Sensor Registration via C-GRASP Global Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for sensor registration, particularly in global optimization problems, are limited by the need for good initial estimates, knowledge of function gradients, and are prone to converging on local minima instead of global minima, making them computationally inefficient and unreliable for dynamic measurement systems with multiple sensors of varying types.
Innovation Solution
A two-step method is employed, where a systematic error function is minimized using Continuous Greedy Randomized Adaptive Search Procedure (C-GRASP) to determine correction factors, followed by an assignment method to assign objects between sensors, decomposing the problem into a systematic error problem and an assignment problem for more efficient sensor registration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional global optimization methods are used for sensor registration, then the problem can be solved with a single approach, but the method requires good initial estimates and knowledge of function gradients, leading to convergence on local minima and computational inefficiency
Solution Approach 1:
The patent segments the sensor registration problem into two distinct sub-problems: (1) determining systematic errors between sensors, and (2) assigning objects to sensors. This segmentation allows each sub-problem to be solved with appropriate methods without the limitations of traditional unified global optimization approaches.
Solution Approach 2:
The patent performs preliminary action by first determining systematic errors between sensors before performing object assignment. This preliminary error characterization enables subsequent assignment algorithms to work with corrected data, improving both reliability and computational efficiency.
2Measurement precision
If traditional optimization methods are used, then a unified approach is applied, but the method is prone to converging on local minima instead of global minima
Solution Approach 1:
By segmenting the optimization problem into systematic error determination and object assignment phases, the patent eliminates the risk of converging on local minima that plagues unified optimization methods. Each phase can be solved with deterministic or less sensitive algorithms.
Solution Approach 2:
The patent introduces an intermediary step of systematic error characterization that mediates between raw sensor data and final object assignment. This intermediary processing removes systematic biases before assignment, improving both precision and convergence reliability.
3Adaptability or versatility
If a single unified method is used for sensor registration, then implementation is simple, but the method cannot efficiently handle dynamic measurement systems with multiple sensors of varying types
Solution Approach 1:
The patent segments the sensor registration methodology into modular components that can be independently applied to different sensor types and configurations. This segmentation provides flexibility for handling diverse sensor arrays while maintaining manageable complexity through standardized interfaces.
Solution Approach 2:
The patent creates a universal framework where the systematic error determination method can be applied across different sensor types (imaging sensors, non-imaging sensors, sensors of the same or different kinds). This universal approach enhances adaptability while the modular structure controls complexity.
Data Source
AI summary
Disclosed are method and apparatus for registering multiple sensors collecting data from multiple objects. Sensor registration is decomposed into a two-step procedure. The first step corrects systematic errors. The second step assigns objects measured by one sensor to objects measured by a second sensor. Systematic errors are corrected by generating the global minimum of a systematic error function. One embodiment for generating the global minimum uses a Continuous Greedy Randomized Adaptive Search Procedure.