Jochen Kerdels

Realtime motion compensation for ROV-based tele-operated underwater manipulators

This paper presents a novel underwater movement compensation algorithm for stabilization of manipulator position utilizing not ROV movements for disturbance comensation, but overlaid manipulator movements. A model based estimator is used to predict vehicle movement and provide the manipulation system with the necessary time to compensate for the estimated motion. It describes the conceptual benefits of this approach compared with common station-keeping algorithms, and shows how previous methods can be combined with the new approach in order to further improve manipulator position accuracy. The method is validated in a number of experiments, which show its feasability and outstanding performance.

A Robust Vision-Based Hover Control for ROV

The main field of application for small- and middle-class ROVs is the inspection of underwater structures or other objects of interest. Approaching such an object, one would want to hold a steady position in front of the object to study it in detail without having to concentrate on the control of the vehicle. This kind of hover control could be implemented by using an inertial measurement unit (IMU), but most of the small-and middle-class ROVs do not have one. Furthermore, even the best IMUs tend to drift. On this account our approach, which is presented in this paper, uses video data to estimate the movements of the vehicle and uses these data to keep the vehicle hovering in front of a particular structure. The used vision algorithms are aimed at real world applications and are robust enough to handle various light and visibility conditions.

A practical underwater 3D-Laserscanner

A number of attempts have been made to use the benefits of 3D-Laserscanning techniques in the underwater environment. Unfortunately, due to a number of operative problems with such devices, their accuracy and therefore applicability remains quite low. This paper specifically focuses on these practical issues by expanding on previous works in this area and improving their usability. The result is a calibration procedure for triangulation-based 3D-laserscanners for the underwater environment which provides a very promising precision and reliability, but at the same time does not demand exaggerated deployment overhead.

C-Manipulator: An Autonomous Dual Manipulator Project for Underwater Inspection and Maintenance

We present the new project C-Manipulator (funded by the German Ministry of Economics (BMWI), Grant No. 03SX231). The goal of C-Manipulator is the development of an autonomous, modular, dual manipulator system for underwater applications. This paper provides an overview over the project. It explains shortly the relevance of autonomous underwater manipulation. Then it describes briefly the state-of-the-art, explains the new vision-based control approach featuring visual servoing techniques and the planned manipulator system design featuring the Sub-C Network. Furthermore, a new developed indoor test-bed using a gantry crane for UUV-simulation is introduced, which will be used to test the manipulator system under realistic conditions and to prepare the system for a final test in the Baltic sea, which is planned for 2009.Copyright

Sensorless Computer Control of an Underwater DC Manipulator

Virtually all manipulators used on todays underwater remote operated vehicles (ROV) are controlled manually by human operators. Thus, extensive sensory information about speed, position, and load for the different joints of the actuator is often not needed. Especially the de-manipulators on small- and middle-class ROVs often offer no sensory information at all. Given that we use such a de-manipulator among others in a research project dealing with computer-controlled autonomous underwater manipulation, we needed the speed and position information and an estimation of the force applied by the manipulator, especially of the gripper. As the refit of an existing underwater manipulator with additional sensors is quite challenging and error-prone, we developed a sensorless control for the dc-manipulator based on the back EMF of the actuators.

Image segmentation based on height maps

In this paper we introduce a new method for image segmentation. It is based on a height map generated from the input image. The height map characterizes the image content in such a way that the application of the watershed concept provides a proper segmentation of the image. The height map enables the watershed method to provide better segmentation results on difficult images, e.g., images of natural objects, than without the intermediate height map generation. Markers used for the watershed concept are generated automatically from the input data holding the advantage of a more autonomous segmentation. In addition, we introduce a new edge detector which has some advantages over the Canny edge detector. We demonstrate our methods by means of a number of segmentation examples.

A Possible Encoding of 3D Visual Space in Primates

Killian et al. were the first to report on entorhinal neurons in primates that show grid-like firing patterns in response to eye movements. We recently demonstrated that these visual grid cells can be modeled with our RGNG-based grid cell model. Here we revisit our previous approach and develop a more comprehensive encoding of the presumed input signal that incorporates binocular movement information and fixation points that originate from a three-dimensional environment. The resulting volumetric firing rate maps exhibit a peculiar structure of regularly spaced activity columns and provide the first model-based prediction on the expected activity patterns of visual grid cells in primates if their activity were to be correlated with fixation points from a three-dimensional environment.

A Noise Compensation Mechanism for an RGNG-Based Grid Cell Model

Grid cells of the entorhinal cortex provide a rare view on the deep stages of information processing in the mammalian brain. Complementary to earlier grid cell models that interpret the behavior of grid cells as specialized parts within a system for navigation and orientation we developed a grid cell model that facilitates an abstract computational perspective on the behavior of these cells. Recently, we investigated the ability of our model to cope with increasing levels of input signal noise as it would be expected to occur in natural neurobiological circuits. Here we investigate these results further and introduce a new noise compensation mechanism to our model that normalizes the output activity of simulated grid cells irrespective of whether or not input noise is present. We present results from an extended series of simulation runs to characterize the involved parameters.

A topology-independent similarity measure for high-dimensional feature spaces

In the field of computer vision feature matching in high dimensional feature spaces is a commonly used technique for object recognition. One major problem is to find an adequate similarity measure for the particular feature space, as there is usually only little knowledge about the structure of that space. As a possible solution to this problem this paper presents a method to obtain a similarity measure suitable for the task of feature matching without the need for structural information of the particular feature space. As the described similarity measure is based on the topology of the feature space and the topology is generated by a growing neural gas, no knowledge about the particular structure of the feature space is needed. In addition, the used neural gas quantizes the feature vectors and thus reduces the amount of data which has to be stored and retrieved for the purpose of object recognition.

Sensor Proccessing and Behaviour Control of a Small AUV

This paper presents the design of the electronics, the sensor data processing scheme and the implementation of the behaviour based control for a small autonomous underwater vehicle (AUV). The robot has a volume of 12dm3 and is used as a demonstration vehicle for underwater ai-technologies. The thruster configuration allows the μAUV to hover and its active light sensors enable obstacle detection abilities. The control features several behaviours like obstacle detection and depth sensor calibration and is completely autonomous. An ATMegal28 functions as onboard CPU and due to the limited computing power we use a special scheduling algorithm which also acts as the behaviour arbiter.