Geir Hovland

Skill acquisition from human demonstration using a hidden Markov model

A new approach to skill acquisition in assembly is proposed. An assembly skill is represented by a hybrid dynamic system where a discrete event controller models the skill at the task level. The output of the discrete event controller provides the reference commands for the underlying robot controller. This structure is naturally encoded by a hidden Markov model (HMM). The HMM parameters are obtained by training on sensory data from human demonstrations of the skill. Currently, assembly tasks have to be performed by human operators or by robots using expensive fixtures. Our approach transfers the assembly skill from an expert human operator to the robot, thus making it possible for a robot to perform assembly tasks without the use of expensive fixtures.

TCSC Allocation Based on Line Flow Based Equations Via Mixed-Integer Programming

Research effort has been given to locate the optimal locations of thyristor-controlled series capacitor (TCSC) and their initial compensation levels using mixed-integer programming (MIP). As a useful technique for combinatorial optimisation over integer and continuous variables, the MIP approach can provide robust performance as well as high computational efficiency while solving complex optimal problems. Previous work using MIP employed dc load flow model ignoring reactive power balance, power loss and transformer tap ratios. In this paper, a new planning method is developed based on recently reported line flow equations and basic linearisation of binary-continuous products. The objectives of the planning strategy are to improve system loadability, voltage profile in the network, as well as to minimise the investment cost by choosing proper locations and settings of devices. Simulation results are presented and discussed for IEEE 9-, 57-, 118-, and 300-bus systems.

Dynamic sensor selection for robotic systems

A new technique for selecting, in real time, different sensing techniques for a robotic system has been developed. The proposed method is based on stochastic dynamic programming, which provides an effective solution to multi-stage decision problems. At each stage in the decision process a sensor selection controller has the option of consulting a new process monitoring technique to improve the knowledge of the task or terminating the decision process without any further information gathering. The sensor selection controller has been successfully implemented for the real-time control of a planar robotic assembly task in a discrete event control framework. One of the monitoring methods used is based on hidden Markov models, where the average recognition rate was 87%. The rate of 87% was chosen to show the effectiveness of the dynamic sensor selection method. The experiments show that the method performs better than any individual process monitor. A successful event recognition rate of 97% with an average CPU time of 0.38 seconds is achieved when two force monitors and one position monitor are available to the sensor selection controller.

Compliance error compensation technique for parallel robots composed of non-perfect serial chains

The paper presents the compliance errors compensation technique for over-constrained parallel manipulators under external and internal loadings. This technique is based on the non-linear stiffness modeling which is able to take into account the influence of non-perfect geometry of serial chains caused by manufacturing errors. Within the developed technique, the deviation compensation reduces to an adjustment of a target trajectory that is modified in the off-line mode. The advantages and practical significance of the proposed technique are illustrated by an example that deals with groove milling by the Orthoglide manipulator that considers different locations of the workpiece. It is also demonstrated that the impact of the compliance errors and the errors caused by inaccuracy in serial chains cannot be taken into account using the superposition principle.

Kinematic error calibration of the Gantry-Tau parallel manipulator

This paper presents a new error kinematic model containing twelve parameters for the three linear actuators of the Gantry-Tau parallel kinematic manipulator. Three different types of measurement equipment, laser interferometers, linear encoders and double-ball bars, can successfully be used to calibrate the linear actuators. A method for selecting an optimal set of Cartesian coordinates is presented and a small set of fifty coordinates have been found to be sufficient. The methods presented in this paper allow for a fast and easy set up calibration procedure of the Gantry-Tau after assembly and installation, which is a pre-requisite for flexible automation with parallel kinematic manipulators

Frequency-domain force measurements for discrete event contact recognition

Discrete event recognition based on force measurements in the frequency-domain, is presented. The force signals arise from interaction between the workpiece and the environment in a planar assembly task. The discrete events are modeled as hidden Markov models (HMMs), where the models are trained off-line with the Baum-Welch re-estimation algorithm. After the HMMs have been trained, we use them online in a robotic system to recognise discrete events as they occur. Event recognition with an accuracy as high as 98% was accomplished in 0.5-0.6 s with a relatively small training set.

Combining force and position measurements for the monitoring of robotic assembly

A method for combining dynamic force and static position measurements for the monitoring of assembly is presented. A multilayer perceptron (MLP) network is used as a classifier where the individual network outputs correspond to contact state transitions occuring during the assembly process. When a contact state transition occurs, the MLP output with the largest value is chosen. The recognised contact state is sent to a discrete event controller which guides the workpiece through a series of contact states to the final desired configuration. The MLP has been successfully implemented on a Motorola 68040 based VxWorks board with successful recognition rates of 94.4% and 92.0% on a training set and an independent test set, respectively.

Hybrid dynamic modeling and control of constrained manipulation systems

Discrete event systems are presented as a powerful framework for a large number of robot control tasks. This paper presents a general description of the discrete event modeling and control synthesis for robot manipulation. Additionally, methods for the effective monitoring of the process based on the detection and identification of discrete events are given. The effectiveness and versatility of the approach are demonstrated through a wide variety of experiments. Applications are demonstrated in assembly, online training of robots, advanced perception capabilities, human-robot shared control and the understanding of human manipulation skills.

Automatic elasticity tuning of industrial robot manipulators

We present a method for automatic elasticity tuning of industrial robot manipulators. The main contributions of the work are: a) The parameters of a mechanical mass-spring-damper equivalent of any order are solved given, only partial state information (motor encoder position and motor torque). b) The method is fully automatic with no operator input and can easily be applied in the field to update the dynamic model parameters. The ability to automatically update the elasticity parameters is particularly useful when the robot operators mount flexible tooling or equipment on the robot arms. c) The method separates friction and elasticity identification. d) The method is demonstrated on an industrial ABB robot. e) We combine an important result from the vibration literature (Yam and Elhay, 1996), with the solution of inverse eigenvalue problems (Gladwell, 1986). To our knowledge, this is the first time that these methods have been combined and applied to the identification of flexible robot manipulators. The main advantage of the method compared to other identification methods is the fact that only motor encoder position and motor torque are required to identify the springs, masses and dampers of an Nth order system.

Collision-Free Workspace Design of the 5-Axis Gantry-Tau Parallel Kinematic Machine

This paper describes the workspace of the 5-axis variant of the Gantry-Tau parallel manipulator. The 5-axis movements are achieved by extending the 3-axis Gantry-Tau with two linear telescope link actuators. A new analytic inverse kinematic solution for the 5-axis machine and an analysis of the collision-free zones of the machine are presented. The results show that by careful design of the manipulated platform, a collision-free 5-axis workspace can be achieved. The achievable re-orientation of 30 degrees in the entire workspace is large for a parallel kinematic machine