Safety is one of important factors in control of mechatronic systems interacting with humans. In order to evaluate the safety of such systems, mechanical impedance is often utilized as it indicates the magnitude of reaction forces when the systems are subjected to motions. Namely, the mechatronic systems should have low mechanical impedance for improved safety. In this paper, a methodology to design controllers for reduction of mechanical impedance is proposed. For the proposed controller design, the mathematical definition of the mechanical impedance for open-loop and closed-loop systems is introduced. Then the controllers are designed for stable and unstable systems such that they effectively lower the magnitude of mechanical impedance with guaranteed stability. The proposed method is verified through case studies including simulations.

This paper presents an approach for fast modeling and identification of robot dynamics. By using a data-driven machine learning approach, the process is simplified considerably from the conventional analytical method. Regressor selection using the Lasso (l 1 -norm penalized least squares regression) is used. The method is explained with a simple example of a two-link direct-drive robot. Further demonstration is given by applying the method to a three-link belt-driven robot. Promising result has been demonstrated.

For robots with joint elasticity, discrepancies exist between the motor side information and the load side (i.e., end-effector) information. Therefore, high tracking performance at the load side can hardly be achieved when the estimate of load side information is inaccurate. To minimize such inaccuracies, it is desired to calibrate the load side sensor (in particular, the exact sensor location). In practice, the optimal placement of the load side sensor often varies due to the task variation necessitating frequent sensor calibrations. This frequent calibration need requires significant effort and hence is not preferable for industries which have relatively short product cycles. To solve this problem, this paper presents a sensor frame identification algorithm to automate this calibration process for the load side sensor, in particular the accelerometer. We formulate the calibration problem as a nonlinear estimation problem with unknown parameters. The Expectation-Maximization algorithm is utilized to decouple the state estimation and the parameter estimation into two separated optimization problems. An overall dual-phase learning structure associated with the proposed approach is also studied. Experiments are designed to validate the effectiveness of the proposed algorithm.

This paper presents the dynamics modeling and dynamic identification of a dual-blade wafer handling robot. An explicit form dynamic model for this 8-link parallel robot is proposed. The dynamic model is transformed into a decoupled form to enable dynamic parameters identification with least-square regression. A well conditioned trajectory is chosen for identification experiment. Both viscous friction and Coulomb friction are considered to make the model more reliable. Model has been validated by experiments.

Traditional kinematic analysis of manipulators, built upon a deterministic articulated kinematic modeling often proves inadequate to capture uncertainties affecting the performance of the real robotic systems. While a probabilistic framework is necessary to characterize the system response variability, the random variable/vector based approaches are unable to effectively and efficiently characterize the system response uncertainties. Hence in this paper, we propose a random matrix formulation for the Jacobian matrix of a robotic system. It facilitates characterization of the uncertainty model using limited system information in addition to taking into account the structural inter-dependencies and kinematic complexity of the manipulator. The random Jacobian matrix is modeled such that it adopts a symmetric positive definite random perturbation matrix. The maximum entropy principle permits characterization of this perturbation matrix in the form of a Wishart distribution with specific parameters. Comparing to the random variable/vector based schemes, the benefits now include: incorporating the kinematic configuration and complexity in the probabilistic formulation, achieving the uncertainty model using limited system information (mean and dispersion parameter), and realizing a faster simulation process. A case study of a 6R serial manipulator (PUMA 560) is presented to highlight the critical aspects of the process. A Monte Carlo analysis is performed to capture the deviations of distal path from the desired trajectory and the statistical analysis on the realizations of the end effector position and orientation shows how the uncertainty propagates throughout the system.

This paper presents the formulation of a reduced-order linear discrete–path approximation in state space and its solution as a function of path lengths for a 3D curvature-based beam model (CBM). Solutions to both forward and inverse problems are discussed; the former is essential for real-time deformed shape visualization whereas the latter is much needed for haptic force feedback. The method is illustrated with an application example where a 2D beam is characterized by a 6 th order CBM. Practical implementation shows that when external forces as system input are expressed in global coordinates, the CBM can be decoupled into two 2 nd order systems enabling parallel computing of the deformed shape and the orientation and moment, and effectively reducing the table size for storing the operating conditions. The proposed real-time computation method has been validated by verifying results against published experimental and MSM simulated data.