Multi-Axis Position Sensing With De Bruijn Grid Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical position sensors face challenges in achieving high accuracy and efficiency for multi-axis measurements due to computational complexity, noise sensitivity, and high costs, particularly in 3D printing and machine tools, where they require more information and processing power than needed for final position results.
Innovation Solution
A method for multi-axis position sensing using an imaging device to capture a partial image of a grid pattern reference scale, employing differential coding, multilevel halftone grids, and linear summation to interpolate positions efficiently, with Fourier analysis for alignment and error correction, enabling high-aspect-ratio sampling and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If 1D sensors using LFSR code sequences are used for position measurement, then the decoding is simple and effective, but extending to multi-axis measurement is elusive and requires multiple separate sensors
Solution Approach 1:
The patent transitions from 1D linear code sequences to 2D grid patterns arranged in matrix form. Each grid position contains a symbol representing a codeword, enabling simultaneous multi-axis position measurement in a single sensor field of view. The grid structure allows encoding of multiple spatial dimensions (x, y, and rotational axes) that can be decoded independently while sharing the same physical sensor array.
Solution Approach 2:
The 2D grid pattern serves multiple functions simultaneously: it provides position encoding for multiple axes, enables self-location under various orientations, and allows error detection and correction. A single sensor capturing the grid pattern can determine positions along multiple orthogonal axes without requiring separate sensor systems for each axis.
2Adaptability or versatility
If de Bruijn Tori patterns are used for self-location coding, then 4-orientability is achieved, but the decoding becomes computationally heavy and exponentially complex under rotation or noise
Solution Approach 1:
The grid pattern is divided into independently decodable rows and columns, each containing LFSR-based code sequences. This segmentation allows the decoding process to be broken down into simpler one-dimensional decoding operations rather than requiring complex two-dimensional pattern recognition. Each row and column can be decoded separately using efficient LFSR algorithms, reducing overall computational complexity.
Solution Approach 2:
The patent replaces complex mathematical decoding algorithms with simpler LFSR-based decoding mechanisms. By using linear feedback shift register sequences with known algebraic structures, the system achieves efficient decoding through bitwise operations and linear algebra rather than requiring exhaustive pattern matching or complex computational geometry algorithms.
3Measurement precision
If high-resolution features are printed on the reference scale to improve position accuracy, then measurement precision increases, but manufacturing costs increase exponentially
Solution Approach 1:
The patent uses a 2D grid arrangement of symbols where each symbol represents a codeword, rather than requiring continuous high-resolution grayscale patterns. This discrete symbolic representation allows coarser feature sizes to achieve the same information density, making the scale manufacturable with standard lithography processes while maintaining high measurement precision through the information content of the coded symbols.
Solution Approach 2:
The system changes from analog grayscale intensity variations to discrete symbolic representations with distinct visual states. This parameter change from continuous to discrete allows the use of larger, more easily manufactured features while maintaining high measurement resolution through the combinatorial information encoded in the symbol arrangements and patterns.
4Measurement precision
If more camera pixels are used to capture more information from the reference scale, then interpolation accuracy improves, but the amount of data processing and computational requirements increase
Solution Approach 1:
The patent extracts only the essential position information from the captured image by directly decoding the LFSR code sequences embedded in the grid pattern. Rather than processing all pixel data for general image analysis, the system selectively extracts codewords from specific grid positions and decodes them using efficient algebraic algorithms, achieving high precision with minimal computational overhead.
Solution Approach 2:
The reference scale contains redundant codeword representations at multiple grid positions that encode the same position information. This copying of information allows the system to use fewer camera pixels while maintaining accuracy, as the redundant codes provide multiple opportunities to extract the required position data without requiring every pixel to contribute uniquely to the measurement.
Data Source
AI summary
Multi-axis self-location method and apparatus wherein de Bruijn sequences on 2 or more axes are convolved into an array of symbols such as halftone dots to form a reference scale. The position of an imaging device such as a camera relative to the reference scale is ascertained from the captured camera image by bit-wise reconstitution of axis position codes with simple, predominantly linear operations over small neighbourhoods. Judicious choice of differential coding, LFSR generator polynomials, mathematical operators, and deconvolution kernels enables code digits of an axis to be regenerated while simultaneously cancelling out the contributions of other axes. Also optionally provided are uniform DC-balanced variants yielding greatly improved position interpolation, isometric implementations decodable from high-aspect-ratio sample windows, robust concatenated error correction, and extensions into n-space.