Content uploaded by Bruno Siciliano
Author content
All content in this area was uploaded by Bruno Siciliano on Dec 14, 2014
Content may be subject to copyright.
The Potential of Model-Based Control Algorithms for Improving Industrial Robot Tracking Performance
Abstract
First Page of the Article
... The dynamics calibration of the robot, as well as model-based control strategies can improve the performance of the machine reducing the tracking error [14]. The robot dynamics model can also be extended even considering the joints and links stiffness defining the also so-called extended flexible joint dynamics model [15]. ...
Nowadays robotic machining is considered an immediate feasible alternative to a CNC machine where material removal is required. The main advantages of using industrial robots (IRs) instead of CNC machine is the lower purchase cost and the wider working volume. Despite great interest from industrial end-users and a wide range of possible applications a real IRs usage in machining applications tend to remain a niche of the market. As reported in Iglesias:2015, within the total sales of 2011, the percentage of industrial robots used in material removal operations is less than 5%. Despite this information is a bit dated (2011), it provides an idea of the IRs market trends.
Although, during the years, IRs have made great advances in terms of performance (improvements in positioning accuracy and reduction of tracking error during high dynamics movements) they still present lacking to face machining applications. The problems to be addressed are multiple ranging from kinematic inaccuracies compensation (i.e. kinematics calibration), robot dynamics performance improvements (i.e. dynamics calibration) and the evaluation of the influence on robotic structure of the technological process.
Within this context, this thesis introduces and investigates new techniques for the calibration dynamics model. Such techniques are used to estimate dynamics parameter set. The calibrated models are used to implement model-based control strategies (external with respect to the robot controller) and applied in machining applications.
Starting from the state-of-the-art on dynamic modelling, this work evaluates both rigid body lumped parameters model and elasto-dynamics model (single joint model).
In particular we introduce the concept of local dynamics model calibration. The idea is to calibrate the dynamics model in a sub-region of the robot workspace achieving high torques prediction accuracy only in a predefined workspace bounded area. Such approach is motivated from the analysis of machining tasks, that are usually defined in a constrained workspace sub-region.
In the thesis we also propose an innovative calibration procedure for the rigid body dynamics model addressing all the issues related to design of exciting signal for identification experiments. The calibration algorithms and the dynamics parameters set, experimentally estimated, have been validated on two practical cases.
In the first case, we have used the rigid body dynamic model to realize a virtual force sensor based on the motor extra-currents estimation. Such sensor, exploiting the locally calibrated model, is able to accurately predict external forces that acts on robot end-effector (EE) by measuring only motors torques and joints positions. The external forces estimated can be used as a feedback signal in a force control loop. The aim is to measure and control the force exchanged between the robot and the environment during a polishing task.
In the second practical case we have investigated the use of the joint stiffness matrix for the optimization of a machining path. In particular, the kinematics configuration of the robot, assumed during the execution of a milling path, is optimized on the base of the robot compliancy and on the base of the milling forces action.
Finally, last chapter of the thesis discusses the use of an IR for hard material deburring. In this chapter, we introduce a control algorithm based on a human mimicking strategy. In other words, trying to replicate a human worker behaviour, the use of an IR is enabled for the deburring of hard material (i.e. a not trivial robotic application). The aim is to demonstrate how a smart use of an IR allows to obtain good results, also without using model-based control strategies, providing an industrial ready-to-use solution.
In summary, the thesis deals with the analysis of IRs dynamics modelling and calibrations. We propose a novel and effective strategies for dynamics parameters estimation. The dynamics models calibrated in this way are able to provide accurate torques prediction in bounded workspace sub-regions. We have proven the effectiveness of these calibration methodologies addressing two different machining task. We have also investigated the use on an IR in a third machining application proposing a smart use of the manipulator without using any model-based control strategies.
... For example, model-predictive control [10], intended as one of its many nuances such as GPC [11] or receding-horizon control [12] have proved to be very effective for achieving at the same time high speed and pronounced vibration damping [13,14]. In general the importance and the possible performance improvement brought by nonlinear and modelbased strategies is well recognized by the roboticist community [15]. ...
In this paper, a simple nonlinear control strategy for the simultaneous position tracking and vibration damping of robots is presented. The control is developed for devices actuated by speed-controlled servo drives. The conditions for the asymptotic stability of the closed-loop system are derived by ensuring its passivity. The capability of achieving improved trajectory tracking and vibration suppression is shown through experimental tests conducted on a three-axis Cartesian robot. The control is aimed to be compatible with most industrial applications given the simplicity of implementation, the reduced computational requirements, and the use of joint position as the only measured signal. © 2016 Higher Education Press and Springer-Verlag Berlin Heidelberg
... It is widely demonstrated [1], [2] that the design of modelbased controllers represents a feasible solution to improve the performance of manipulators in terms of tracking error accuracy at high speed movements. Due to this reason, the study of methodologies for the identification of the robot dynamic model have been deeply investigated during last three decades. ...
The identification of the dynamic model of a robotic manipulator represents a fundamental step for designing high performance model-based controllers. Despite the huge number of works presented on this topic, the symbolic dynamic model reduction (i.e., the identification of the set of parameters observable through the measure of joint torques and positions) still remain a challenging task, characterized from tailored solutions, adapted from time to time to specific families of mechanisms. The work here presented, introduces an automatic and analytical reduction of the dynamic model, based on a multi-dimensional Fourier series decomposition of the dynamic equations. The procedure enables to obtain symbolically the base dynamic parameters (BP) starting from a given kinematic structure. The Fourier based model reduction can be applied indifferently both to open-and closed-chain kinematics. A simulated example shows the effectiveness of the proposed algorithm.
... It is long [1], [2], and generally accepted that the dynamic calibration of industrial robots (IRs) is of utmost importance in increasing the predictability and accuracy of model-based control strategies. Since early works [3], [4], many researchers have investigated some methodologies all involving a linear reduction of the rigid-body model into a minimum set of lumped dynamic parameters to be estimated [5], e.g. a complete-observable linear map from joint position, velocity and acceleration to motor torques. ...
Currently, opposite to other manufacturing sectors like automotive, food or metal processing, robots are still out of Footwear industry; only technical shoe producers have introduced robots to assist in the injection moulding process but there are not other relevant applications in use. The described difficulties in shoe manufacturing ask for the use of state of the art robotic technology to reach the required degree of flexibility, autonomy and dexterity in order to make the robot a valuable resource for footwear companies. In this scenario, the paper describes an innovative robotic cell for robotized roughing operations. The robot controller has been designed to be easily integrated with standard industrial controller by exploiting the so-called sensor tracking option provided by many different robot suppliers. Experimental results are reported and demonstrate the feasibility of the solution proposed.
... It is long [1], [2], and generally accepted that the dynamic calibration of industrial robots (IRs) is of utmost importance in increasing the predictability and accuracy of model-based control strategies. Since early works [3], [4], many researchers have investigated some methodologies all involving a linear reduction of the rigid-body model into a minimum set of lumped dynamic parameters to be estimated [5], e.g. a complete-observable linear map from joint position, velocity and acceleration to motor torques. ...
Notwithstanding The research on dynamic modelling of Industrial Robots (IRs hereafter) covers The last Three decades, improvements are necessary To enable IRs adoption in Technological Tasks where high dynamics or interaction with environment is needed, e.g. deburring, milling, laser cutting etc. Indeed, This class of applications displays even more The necessity of high-Accuracy Tracking especially in workspace sub-regions, while common IR dynamic calibration methods often span The workspace at large (in Term of positions and high velocities) resulting in an averagely fitting models. Open issues are Therefore on The applicability/scalability of standard methods in workspace sub-regions and on The metrics used for The calibration performance evaluation. The paper proposes an algorithm designed To high-Accuracy local dynamic identification, comparing it with The results achievable by a common IRs dynamic calibration method and by The same method scaled To a workspace sub-region. In addition, unlike from standard, The here reported experimental comparison is made by evaluating The Torque prediction error for IRs robot moving along path programmed by standard/commercial IR motion planner and not along path belonging To The same Template-class of Trajectory used in identification phase.
Dynamic compensation control algorithms can improve the performance of conventional industrial manipulators if an accurate dynamic model is available. It is true that usually the structure of the dynamic model is known but the actual values of the dynamic parameters are either unknown or not accurately known. It is thus required to perform an identification of these parameters based on measurements made on the manipulator during assigned motion. This paper describes the identification of the dynamic parameters for a conventional industrial robot. Algorithms are available to perform the identification when ideal conditions are present. Despite of the fact that the conditions for the experiments are far from those provided by research set-ups, the obtained results have confirmed that identification of dynamic parameters can be effectively performed even on an industrial set-up.
It is well-known that an accurate knowledge of the dynamic model of a manipulator can significantly improve the trajectory tracking capabilities by resorting to model-based control algorithms. While the structure of the dynamic model can be simply obtained by considering rigid- body motion laws, the actual values of the dynamic parameters (inertias, masses, centers of gravity) are either unknown or not accurately known. Algorithms have been proposed which estimate these parameters from measurements taken during manipulator motion. This paper deals with identification of dynamic parameters for a conventional industrial manipulator, performed via an algorithm based on the energy theorem for mechanical structures. The purpose here is to show that such an identification leads to good results even with all the limitations set by the conventional industrial set-up. Also, the results are compared with those obtained with identification algorithms based on the dynamic model of the manipulator.
Dynamic modelling of gear-driven rigid robot manipulators is discussed in this work. The dynamic effects of the motion of the motors driving the joints through gears are analyzed. A complete model is derived using the Lagrange formulation in which the contributions of rotor inertias and rotor-link interactions are evidenced. The resulting equations of motion are shown to be linear in terms of a suitable set of dynamic parameters for the augmented links (links with motors). These are utilized for model derivation using the recursive Newton-Euler formulation. The example of an elbow manipulator is developed.
State-of-the-art robots do not perform up to their physical speed or load capabilities. Currently, performance is limited by the form of control implemented, which gives rise to speed and load dependent errors and overshoots. In this paper, a new control algorithm is proposed which entails smooth real time reference paths and an adaptive feedforward path-tracking algorithm. The result is high accuracy independent of speed and load. With the trade-off between speed and accuracy eliminated, the full operational capabilities of the axis motors can be exploited. The utility of these principles is shown in a video tape demonstration of a new controller for a Unimation PUMA robot.
The manipulator trajectory tracking control problem revolves around computing the torques to be applied to achieve accu rate tracking. This problem has been extensively studied in simulations, but real-time results have been lacking in the robotics literature. In this paper, we present the experimental results of the real-time performance of model-based control algorithms. We compare the computed-torque control scheme with the feedforward dynamics compensation scheme. The feedforward scheme compensates for the manipulator dy namics in the feedforward path, whereas the computed-torque scheme uses the dynamics in the feedback loop for lineariza tion and decoupling. The parameters in the dynamics model for the computed-torque and feedforward schemes were esti mated by using an identification algorithm. Our experiments underscore the importance of including the off-diagonal terms of the manipulator inertia matrix in the torque compu tation. This observation is further supported by our analysis of the dynamics equations. The manipulator control schemes have been implemented on the CMU DD arm II with a sam pling period of 2 ms.
At the DFVLR Institute of Flight System Dynamics a mathematical model of the dynamics of the Manutec r3 industrial robot has been worked out in detail. This model has different degrees of complexity, depending on the effects one is interested in investigating. In this report the DFVLR model 1, which is a multibody model with 6 degrees of freedom, and the DFVLR model 2, which is a model with 3 degrees of freedom, are described. The model concept is outlined, the numerical values of all the model parameters are given, and FORTRAN 77 subroutines are documented to numerically evaluate the equations of motion of the two models. Most of the model parameters have been acquired by measurements in the laboratory. The FORTRAN 77 subroutines are generated symbolically by a multibody program. These mathematical models may be used for computer-aided dynamic studies, such as, for example, controller design, dynamic trajectory planning and dynamic simulation.
The feedback algorithm widely employed in the various current control strategies is a digital implementation of a conventional proportional plus derivative (PD) control. For dynamic control of robots, this algorithm can be shown to lead to unsatisfactory trade-offs between static accuracy, system stability, insensitivity to model structure inaccuracies and high frequency noise rejection. These trade-offs become even more serious as sampling rate decreases. To reduce these trade-offs, a more realistic discrete time system model of a robot arm and a more sophisticated feedback compensator are required. In this paper, a discrete time system model of a robot arm is derived and limitations of PD controllers are shown. They are consistent with our simulation and experimental results. Also a Lag-lead compensator is designed by a frequency-response analysis based on this model, along with the algorithm for realizing this compensator. Simulations and tests are conducted, which show encouraging results in reducing the trade-offs.
Experimental evaluations have demonstrated the effect of dynamic compensation on the trajectory tracking accuracy of industrial manipulators. High gear ratios reduce the compu tational complexity of the system dynamics but do not elimi nate the requirement for accurate modeling. For the PUMA-560 case study, the motor dynamics were experimen tally determined and combined with link dynamics to accu rately model the complete system. The impact from compen sating for specific manipulator dynamics by feedforward and computed-torque controllers was then experimentally deter mined. Feedforward dynamic compensation reduced the peak trajectory tracking errors of an independent joint Pro portional Derivative controller by a factor of three. The com bination of high sample rates and feedforward dynamic compensation produced a control algorithm with excellent peak and final error.