The research of the measurement group at Linköping university has for some time been focused on the role of measurements in robotics. The main objective of the group's research work has been to provide the robot with sensor information in order to increase its autonomy.

In this work I have studied problematic aspects of robot use that are related to robot safety as seen from the measurement technology point of view. The work can be divided into three subparts:

• a sensor for person presence detection

• a tool for robot control with built in emergency stop

• object recognition, a requirement for increased autonomy that can decrease human presence close to the robot.

In robot safety the aim was to develop a man-detecting sensor that could be used for safety purposes. The efforts were concentrated upon a capacitive sensor for the following reasons:

• man's electrical characteristics differ strongly from those of the air forced out of its place by the human body

• man can not change its electrical characteristics.

Both model calculations and practical tests have been made. Persons are detected as intended.

Industrial robots are equipped with a teach pendant, a hand held control device. It is connected to the control system of the robot and is used for work in close interaction with the robot. Work close to the robot might for instance be necessary during program verification or adjustments. It is important for a person, physically close to the robot, to be able to change command or initiate an emergency stop rapidly and easily. To that aim a user-friendly control tool with a built-in emergency stop is proposed. The tool is based on a glove that measures the finger flexions. The finger flexions form patterns that correspond to different commands. The commands are recognised by methods based on fuzzy principles. The glove is tested on gestures inspired by the manual alphabet used by deaf people. All gestures/commands were correctly recognised.

In the sensor system used by the measurement group a 2-D vision system is used for the localisation and recognition of different work pieces. This system does not in all cases provide the 3-D information needed, e. g. whether an object is a cylinder or a cube. A two-camera system or special illumination techniques can provide the information. A very simple laser scanner provides however enough information for the detection, i. e. gives unique signatures for differently shaped objects. We have investigated this simple and relatively cheap solution. Fourier transformation is used to recognise the differences in the detected laser scanner signal. The result shows that it is possible to recognise objects by this method.