Multi-Sensor Object Tracking With Confidence-Weighted Kalman Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems face challenges in integrating data from sensors of varying quality and precision, leading to intermittently missing data and sudden deterioration in track quality when relying on lower capability sensors.
Innovation Solution
A modified Kalman filter process that applies weighted 'Kalman gain' values to combine tracking data from multiple sensors, adjusting states and covariances based on relative confidence, allowing for improved data integration and maintaining track quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only the output from the sensor providing the highest quality output is used, then measurement precision is improved, but reliability deteriorates when the best sensor is temporarily not detecting the object
Solution Approach 1:
The patent combines outputs from multiple tracking data sources (sensors) into a single integrated track estimate. When the best sensor is unavailable, the system merges data from lower-quality sensors to maintain continuous tracking, resolving the contradiction between using only high-precision data and maintaining reliability during sensor outages.
Solution Approach 2:
The system dynamically adjusts which sensor data to use based on availability and quality. The data source selection is not static but adapts in real-time to current sensor performance and availability conditions, allowing the system to switch between sensors optimally to maintain both precision and reliability.
2Reliability
If data from multiple sensors of varying quality is integrated, then reliability is improved, but measurement precision deteriorates when lower quality sensors are used
Solution Approach 1:
The patent applies different quality weights to different data sources based on their individual characteristics. Each sensor's contribution to the final track estimate is locally optimized according to its specific quality level, allowing the system to maintain high precision when using lower-quality sensors by appropriately weighting their contributions.
Solution Approach 2:
The system changes the parameters (weights) assigned to different data sources based on their quality and availability. By dynamically adjusting these parameters, the system can optimize the balance between reliability and precision, giving more weight to higher-quality sensors when available and appropriately compensating when using lower-quality alternatives.
3Adaptability or versatility
If a weighted combination of multiple sensors is used, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent implements a practical weighting system that uses sufficient (but not excessive) complexity to achieve the desired adaptability. The data source selection and weighting mechanism uses a manageable level of computational complexity to provide flexible sensor integration, applying the principle of using just enough complexity to solve the problem without unnecessary overhead.
Data Source
Figure 1~2
Figure 3
AI summary
A system, computing device and a computer-implemented method of tracking an object are disclosed. Embodiments obtain (302) a plurality of estimated states and covariances relating to a tracked object and associated with a respective plurality of tracking data sources, and process (304) the obtained plurality of estimated states and covariances using a modified Kalman filter process to generate an updated state and updated covariance. Embodiments generate (306) object tracking information using the updated state and updated covariance. The modified Kalman filter process comprises computing the updated state and the updated covariance in an update phase, wherein the computing of the updated state and the updated covariance is weighted according to relative confidence in the modified Kalman filter process and each of the plurality of tracking data sources.