The mechanic apprentices of the company Rieter AG manufacture during their four years apprenticeship a robot with 5 axes. The ASM robot is a folding arm robot to be able to be moved by DC-engines and stepping motors. At the end of the apprenticeship the robot can be controlled only by two joysticks. Any regulation or intelligent InterfaceE are missing.

In a practical training work a controller was developed that rules the robot over this single interface. The controller consists of a micro controller system with a Infineon 16Bit micro Controller C167. Target of this thesis (diploma) is to test the controller and to equip the individual function groups. Further a software should be developed, that provides programming the controller by a Host Software on computer, or even on the controller itself. The input of the positions should be possible to give by TeachIn. The software consists of a real-time-multitasking-operation-system RTX166 as well as a regulation, that runs parallely and that is indenpendent from the Operation System (OS). The Operation System takes over the menu guidance and the communication to the Host Software on the computer. Now a redesign is necessary, which settles the errors. The regulation as well as a part of the menu guidence are implemented in the software. The storing of scripts and functions to manipulate the positions in view to editing by the menu guidance are implemented. Then, the real time system RTX166 is completely executable in the targetsystem. It is possible to give positions which are getting reached. Work are still to be carried out in detail and the menu guidance must still be completely implemented.

The EPF Lausanne is developing a robot for the EXPO'02 whose task it is to speak to visitors by itself. If the visitors would like to get a guided tour around the exhibition room, they just have to tell the robot. For the localization of people a camera is installed in one eye. Two students of the Zurich University of Applied Sciences (ZHW) work on face detection to localize people. The idea of face detection is to look at the picture from the cam for skin colored areas. Many things in a room can be skin colored and therefore will be detected.

The main task of the diploma project is to develop a program which decides if the skin colored area found in a picture is a face or not.

The difference in brightness of a found area gives some information about where the eyes and the mouth are. A statistic is applied on the region. It considers the contour of the skin color mask for covering the background. The results of the stat are the most interesting places, where the eyes and the mouth might be. The interesting places are divided into three categories, possible eyes, possible mouth and interferences. The decision, whether the detected area is a face or not is based on an interpretation of the interesting places.

The face detection works as the students assumed. Faces can be separated from other skin colored regions in different illuminations. The Boolean result, whether a region is a face, is not absolutely reliable. The test with the accumulated pictures has been successful. The rate of reliability in determining whether a skin colored area is a face could be increased from 40% to 80%. The disadvantage is that every fourth head is lost and cannot be detected as a face.

To be absolutely certain to detect only faces this algorithm should be combined with another sensor. In this case a combination with the laser sensor of the robot which scans the environment might be the easiest and most powerful solution.

The deltarobot is an important development among the category of parallel-robots. It consists of three arms which at the end are linked with each other by a tool-center point (TCP). The actuators are on the basic platform fixed. Consequently the mass to be moved is minimal and therefore allows for high acceleration.

It was the object of this thesis to develop a model of a deltarobot with which to demonstrate its basic functions. According to specifications the mechanical part was realised with Lego components.

In a first step the robot had to be enabled to find target positions. Based on this a simple "pick and place" function had to be developed.

The first step was realised by applying the theory of inverse kinematics, which describes the calculation of the pin-joint angles if the TCP-coordinates are given. This calculation requires the application of higher mathematical functions which are not integrated in the software tools at use. To approximate the respective functions algorithms were implemented and tested for adequate accuracy.

After this it was possible to programme the control software containing the calculation of the pin-joint angles, the initialisation and the position control.

The realisation of the Pick and Place function was solved with a second Lego-computer (RCX) as there were not enough ports at disposal. This one is used to control the gripping device. To establish a coordinated cooperation between the finding of the target position and the gripping device, both RCX communicate via infrared port. In contrast to the first programme the position is not calculated by the same RCX as it is initiated. Now it is the master which determines the position and sends it to the slave.

To demonstrate the Pick and Place two cylinders are shown rotating in a square.

The project was successfully completed and the product was delivered to our industrial partners.

Modern image processing is gaining increased importance for many medical applications. Under the aspect of ?just in time? care which is being required more and more today, the analysis-methods are becoming more precise, complex and costly. In order to comply to the requirements of modern society, the doctors try to do their examinations locally. This diploma-thesis was written under the aspect of satisfying the demand for economical measurement devices.

In this paper the realisation of a system for reading depth-data and thereby producing a grid model of an object is described. As triangulation?methods are used, the 3D structure can be fond without any mechanical contact to the object. The depth data is found with an optical 3D measuring method which uses an active light source (line laser) to create a line (light section) on the object.

With a sensor (webcam) the light-section is digitised. Only the red part of the obtained bitmap picture is used for further processing. With this data a single laser line can be reduced to many sub-functions of one pixel width. Together with information on the measurement setup the transformation then generates the original scene from the camera data.

By using curve-processing-algorithms the 3D-surfaces are smoothened and disturbances and noise are suppressed. These algorithms are applied first of all to the subfunctions (as approximations) and secondly (as interpolations) to determine the intermediate values of neighbouring light-sections. Thereby enabling variable resolutions of the grid-model.

The output is optionally in a data file for viewing and further processing in Matlab, or in vector graphic for mat (Object File Format) for a 3D viewer. In Matlab the depth picture (contour model) and the equidistant grid model are generated. The vector graphic only shows the equidistant gridmodel (wire frame).

Due to instructions of the Institute for Mechatronic System at ZHW a CANopen Multi-I/0 Node has been developed for the project SMARTparking. A product was required that allows operating with different sensors and actuators according to the CANopen standard. The work study was based on the previous project Har01/1 and the source code of a CANopen-Stack.

The existing software was adopted as well as expanded for two different Fujitsu microcontroller. The new CANopen stack is already successfully used in similar diploma thesis.

The hardware which was adopted from the previous project, with the measures of 46 x 70, is able to control 2 sonar ranging modules, 4 DC low-cost servos, 4 infrared sensors, 8 digital I/0 ports and 8 LEDs. The required drivers were adapted and some were newly developed before they were integrated successfully in CANopen Standard. The servo calibration, the baudrate, the Nod-ID and the digital I/0 port can be adjusted with an onboard menu. The Voltage supply of the node may vary from 10 to 48 Volt.

With this newly developed Multi-I/0 Node all required sensors and actuators may be configured and used from CANopen. The time to finish our project was limited, therefore we had to do without extensive tests but it is our goal to catch up on this before the CANopen Multi-I/0 Node goes into production.

The Institute for Mechatronic Systems uses DC servo drivers for automobile power braking and steering in the context of its SmartParking project. To date, the motors have been driven with IMS universal boards or third-party products. The project assignment is to have this task performed by a CANopen servo controller module.

The degree dissertation includes the development of a pilot production quality compact CANopen node with the following hardware features:

• wide input voltage range (10V...60V)
• external power supply for voltages above 60V
• limit switch connection
• CAN/CANopen steering
• retail price approx. CHF 300

The assignment also comprises writing driver and test software for the already existing hardware, and subsequent initiation. Finally, the device profile for drivers is to be implemented according to standard.

The hardware was tested using the specifically developed driver and test software and, following decisive changes to the operating software, initiation was successfully completed. The device profile was realised with all compulsory entries and linked up with a CANopen stack using the communication profile. The driver is steered using a position and speed regulator with trapezoid speed progression.

The entire system has been successfully tested for its main functions, confirming that a fully operational servo controller module for drivers in line with the CANopen standard has been developed.

Abstract

Miniaturized, passive magnetic sensors for wireless data-transfer are of a great interest where a sparingly use of the energy is important. Possible uses are: cardiac peacemaker, hearing aid, wrist watches, cellular phones and so on. The core of these sensors is a miniaturized reedrelay-structure. For the development of these structures, numeric models are used since the measurement of the forces at the contacts of the relay on a experimental way demand a disproportionately effort. The function of the relay is based on magnetic forces. If the contacts live, also electromagnetic forces influence the system (because of the small distance of the contacts).

Within the scope of this dissertation the electrostatic forces of the MicroReed-structures from the company ASULAB SA are modeled with the SESES (SEmiconductor SEnsor and actuator Simulation), a program of finite elements.

Several models with a different degree of complexity have been used. In the simplest case, parallel relay contacts without scatter-field has been looked at. Such a model with idealized force-characterization (analytic solution of the corresponding field-problem) is essential for the modelvalidation. In the most complex case, ASLUAB structures considering the scatter-field have been modled. Core of the dissertation is the validation of the electrostatic forces. The dimensions of the modeled environment (surrounding the contacts) and the choice of the finite elements Machine were the points of investigation.

The electrostatic forces have been measured with the following methods: (a) With derivation of the electromagnetic energy in the entire model that has been modeled for several contact-distances. (b) With evaluation of the Maxwell-tensor at elements that copy the contacts.

Special emphasis was given to investigation of convergence of energy and power.