A dynamic sensor interpreter for roboti...

Changing configurations of robots e.g. in modular robotics, due to retrofitting or maintenance often requires the addition of new sensors that are previously unknown to the system. Usually, it takes quite some effort for the logical integration until the sensor data can actually be used in the system, like updates of the host software, installing drivers or software modules. This can be a time consuming and potentially also an error-prone task. We propose an approach to speed up this process by including an electronic data-sheet standard in the sensor node that contains meta information such as the physical quantity that is measured, the encoding of the data, data rate, calibration data or measurement uncertainty. Based on this information, a Network Capable Application Processor (NCAP) on the host side can automatically integrate the device in the system by adding the information to a so-called parameter server and providing the sensor data (including calibration) as Robot Operating System (ROS) topics. It is important to note that this procedure does not require any change of the software on the host side. Furthermore, the datasheet also allows to validate if a certain sensor suffices the requirements for a certain task, which is of particular interest in modular approaches, where configuration are frequently changed over the lifetime of the system. Consequently, the proposed approach not only helps to speed up the integration of new devices but also contributes to the stability of the system as automation reduces the risk of errors and as avoidance of additional software modules and drivers reduces the risk of incompatibilities. We demonstrate the benefits of the approach in the retrofitting of a forestry crane with sensors. With the addition of such sensors, such machines turn into autonomous systems enabling the automation e.g. of ordering wood logs.