LMB Filter Object Tracking with Gibbs Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object tracking systems for autonomous driving face challenges in dense urban environments due to multiple measurement hypotheses from sensors like cameras and lidar, which can lead to cluttered measurements and false detections, resulting in the loss of important information by ignoring potentially relevant hypotheses.
Innovation Solution
An extension of the Labeled Multi-Bernoulli Filter (LMB) is developed to efficiently integrate multiple measurement hypotheses, using Gibbs sampling for real-time processing and handling complex scenarios with multiple objects and clutter measurements, allowing for the joint prediction and update of object states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple measurement hypotheses are integrated using LMB filter extension, then tracking reliability is improved, but computational complexity increases
Solution Approach 1:
The measurement hypotheses are segmented and processed individually through the LMB filter framework. Each hypothesis is evaluated separately with its own probability weight, allowing the system to handle multiple hypotheses without treating them as a monolithic complex problem. This segmentation enables manageable computational processing while maintaining reliability through comprehensive hypothesis evaluation.
Solution Approach 2:
The patent changes the parameter representation by introducing probability weights for each measurement hypothesis and using these weights in the LMB filter update equations. By parameterizing the hypotheses with confidence levels and integrating them through probabilistic filtering, the system achieves reliable tracking while keeping computational complexity tractable through efficient parameter updates rather than exhaustive hypothesis evaluation.
2Loss of information
If multiple measurement hypotheses are processed, then information loss is reduced, but processing time increases
Solution Approach 1:
The measurement hypotheses are prepared and weighted in advance before being fed into the LMB filter. Detection algorithms pre-process the raw measurements and assign probability weights to each hypothesis based on initial evaluation metrics. This preliminary action ensures that no information is lost during hypothesis generation, while the pre-computed weights enable efficient processing during the tracking phase, reducing real-time processing time.
Solution Approach 2:
The patent replaces exhaustive hypothesis evaluation with probabilistic filtering. Instead of mechanically processing every possible hypothesis combination through complex evaluation routines, the system uses the LMB filter's probabilistic framework to efficiently update track states based on weighted hypotheses. This substitution of mechanical processing with probabilistic inference significantly reduces processing time while preserving information from multiple hypotheses through their probability weights.
3Reliability
If detection thresholds are lowered to capture more hypotheses, then detection completeness is improved, but false detections increase
Solution Approach 1:
The system dynamically adjusts the effective contribution of each hypothesis based on its probability weight rather than using fixed detection thresholds. By making the hypothesis evaluation dynamic and probability-driven, the system can incorporate low-threshold hypotheses that might represent true objects while automatically down-weighting false detections through their lower probability scores. This dynamic approach improves detection completeness without proportionally increasing false detections.
Solution Approach 2:
The LMB filter provides feedback through the probability weights assigned to each hypothesis and track. Hypotheses that consistently produce false detections receive lower probability weights in subsequent iterations, while true hypotheses are reinforced. This feedback mechanism allows the system to lower detection thresholds for completeness while automatically suppressing false detections through probabilistic feedback from the tracking performance, resolving the contradiction between detecting all objects and filtering false positives.
Data Source
AI summary
A method and system for integrating multiple measurement hypotheses in an efficient labeled multi-Bernoulli (LMB) filter. The LMB filter estimates a plurality of tracks for a plurality of objects, each track of the plurality of tracks having a unique label, a probability, and a state, wherein each track of the plurality of tracks is associated to an object of a plurality of objects to be tracked, each object having an object state. The method receives one or more measurement hypotheses of the multiple measurement hypotheses for each object of the plurality of objects; updates each track of the plurality of tracks based on the respective track and the one or more measurement hypotheses of the multiple measurement hypotheses; determines, for each combination of track of the plurality of tracks and measurement hypothesis, a likelihood ηi(j, k); samples, for each iteration of a plurality of iterations, an update hypothesis γ(t), based on an association of each track of the plurality of tracks to one of: a measurement hypothesis, an events missed detection, or a track dying detection; determining the state of each track of the plurality of tracks based on its respective associations in the updated hypotheses γ(t); extracts, for each track of the plurality of tracks, an existence probability; predicting the object state of each object of the plurality of objects with respect to a next measurement time; determines, whether another update is to be performed; and if another update is to be performed, repeats again the method steps from and including updating each track of the plurality of tracks.