Angel Valera

Force Estimation and Control in Robot Manipulators

Abstract In this paper we present some results on model based force estimation and how these estimates can be integrated in a common robot force control scheme. A generalization of the force estimation method proposed in [Hacksel and Salcudean, 1994] is done and a force observer able to follow ramp environmental forces is introduced. An extension of this method for robotic manipulators is also experimentally verified using an industrial robot with an open control system architecture during a force control task.

Dynamic parameter identification in industrial robots considering physical feasibility

The issue of identification of dynamic parameters in open-chain industrial manipulators is addressed with emphasis on the physical feasibility of the identified set of parameters. The dynamic model on which the identification procedure is based considers rigid-link robots including a complete actuator dynamics modeling and is obtained starting from the Gibbs–Appell equations. Friction at the joints is also considered. The dynamic equations of the model are written linearly with respect to the dynamic parameters to be identified. The matrix form linear system is solved through a quadratic optimization procedure with non-linear constraints in order to ensure the physical feasibility of the identified parameters. The procedure is tested using a PUMA 560 industrial robot. A comparison between control actions and torques obtained from the Inverse Dynamic Problem considering identified parameters is performed in order to establish the validity of the proposed procedure. The set of physically feasible dynamic parameters is used in an integration of the equations of motion of the robot and the results of the simulation are compared with the robot actual movement.

Model-Based Control of a 3-DOF Parallel Robot Based on Identified Relevant Parameters

This paper presents in detail how to model, identify, and control a 3-DOF prismatics-revolute-spherical parallel manipulator in terms of relevant parameters. A reduced model based on a set of relevant parameters is obtained following a novel approach that considers a simplified dynamic model with a physically feasible set of parameters. The proposed control system is compared with the response of a model-based control that considers the complete identification of the rigid-body dynamic parameters, friction at joints, and the inertia of the actuators. The control systems are implemented on a virtual and an actual prototype. The results show that the control scheme based on the reduced model improves the trajectory tracking precision when comparing with the control scheme based on the complete set of dynamic parameters. Moreover, the reduced model shows a significant reduction in the computational burden, allowing real-time control.

Multi Sensor Fusion Framework for Indoor-Outdoor Localization of Limited Resource Mobile Robots

This paper presents a sensor fusion framework that improves the localization of mobile robots with limited computational resources. It employs an event based Kalman Filter to combine the measurements of a global sensor and an inertial measurement unit (IMU) on an event based schedule, using fewer resources (execution time and bandwidth) but with similar performance when compared to the traditional methods. The event is defined to reflect the necessity of the global information, when the estimation error covariance exceeds a predefined limit. The proposed experimental platforms are based on the LEGO Mindstorm NXT, and consist of a differential wheel mobile robot navigating indoors with a zenithal camera as global sensor, and an Ackermann steering mobile robot navigating outdoors with a SBG Systems GPS accessed through an IGEP board that also serves as datalogger. The IMU in both robots is built using the NXT motor encoders along with one gyroscope, one compass and two accelerometers from Hitecnic, placed according to a particle based dynamic model of the robots. The tests performed reflect the correct performance and low execution time of the proposed framework. The robustness and stability is observed during a long walk test in both indoors and outdoors environments.

Mechatronic Development and Dynamic Control of a 3-DOF Parallel Manipulator

The aim of this article is to develop, from the mechatronic point of view, a low-cost parallel manipulator (PM) with 3-degrees of freedom (DOF). The robot has to be able to generate and control one translational motion (heave) and two rotary motions (rolling and pitching). Applications for this kind of parallel manipulator can be found at least in driving-motion simulation and in the biomechanical field. An open control architecture has been developed for this manipulator, which allows implementing and testing different dynamic control schemes for a PM with 3-DOF. Thus, the robot developed can be used as a test bench where control schemes can be tested. In this article, several control schemes are proposed and the tracking control responses are compared. The schemes considered are based on passivity-based control and inverse dynamic control. The control algorithm considers point-to-point control or tracking control. When the controller considers the system dynamics, an identified model has been used. The control schemes have been tested on a virtual robot and on the actual prototype.

Event-Based Localization in Ackermann Steering Limited Resource Mobile Robots

This paper presents a local sensor fusion technique with an event-based global position correction to improve the localization of a mobile robot with limited computational resources. The proposed algorithms use a modified Kalman filter and a new local dynamic model of an Ackermann steering mobile robot. It has a similar performance but faster execution when compared to more complex fusion schemes, allowing its implementation inside the robot. As a global sensor, an event-based position correction is implemented using the Kalman filter error covariance and the position measurement obtained from a zenithal camera. The solution is tested during a long walk with different trajectories using a LEGO Mindstorm NXT robot.

Technical Communique: PD control of robot manipulators with joint flexibility, actuators dynamics and friction

This paper presents a PD controller for robot manipulator considering simultaneously joint flexibility, actuators dynamics as well as friction. The convergence analysis of the proposed PD controller is carried out using passivity. Global convergence to a desired fixed angular position is proved using only angular position and velocity measurements.

Integración de dispositivos físicos en un laboratorio remoto de control mediante diferentes plataformas: Labview, Matlab y C/C++

Remote control laboratories allow to work and made experiments with real physical devices that are located in remote environments. This equipment is accessible by students at any time and from anywhere with an Internet connection. In this paper a set of physical devices embedded within an overall scheme of remote laboratories (AutomatL@bs) is presented. By means of this remote laboratory the extudent can accomplish the experiments remotely. Special attention to the different possibilities of communication and interconnection between the interface and the physical device will be provided. The interface is made by means of Easy Java Simulations, so the direct connection between this and the physical devices located in remote laboratories will be analyzed.

Efficient Lens Distortion Correction for Decoupling in Calibration of Wide Angle Lens Cameras

In photogrammetry applications, camera parameters must be as accurate as possible to avoid deviations in measurements from images. Errors increase if wide angle lens cameras are used. Moreover, the coupling between intrinsic and extrinsic camera parameters and the lens distortion model influences the result of the calibration process notably. This paper proposes a method for calibrating wide angle lens cameras, which takes into account the existing hard coupling. The proposed method obtains stable results, which do not depend on how the image lens distortion is corrected.

Calibration of a trinocular system formed with wide angle lens cameras

To obtain 3D information of large areas, wide angle lens cameras are used to reduce the number of cameras as much as possible. However, since images are high distorted, errors in point correspondences increase and 3D information could be erroneous. To increase the number of data from images and to improve the 3D information, trinocular sensors are used. In this paper a calibration method for a trinocular sensor formed with wide angle lens cameras is proposed. First pixels locations in the images are corrected using a set of constraints which define the image formation in a trinocular system. When pixels location are corrected, lens distortion and trifocal tensor is computed.