Probabilistic Sampling for Track Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing track association methods face challenges in efficiently handling sensor bias and exploring the large solution space for track-to-track association, leading to computational inefficiencies and incomplete characterization of the probability distribution, especially when dealing with multiple sensors and large numbers of tracks.
Innovation Solution
A probabilistic sampling approach that generates multiple full association hypotheses and a soft association matrix, marginalizing over possible sensor biases to prioritize seed track pairs and efficiently approximate the solution space, allowing for near-optimal track association with improved computational efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If exhaustive search of full solution space is performed to identify all association hypotheses, then completeness of probability distribution characterization is improved, but computational time and resources become prohibitive
Solution Approach 1:
The patent applies partial action by performing a limited number of sampling iterations (e.g., 10-100 iterations) rather than exhaustive search of the entire solution space. This partial exploration provides sufficient approximation of the probability distribution for practical decision-making while avoiding prohibitive computational costs. The sampling process generates a representative subset of association hypotheses that capture the essential probability characteristics without enumerating all possible hypotheses.
Solution Approach 2:
The probabilistic sampling algorithm serves itself by using the sampled association hypotheses to directly estimate the probability distribution and make tracking decisions. The sampling process inherently provides both the association estimates and the probability distribution characterization needed for decision-making, eliminating the need for separate exhaustive computation phases.
2Measurement precision
If sensor bias is fully accounted for in track association, then measurement accuracy is improved, but solution space complexity grows exponentially
Solution Approach 1:
The patent changes the approach from deterministic parameter matching to probabilistic parameter sampling. Instead of seeking exact matches between tracks while accounting for sensor bias, the system samples association hypotheses probabilistically, allowing bias effects to be naturally incorporated through the sampling distribution. This transforms the complex exact-matching problem into a more manageable probabilistic estimation problem.
Solution Approach 2:
The probabilistic sampling process acts as an intermediary between raw sensor measurements and final track association decisions. Rather than directly comparing tracks with full bias modeling, the sampling algorithm introduces intermediate association hypotheses that probabilistically bridge the sensor measurements, naturally accounting for bias without requiring explicit complex bias parameter estimation.
3Productivity
If probabilistic sampling is used to approximate solution space, then computational efficiency is improved, but exact probability distribution characterization is lost
Solution Approach 1:
The patent incorporates feedback by using the sampled association hypotheses to refine and update the probability distribution estimates iteratively. Each sampling iteration provides feedback about the likelihood of different associations, which is used to adjust the sampling distribution and improve the accuracy of the probability characterization over time, converging toward the true distribution without requiring exhaustive search.
Data Source
AI summary
A method for solving the track association problem updates and samples a marginal association likelihood conditioned upon existing track assignments and marginalized over possible sensor biases to assign tracks and build a full association hypothesis. The “sampling” is repeated multiple times for a given seed track pair and for different seed track pairs to quickly generate hypotheses that approximate the solution space. The probabilistic track association includes a plurality of likely full association hypotheses and a soft association matrix that probabilistically reflects the likelihood of the track association for a pair of sensors. Efficacy can be enhanced by sensing and processing non-metric features (e.g. size, shape, color) to supplement the metric features (e.g. location, velocity). Efficiency may be enhanced by prioritizing the list of seed track pairs in order of decreasing likelihood, saving only full association hypotheses that are both unique and close to the current most likely hypothesis and terminating the search based on a staleness criteria. Probabilistic sampling may be used for such diverse applications as missile defense, autonomous vehicles and package handling.