A distributed architecture for unmanned aerial systems based on publish/subscribe messaging and simultaneous localisation and mapping (SLAM) testbed

• Main Author: Saib, Kamil Nuther
• Format: Others
• Language: en
• Published: 2018
• Online Access: https://hdl.handle.net/10539/24843

A dissertation submitted in fulfilment for the degree of Master of Science. School of Computational and Applied Mathematics, University of the Witwatersrand, Johannesburg, South Africa, November 2017 === The increased capabilities and lower cost of Micro Aerial Vehicles (MAVs) unveil big opportunities for a rapidly growing number of civilian and commercial applications. Some missions require direct control using a receiver in a point-to-point connection, involving one or very few MAVs. An alternative class of mission is remotely controlled, with the control of the drone automated to a certain extent using mission planning software and autopilot systems. For most emerging missions, there is a need for more autonomous, cooperative control of MAVs, as well as more complex data processing from sensors like cameras and laser scanners. In the last decade, this has given rise to an extensive research from both academia and industry. This research direction applies robotics and computer vision concepts to Unmanned Aerial Systems (UASs). However, UASs are often designed for specific hardware and software, thus providing limited integration, interoperability and re-usability across different missions. In addition, there are numerous open issues related to UAS command, control and communication(C3), and multi-MAVs. We argue and elaborate throughout this dissertation that some of the recent standardbased publish/subscribe communication protocols can solve many of these challenges and meet the non-functional requirements of MAV robotics applications. This dissertation assesses the MQTT, DDS and TCPROS protocols in a distributed architecture of a UAS control system and Ground Control Station software. While TCPROS has been the leading robotics communication transport for ROS applications, MQTT and DDS are lightweight enough to be used for data exchange between distributed systems of aerial robots. Furthermore, MQTT and DDS are based on industry standards to foster communication interoperability of “things”. Both protocols have been extensively presented to address many of today’s needs related to networks based on the internet of things (IoT). For example, MQTT has been used to exchange data with space probes, whereas DDS was employed for aerospace defence and applications of smart cities. We designed and implemented a distributed UAS architecture based on each publish/subscribe protocol TCPROS, MQTT and DDS. The proposed communication systems were tested with a vision-based Simultaneous Localisation and Mapping (SLAM) system involving three Parrot AR Drone2 MAVs. Within the context of this study, MQTT and DDS messaging frameworks serve the purpose of abstracting UAS complexity and heterogeneity. Additionally, these protocols are expected to provide low-latency communication and scale up to meet the requirements of real-time remote sensing applications. The most important contribution of this work is the implementation of a complete distributed communication architecture for multi-MAVs. Furthermore, we assess the viability of this architecture and benchmark the performance of the protocols in relation to an autonomous quadcopter navigation testbed composed of a SLAM algorithm, an extended Kalman filter and a PID controller. === XL2018