating a closed-loop system that improves accuracy over time and corrects past association errors through learned patterns.
2Measurement precision
If machine-learned models are used to improve object association accuracy, then association errors are reduced, but computational complexity and processing time increase
Solution Approach 1:
The computational system is segmented into specialized components: feature extraction modules, temporal pattern recognition models, and classification probability calculation units. This segmentation allows each component to handle specific aspects of the problem efficiently, reducing overall computational complexity while maintaining high classification accuracy through distributed processing.
Solution Approach 2:
The system performs preliminary feature extraction and temporal pattern analysis before final classification decisions. By pre-processing data and identifying relevant patterns in advance, the system reduces the computational burden during real-time decision-making, enabling accurate classifications without excessive processing complexity.
3Duration of action of stationary object
If multiple object tracks are maintained over multiple time intervals, then tracking continuity is improved, but data processing requirements and computational load increase
Solution Approach 1:
The system extracts only the essential features and temporal patterns needed for accurate tracking, discarding redundant data. By focusing on critical information such as object state changes, position trajectories, and classification probabilities, the system maintains long-term tracking continuity while reducing computational load through selective data retention.
Solution Approach 2:
The system applies partial processing to object tracks with lower priority or lower probability of relevance, while applying full processing to high-priority objects. This selective approach allows the system to maintain tracking for multiple objects over extended periods without processing all data at maximum depth, balancing tracking continuity with processing efficiency.
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.