S. Satorres Martínez

Robotic Software Architecture for Multisensor Fusion System

As robotic systems are becoming more complex, distributed, and integrated, there is the need to build sophisticated frameworks to embody intelligence in the robot. This paper presents a component-based software architecture for the integration of industrial robotic platforms with multisensor systems where data fusion is relevant. The architecture is based on an open software robotic platform that is able to fulfill capabilities such as openness, interoperability, adaptability, and modularity, with the addition of hardware in the loop simulator. The resulting robot platform permits easy implementation of model- and sensor-based control concepts, proving to be interesting for robotics research. In addition, it allows the integration of standard industrial components. The entire system has been successfully developed, implemented, and demonstrated for robotic tasks, where multiple and different sensors are required.

A machine vision system for defect characterization on transparent parts with non-plane surfaces

This contribution presents a machine vision system capable of revealing, detecting and characterizing defects on non-plane transparent surfaces. Because in this kind of surface, transparent and opaque defects can be found, special lighting conditions are required. Therefore, the cornerstone of this machine vision is the innovative lighting system developed. Thanks to this, the defect segmentation is straightforward and with a very low computational burden, allowing real-time inspection. To aid in the conception of the imaging conditions, the lighting system is completely described and also compared with other commercial lighting systems. In addition, for the defect segmentation, a new adaptive threshold selection algorithm is proposed. Finally, the system performance is assessed by conducting a series of tests using a commercial model of headlamp lens.This contribution presents a machine vision system capable of revealing, detecting and characterizing defects on non-plane transparent surfaces. Because in this kind of surface, transparent and opaque defects can be found, special lighting conditions are required. Therefore, the cornerstone of this machine vision is the innovative lighting system developed. Thanks to this, the defect segmentation is straightforward and with a very low computational burden, allowing real-time inspection. To aid in the conception of the imaging conditions, the lighting system is completely described and also compared with other commercial lighting systems. In addition, for the defect segmentation, a new adaptive threshold selection algorithm is proposed. Finally, the system performance is assessed by conducting a series of tests using a commercial model of headlamp lens.

A sensor planning system for automated headlamp lens inspection

Usually, the automated systems for quality control based on computer vision have been centered on the design of algorithms for detecting different types of defects. Nevertheless, the issues related to planning suitable sensor poses for the inspection task have received less attention. In addition, the applications where a vision sensor can only sample a portion of a part from a single viewpoint, the sensor planning problem becomes critically important. This is the case of the automated inspection of vehicle headlamp lens, that due to its geometry and dimensions, requires multiple sensor poses to observe the whole part. Moreover, the customer requirements that define the maximum defect size should also be taken into account in the inspection process. This paper presents a vision sensor planning system applied to the quality control of headlamp lenses. The system uses the lens CAD, a vision sensor model and the customer requirements, included through a fuzzy approach, to achieve an optimal set of viewpoints. To compute the number and distribution of the viewpoints, a genetic algorithm is used. Experimental results demonstrate the effectiveness of the sensor planning system on commercial lenses.

A dynamic lighting system for automated visual inspection of headlamp lenses

This paper presents a novel lighting system to reveal surface defects on transparent parts with non-planar surface. Thanks to this system, the defect segmentation is straightforward and fast to compute and so a real-time inspection of these parts is possible. To aid in the conception of the imaging conditions, the lighting system is completely described and also a comparative study with others commercial lighting systems is done. Then, a new adaptive threshold selection algorithm for defect segmentation is proposed. Finally, the system performance is assessed using a commercial model of headlamp lens.

Model predictive position/force control of an anthropomorphic robotic arm

When a robotic manipulator executes a task in restricted and unknown environments, it is necessary to implement a position/force control in free space and constrained space. In classic controllers, position/force control may be carried out by means a controller switching, once the contact point has been reached. In this paper, a new position/force predictive controller for an anthropomorphic robotic arm is proposed. This work proposes an unique control law for controlling both variables (position or force). In a model predictive framework, where several goals have to be reached, it is known that cost function includes several parameters (weights) whose modification allows to prioritize goals to achieve. This fact has been taken into account to set up the controller. The proposed control law is validated by experiments involving a 7-dof anthropomorphic robotic arm executing a trajectory tracking task in a restricted environment. As experimental results show, it is possible to control position and force by adjusting controller weights.

Visual Predictive Control of Robot Manipulators Using a 3D ToF Camera

This paper presents a visual servo strategy for an anthropomorphic manipulator with eye-to-hand configuration. As novelty, a time-of-flight (ToF) camera, together with a model-based predictive control (MPC), have been used to execute the manipulator tasks in an unstructured environment. The ToF camera provides 3-D data of a target object and the MPC cost function achieves convergence of robot motion. With this configuration, there is no requirement to estimate a 3-D feature point position from a 2-D image and the camera model. A procedure for the target segmentation on the depth data, which is based on a previous training, is also presented. Experimental results, obtained in a robotic platform, show that the proposed approach is well suited for visual servoing purposes.

High-accuracy automatic system to assemble vehicle headlamps

In this work, the conception, design and development of a prototype for the automatic assembly and dimensional correction of vehicle headlamps is presented. Firstly, we analyze different measurement systems that can be used to determine the position of the elements that compounds the headlamp; secondly, we show the main conclusions obtained from the application of different technologies of actuators used to control the position of the components during the assembly. Finally, it is described the experimental results which validate the proposed system. They were carried out in both the lab and the assembly line in the factory. The new prototype solve the problem of assembling vehicle headlamps obtaining a final product with minimum dimensional errors.

Sensor planning for cars headlamps inspection using expert knowledge

Abstract This paper presents a vision-sensor-planning application for quality inspection of lens surfaces of car lighting systems. Using the CAD information, expert knowledge of the inspected parts and the camera model, the system automatically synthesizes optimal camera views. The optimal camera views are achieved by a hierarchical genetic algorithm. Results demonstrate the effectiveness of the vision-sensor-planning system.

Image fusion for surface finishing inspection

This paper presents an automatic defect detection system for machined metallic surfaces. Depending on the type of surface treatment a characteristic texture may be added. The aim is to determine flaws even if their orientation and shape are very similar to the surface finishing. For this purpose, a procedure based on merging features obtained under different lighting conditions has been developed. All the devices involved in the image acquisition process are also detailed. Results of the automated inspection suggest that the system works effectively with a low value of false rejections. Finally, ways to further improve the defect detection rate have also been discussed.

An automatic procedure to identify the areas of interest for the automated inspection of headlamp lenses

Many machine vision applications require manipulation of the acquired image to prepare it for processing. One type of pre-processing operation consists in identifying the areas of interest—or regions of interest (ROIs)—for the image analysis. But in occasions, the image areas to be analyzed may vary from one part to another. This is the case of the automated inspection of headlamp lenses that, due to the part positioning mechanism, the image areas to be processed do not always stand in the same image site. This paper presents an automatic procedure to adjust the region of interest in the images acquired for the inspection of a commercial model of headlamp lens. This proposal is included in a machine vision prototype for the automated inspection of such lenses. The results herein presented are achieved with this prototype.