Multi-Scan Radar-Vision Fusion for Uncertainty-Aware Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor fusion systems for autonomous driving are computationally complex and lack quantifiable uncertainty, failing to meet safety regulations and precision requirements for advanced driver-assistance and autonomous driving systems.
Innovation Solution
Implement multi-scan sensor fusion techniques using Dempster-Shafer theory to determine belief and plausibility parameters for radar and vision tracks, enabling high-precision object tracking with quantifiable uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing T2TF techniques use all available pieces of data for global and coherent track-to-target association, then measurement precision is improved, but device complexity and computational requirements increase significantly
Solution Approach 1:
The patent segments the track-to-track fusion process into multiple discrete steps: generating hypotheses for track associations, calculating mass values for each hypothesis, determining belief and plausibility parameters, and selecting final associations. This segmentation breaks down the complex global optimization problem into manageable sequential operations that can be executed with standard vehicle processors.
Solution Approach 2:
The patent considers a predefined number of most likely hypotheses (e.g., top 3-5 hypotheses) rather than exhaustively evaluating all possible track associations. This partial action approach maintains high tracking precision by focusing on the most probable associations while significantly reducing computational complexity compared to evaluating all possible combinations.
2Reliability
If existing T2TF systems process all sensor data for comprehensive object tracking, then reliability is improved, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-defining the number of hypotheses to consider and pre-establishing the fusion algorithm structure before actual sensor data arrives. This allows the system to quickly process incoming radar and vision tracks without requiring complex real-time optimization, thereby maintaining high reliability while reducing processing time.
Solution Approach 2:
The patent implements periodic action by processing sensor data in discrete time steps or scans, where hypotheses are generated and evaluated at regular intervals. This periodic processing approach ensures comprehensive tracking reliability while controlling processing time through structured, time-discrete operations rather than continuous computation.
3Measurement precision
If existing sensor fusion systems provide comprehensive object tracking, then measurement precision is improved, but uncertainty quantification capability is insufficient
Solution Approach 1:
The patent introduces belief and plausibility parameters as intermediary measures that bridge the gap between precise track association and uncertainty quantification. These parameters serve as mediators that provide both the precision needed for accurate tracking and the uncertainty information required for safety-critical decision-making in autonomous driving systems.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This document describes techniques, systems, and methods for multi-scan sensor fusion for object tracking. A sensor-fusion system can obtain radar tracks and vision tracks generated for an environment of a vehicle. The sensor-fusion system maintains sets of hypotheses for associations between the vision tracks and the radar tracks based on multiple scans of radar data and vision data. The set of hypotheses include mass values for the associations. The sensor-fusion system determines a probability value for each hypothesis. Based on the probability value, matches between radar tracks and vision tracks are determined. The sensor-fusion system then outputs the matches to a semi-autonomous or autonomous driving system to control operation of the vehicle. In this way, the described techniques, systems, and methods can provide high-precision object tracking with quantifiable uncertainty.