Kernelized Data Fusion for Vehicle Positioning Uncertainty
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Relevance Vector Machines (RVM) have limited flexibility and struggle to adequately model uncertainty for inputs far away from the training data, especially when dealing with heterogeneous data sources that produce dissimilar data types and require different processing methods.
Innovation Solution
The method involves performing a kernel function on system inputs from heterogeneous data sources to generate kernelised data, which is then used to infer the significance of each data source through a regression technique, such as a Gaussian Process, allowing for the combination of different kernel functions and feature extraction processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If conventional RVM is used with basis functions provided a priori, then computational cost is reduced, but flexibility and ability to model uncertainty for far-away inputs deteriorates
Solution Approach 1:
The patent applies dynamics by replacing the static, fixed basis functions with dynamic kernel functions that adapt to the input data distribution. The kernel functions automatically adjust their behavior based on the proximity of inputs to training data, enabling the model to maintain computational efficiency while gaining the flexibility to model uncertainty for far-away inputs without requiring manual specification of basis functions.
Solution Approach 2:
The patent changes the parameter approach from fixed a priori basis functions to learned kernel parameters. The kernel functions use parameters that are inferred from the data itself, allowing the model to adapt to different data distributions and maintain both computational efficiency and flexibility in modeling uncertainty across the input space.
2Device complexity
If conventional RVM is used with fixed basis functions, then model structure is simplified, but ability to handle heterogeneous data sources deteriorates
Solution Approach 1:
The patent applies local quality by using different kernel functions for different data sources or input regions. Each kernel function can be tailored to the specific characteristics of its corresponding data source, allowing the model to handle heterogeneous data while maintaining a relatively simple overall structure. The kernel functions adapt locally to the data distribution without requiring complex global model restructuring.
3Adaptability or versatility
If kernel functions are applied to heterogeneous data sources, then flexibility and uncertainty modeling improve, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the data processing into separate kernel function applications for different data sources. Each kernel function operates independently on its corresponding data source, which allows the model to handle heterogeneous data with flexibility while managing computational complexity through modular processing. The segmentation enables parallel computation and reduces the overall computational burden compared to a monolithic approach.
Data Source
AI summary
A method and apparatus for processing data, the data including a set of one or more system inputs; and a set of one or more system outputs; wherein each system output corresponds to a respective system input. Each system input can include a plurality of data points, such that at least one of these data points is from a different data source to at least one other of those data points. The method includes performing a kernel function on a given system input from the data and a further system input to provide kernelized data; and inferring a value indicative of a significance of data from a particular data source; wherein the inferring includes applying a regression technique to the kernelized data.