Sensors can be used to measure the position of an object. In the present thesis the effects which limit the usage of sensors in high dynamic positioning applications on a nanometer level are discussed. Various sensor principles and their properties are investigated and compared. Sensors based on the measurement of i.a. magnetic fields, illumination, or even strain are characterized, as well as their range, bandwidth, resolution, linearity and disturbance rejection is determined.
It will be shown that the simultaneous use of multiple sensors and the specific combination of sensors’ data (fusion) enables a higher performance primarily in terms of resolution and dynamics. Several techniques for the fusion are discussed under consideration of various aspects, however the ultimate aim of sensor fusion is similar.
The methods of feedforward control, complementary filtering, Kalman filtering and optimal filtering (robust control) are developed and verified on practical problems in position sensor systems. To treat various challenges in sensor filtering and sensor fusion a methodological approach, containing separable steps of
• problem formulation with well-defined prerequisits and simplifications,
• theory discussion with approach to find a solution,
• analytical proof or reasoning by statistical values out of numerical simulations,
• experiment design, and
• verification on a real time platform are realized.
Table of Contents
1 Introduction
1.1 Motivation for filtering
1.2 Motivation for sensor fusion
1.3 Boundaries of the thesis
1.4 Sensor
1.4.1 Range
1.4.2 Bandwidth
1.4.3 Resolution
1.4.4 Noise
1.4.5 Drift
1.5 External disturbances
1.6 Atomic Force Microscopy
1.7 Dynamic system
1.7.1 Transfer function
1.7.2 Bode diagram
2 Sensor characterization
2.1 Test setup
2.1.1 Mechanical considerations
2.1.2 Amplifier
2.1.3 Non-linearities
2.2 Data Acquisition
2.2.1 Analog to Digital Converter
2.2.2 Signal conditioning
2.2.3 Performance of conversion
2.3 Strain gauges
2.3.1 Operation principle
2.3.2 Thermal properties
2.3.3 Dynamic properties
2.3.4 Resolution
2.3.5 Non-linearities
2.4 Optical sensors
2.4.1 Operation principle
2.4.2 Photodiode
2.4.3 Optimization
2.4.4 Interference
2.4.5 Optical modulation
2.5 Capacitive sensor
2.5.1 Operation principle
2.5.2 Practical considerations
2.6 Magnetic sensors
2.6.1 GMR as displacement sensor
2.6.2 Temperature drift
2.6.3 Demagnetization
2.6.4 Resolution
2.6.5 Bandwidth
2.7 Piezoelectric force sensor
2.8 Summary
3 Fusion
3.1 Model inversion
3.2 Complementary filters
3.2.1 Implementation
3.3 Kalman filter
3.3.1 Theory of operation
3.3.2 Kalman filter for white noise
3.3.3 MISO Kalman filter
3.3.4 Weighted input Kalman filter
3.3.5 Extended Kalman filter
3.3.6 Kalman filter with noise model
3.3.7 Verification
3.4 Robust filters
3.4.1 Definitions
3.4.2 Problem formulation
3.4.3 Filter synthesis
3.4.4 H2/H∞-optimal filtering with compensation for sensor dynamics
3.4.5 Verification
4 Summary and outlook
4.1 Summary
4.2 Contents
4.3 Outlook
Objectives and Topics
The primary goal of this thesis is to investigate and improve position measurement and estimation in high-dynamic, nanometer-scale positioning applications, such as Atomic Force Microscopy, by employing sensor fusion techniques to overcome individual sensor limitations in terms of bandwidth, resolution, and drift.
- Characterization of displacement sensor principles (Magnetic, Optical, Capacitive, Strain gauges)
- Development of signal conditioning and data acquisition systems
- Implementation of model-based filtering techniques (Kalman filtering, Extended Kalman Filter)
- Application of robust control methods (H2-optimal and H∞-optimal filtering) for sensor data fusion
- Verification of fused sensor performance on real-time platforms (STM32)
Excerpt from the Book
1.1 Motivation for filtering
Already out of the philosophical disquisition it can be recognized that sensor noise and sensor bandwidth play an important role for feedback control systems. To confirm that, a feedback control loop with controller C(s), plant G(s), sensor S(s) according to Figure 1.1 is considered. s denotes the complex Laplace variable defined by s = iω ∈ C. The colored noise arising from the sensor and its environment and noise transfer function N(s) is usually identified by measurements. The sensor model S(s) and that of the plant G(s) are both either analytically derived or identified by measurements.
Summary of Chapters
1 Introduction: Introduces the necessity of sensor fusion and filtering in precision positioning and defines the scope of the thesis regarding sensor properties and dynamic systems.
2 Sensor characterization: Analyzes various displacement sensor principles and their noise, resolution, and bandwidth characteristics, including the setup for measurements.
3 Fusion: Discusses data fusion methods, ranging from simple complementary filters to advanced Kalman filtering and robust H2/H∞-optimal filter design.
4 Summary and outlook: Summarizes the achieved improvements in resolution and bandwidth and suggests advanced future research directions like MHE and particle filters.
Keywords
sensor fusion, nanopositioning, high dynamics, high resolution, sensor noise, Kalman filter, H2-optimal filter, H∞-optimal filter, sensor characterization, data acquisition, displacement measurement, signal conditioning, robust control, Atomic Force Microscopy
Frequently Asked Questions
What is the central focus of this thesis?
The thesis focuses on improving the performance of position measurement systems in high-dynamic, nanometer-level applications through sensor fusion.
What are the primary challenges addressed?
The research addresses limitations in individual sensor performance, specifically regarding sensor noise, restricted bandwidth, and long-term drift in high-precision positioning.
What is the main objective?
The objective is to combine data from multiple, physically different sensors using advanced filtering techniques to emulate an ideal sensor with high resolution and high dynamic performance.
Which scientific methods are employed?
The work utilizes model identification, analytical derivation of transfer functions, Kalman filtering (including Extended Kalman Filters), and robust control theory, specifically H2 and H∞-optimal filter synthesis.
What does the main body cover?
The main body is divided into the characterization of specific sensor principles (magnetic, optical, capacitive, strain gauges) and the development, implementation, and verification of fusion algorithms.
Which keywords define this work?
Key terms include sensor fusion, nanopositioning, Kalman filter, H∞-optimal filter, and sensor characterization.
How is the GMR sensor temperature drift mitigated?
The thermal drift is significantly reduced through physical demagnetization of the shielding components and the application of H∞-optimal filtering techniques.
How is the real-time implementation handled?
The algorithms are implemented on a STM32F407 microcontroller featuring a Floating Processor Unit (FPU), allowing for high-speed signal processing and data acquisition.
What role does the Atomic Force Microscope play?
The AFM serves as an emphasized practical application example that requires high-resolution and high-bandwidth positioning stages.
- Arbeit zitieren
- Daniel Piri (Autor:in), 2014, Sensor Fusion for Nanopositioning, München, GRIN Verlag, https://www.hausarbeiten.de/document/286249
