University of Twente Student Theses


Demonstrator combining ROS/TERRA-LUNA

Vos, Peter-Jan (2015) Demonstrator combining ROS/TERRA-LUNA.

[img] PDF
Abstract:The Production Cell is an existing model of a plastic moulding plant and is used as a platform to test multiple hard real-time embedded control systems. The usage of hard real-time systems ensures a reliable control system, but limits the possibilities for additional tasks that can be performed since they can exceed the time window of the hard real-time system. At the University of Twente a ‘bridge node’ is made which links a hard real-time system and a soft real-time system. The LUNA bridge connects the in-house developed hard real-time execution framework LUNA to ROS. ROS is an increasingly popular soft real-time software package to control robots. With the bridge node, called luna-bridge, it is possible to control actuators in hard real-time and update their setpoints based on complex calculations done on another computational platform. To demonstrate the luna-bridge, a newmodule is added to the Production Cell which identifies and extracts certain blocks of the plant. This sorting module is equipped with a webcam, which images are analysed in a ROS-node. The blocks that move around the Production Cell are marked with certain shapes, which are identified in the ROS-node. If a positive match is found with the hard-coded shapes, a new message is sent over the luna-bridge to the hard real-time system. The hard real-time software runs on a QNX-based PC104. An AnyIO-card is used to communicate between the motor/encoder and the LUNA application. The LUNA application receives messages from the ROS node, and converts these to a setpoint. The sorting rod is then controlled to this position and the block is extracted from the belt. The image processing node takes 18 ms to analyse the shapes of one frame. On average, a total of 9.2 frames are captured by the webcam when a block passes. Out off these 9.2 frames the image processing algorithm is able to extract the three shapes in 8.5 frames. Identification of the shapes on these frames is done with 100 % accuracy based on a test with 36 blocks, so that the sorting module is always able to correctly identify the block marked for removal. The sorting module is able to move to its extraction position within 0.4 seconds. The module is therefore always in time to extract the block marked for removal, which will arrive 0.4 seconds later and is successfully removed after another second. However, when the Production Cell is operating at maximum speed, the sorting module can unintentionally push off the block in front of the desired block. Additions to TERRA, or the 20-sim to TERRA code generation, are required to add a timing mechanism to the platformso the sorting module can operate safely while the Production Cell is operating at full speed. The mechanical design of the sorting module can be improved by making it smaller, decouple the rod from the shaft and adding end-switches. The software uses hard-coded shapes which are tagged for removal, a modification so that the removed shapes can be set at run-time would be nicer. In future projects, the luna-bridge can also be integrated in more complex set-ups which require bidirectional communications. The new sorting module is a stand-alone systemwhich is added to the Production Cell. It works without interfering with the normal workings of the Production Cell, and can be removed if necessary. The new module uses a hard real-time LUNA application to control the motion, while the image processing is offloaded to a soft real-time system. The platforms communicate through the luna-bridge, which sends commands from the soft real-time ROS application to LUNA. By using the luna-bridge, it is possible to combine the benefits of a hard real-time system with a soft real-time system.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Systems and Control MSc (60359)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page