Machine learning and geodesy: A survey

The goal of classical geodetic data analysis is often to estimate distributional parameters like expected values and variances based on measurements that are subject to uncertainty due to unpredictable environmental effects and instrument specific noise. Its traditional roots and focus on analytical solutions at times require strong prior assumptions regarding problem specification and underlying probability distributions that preclude successful application in practical cases for which the goal is not regression in presence of Gaussian noise. Machine learning methods are more flexible with respect to assumed regularity of the input and the form of the desired outputs and allow for nonparametric stochastic models at the cost of substituting easily analyzable closed form solutions by numerical schemes. This article aims at examining common grounds of geodetic data analysis and machine learning and showcases applications of algorithms for supervised and unsupervised learning to tasks concerned with optimal estimation, signal separation, danger assessment and design of measurement strategies that occur frequently and naturally in geodesy.

Linear System Identification and Vibration Control of End-Effector for Industrial Robots

This paper presents the discrete state space mathematical model of the end-effector in industrial robots and designs the linear-quadratic-Gaussian controller, called LQG controller for short, to solve the low frequency vibration problem. Though simplifying the end-effector as the cantilever beam, this paper uses the subspace identification method to determine the output dynamic response data and establishes the state space model. Experimentally comparing the influences of different input excitation signals, Chirp sequences from 0 Hz to 100 Hz are used as the final estimation signal and the excitation signal. The LQG controller is designed and simulated to achieve the low frequency vibration suppression of the structure. The results show that the suppression system can effectively suppress the fundamental natural frequency and lower vibration of end-effector. The vibration suppression percentage is 95%, and the vibration amplitude is successfully reduced from ±20 μm to ±1 μm. The present work provides an effective method to suppress the low frequency vibration of the end-effector for industrial robots.

Unknown Disturbance Estimator Design for Non-Minimum Phase Plants Using Parallel Feed-Forward Model

A novel design method of an unknown disturbance observer for non-minimum phase plants is proposed in the present paper. In order to improve the estimation performance, we introduce the approach as a virtual augmented plant by adding a parallel model to the non-minimum phase real plant. The parallel model is designed so that the virtual augmented model becomes the minimum phase. Thus, it is possible to design the unknown disturbance estimator for the minimum phase plant but for the non-minimum phase plant. As the result, it is possible to improve the estimation performances. In this case, it is important to clarify the relationship between the unknown disturbance estimation signal for the real plant and the virtual augmented plant. In the present paper, the unknown disturbance estimation signal of the real plant is re-constructed by using the disturbance estimation of the virtual plant. And the parallel model design method is also proposed. The effectiveness of the proposed method is verified by numerical simulations for several mechanical vibration systems. The results show that the proposed method can improve estimation performances in comparison with conventional methods.

Robust Fault-Tolerant Control for Singular Bilinear Systems with Markovian Jump Related to Output Feedback

In this paper, output feedback controller is applied to solve robust fault-tolerant problem for singular bilinear system with Markovian jump. The aim of this problem is to design a output feedback controller that ensures stochastic stability of the closed-loop system and the -induced gain from the noise signal to the estimation signal remains bounded by a prescribed value, whether the actuators are normal or abnormal. The controller can be constructed by solving a set of linear matrix inequalities. An illustrative numerical example is given to demonstrate the applicability of the approach.

False Peaks Suppression based on Cross-High-Order Spectral QR Decomposition Approach

In signal processing, a frequently encountered problem is harmonic retrieval in additive colored noise, especially false peaks existence in harmonic signal peaks. The purpose of this paper is to develop an efficient approach to clear the false peaks based on cross-high-order spectral QR decomposition approach. Simulation results indicate that spectral density curve is smooth without false peaks existence. The methods have better in resolving power and performance than previous MUSIC approach. Thus, this approach is ideally suited for harmonic retrieval in additive colored noise and short data conditions, and is also accurate to estimation signal parameter in hybrid colored noises.