Proceedings of the 13th WSEAS International Conference on SYSTEMS
qB - Distributed Real Time Control System in UAV design KAMIL KOZAK ROMAN KOTERAS KRZYSZTOF DANIEC ALEKSANDER NAWRAT Department of Automatic Control and Robotics,
Silesian University of Technology Akademicka 16, 44-100 Gliwice POLAND
[email protected],
[email protected],
[email protected],
[email protected] Abstract: - Paper present brief description of qB - Distributed Control System designed for Unmanned Aerial Vehicle UAV. Concept is a result of few years development and was created to fulfill requirements pointed by the market. Described Embedded System operates in real time mode, is fully scalable and portable. Connection of hardware modules methodology and scalable software layer allows to keep product up to date regardless of market progress. Key-Words: - Real Time system, distributed system, UAV, autonomous vehicle.
According to definition, embedded system is a specialpurpose computer system designed to perform one or a few dedicated functions. However embedded system doesn't have to be a single hardware board with single precompiled piece of software. The idea to increase scalability and portability of such a system among different types of UAV platforms is to create distributed control system, that will allow to split single board solutions into smaller hardware modules connected with data buss that will realize "atomic" operations. In such a system it will be easy to exchange hardware modules same as redesign and prepare the system to operate on new mechanical platform . It must be mentioned that the crucial part of control system is its stability of long time operation and deterministic time of task context switching realized by underlying operating system. Real time operating system that works on a single board is easily configurable and allow to clearly define sampling rates of all sensors and periods between control algorithms iterations. Thinks becomes more complicated when distributed system have to be synchronized to guarantee real time mode operation. Software layer designed on top of real time operating system have to be highly portable and scalable to guarantee that it will be universal for different hardware platforms. Additionally as it was described in [8] distributed qB Control System have to cooperate with
1. Introduction Embedded system design is nowadays one of the most popular branches of computer science. Hardware and software platforms, merged together to fulfill single specific task are commonly used in the market. Minimized in size microcomputers with powerful computing units are present in almost every aspect of human life. We can find them in digital clocks, ovens or coffee machines. Most users of those common devices have no idea what sophisticated techniques were used to create those products and how complicated was the design process of those embedded systems. It should be mentioned that embedded system design is treated by European Union as innovative branch of science that should be explored and developed in nearest future. However design of single board, task dedicated systems have few disadvantages, one of them is that any change in hardware platform design requires change in software layer. It's crucial that at the beginning of embedded system development all tasks that should be realized must be known in details. Any mistake made at planning stage of the system may have difficult to solve impact on result of the development. Very often such mistakes are impossible to solve without redesigning of hardware layer and as a result of that changes in software layer.
ISSN: 1790-2769
185
ISBN: 978-960-474-097-0
Proceedings of the 13th WSEAS International Conference on SYSTEMS
and software layers. Additionally it has to fulfill all assumptions that were stated for single board solution. It must operate in real-time manner, without lags and delays, software developer must have the possibility to assign priority to all tasks and signals in the system. Few years of experience in UAV helicopter design allows to predict that level of disrupts on data bus generated by electric motors and fact that mechanical platform is vibrating with high frequency while flying, would be a serious problem that should be solved. Data buss have to be resistant to distortions such as surge, radiated and conducted emission. Described solution would be a good start for any control system not only dedicated for UAV platforms.
monitoring tools to guarantee different modes of operation of UAV platform such us SIL (Software In the Loop) and HIL (Hardware In the Loop). Bottle neck, that will have highest impact on implementation, real time operation mode, synchronization and scalability is a data bus that have to merge all modules of created distributed system. Data bus have to allow developer to assign priorities to specific messages in the system, minimize collisions of data frames to maximize throughput and guarantee consistency of data in whole distributed system. Finally, the goal of this project is to develop innovative, flexible and easily extensible hardware and software platform for autonomous UAV control, that involve advantages of distributed systems to minimize the work that need to be done to adapt it to new UAV mechanical platform.
3. Single board system When all assumptions about system operation are made, developers know in details what have to be done, the process of single board embedded system design can be started. After few weeks of development hardware platform is ready and the first prototype can be merged with software layer. If everything was designed properly after next few weeks embedded system is ready. It fulfills all assumptions made at the planning stage. In the mean time new sensors were developed. To stay in touch with the market new embedded system have to be designed and the work starts from the beginning. Such a scenario is very common in companies that develop electronic systems, it guarantee that there always will be a work that have to be done to improve the product. And noone wants to design universal platforms because profit from such development would me much lower. Single board solutions have a lot of advantages, main of them is that such a system can be minimized in size to the limits, different aspects can be optimized to increase performance of the whole system such as power consumption. In case of the system that have to be minimized is size that's the best way to go. However in case of UAV control system design, size is important but it's not at the first place. Scalability and portability of the system are the main goals that have to be achieved. Such a system consist of two independent projects that have to tightly cooperate with each other. First one is software layer, second one is hardware layer. Both layers have to be portable and scalabable to guarantee that the final product will be universal among different types of UAV mechanical platforms. Design, modes of operation and development techniques of software layer were described in details in
2. Problem Definition Control systems for UAV were designed and described in different research projects [1][3][4][5][6][7]. Innovative idea of distributed control system is a result of few years development of control system for UAV helicopter [8]. Already created and tested system was designed as single board dedicated platform Fig. 1. It allows to monitor all input signals, control all output signals, test different control algorithms, however it can be used only with one mechanical platform or with platforms with similar sets of input and output signals. Any exchange of sensor or upgrade in hardware design is almost impossible without redesigning of the system.
Fig. 1 Single board embedded system
The problem is to develop embedded system with identical functionalities, similar in size and weight to the one already created, that will allow to easily port this solution to other UAV mechanical platforms. System that will allow to exchange any sensor, actuator or computing unit without redesigning of whole hardware
ISSN: 1790-2769
186
ISBN: 978-960-474-097-0
Proceedings of the 13th WSEAS International Conference on SYSTEMS
Scalability of software layer designed for a single board solution Fig. 1 allow to minimize its functionalities to fulfill "atomic" operation of each node in the distributed system Fig. 2. It means that designed scalable and portable software layer is fully reused in every node of the distributed system with minimized functionalities, accurate for every node. Single board solution have some advantages. Mainly it's much easier to configure than distributed system. There is no problem with consistency of data, all sensors sampling periods and control algorithms are scheduled by one operating system on a single processor. In case of distributed systems enumerated things are much more complicated. Operation of such a system can be compared to real time system that works on single board hardware. Nodes in the network are similar to software tasks in the operating system, data bus can be compared to mailboxes, queues or other mediums that allow communication in multitasking systems. At this point similarities finishes and some assumptions have to be made to guarantee stable and real time operation of distributed system. In operating system task scheduling is based on priorities assigned to specific tasks. and similar mechanism have to be introduced in described solution. The biggest problem to solve is a choice of proper data bus. It must be possible to assign priorities to messages, there must be possibility to avoid collisions on the network to maximize throughput and minimize delays. List of available on the market and standardized mediums that allow to exchange data is very long, starting from most common like ethernet, USB, RS232 through more sophisticated ones like profibus, CAN or FireWire. It's possible to enumerate for them advantages and disadvantages however after long period of research CAN bus with CANopen protocol stack was chosen.
Advanced Robotics [8]. Designed qB software layer is fully scalable, portable, real time operating platform that can be easily used in different research projects, as a base of development to minimize the time needed to finish the product. Problem begins when hardware platform have to be designed as universal one. At the beginning it must be stated what does it mean scalable and portable hardware platform. Hardware will be scalable if there will be possibility to extend IO sets, computing power or attach some new hardware modules without redesigning of the already created parts. Portability means that the hardware platform could be moved among different mechanical platforms that have to be controlled without redesigning of the system. The only way that allows such a construction is called a distributed system.
4. Distributed system Distributed system is a set of independent devices connected and working together. The main feature of such a system is its transparency it means that from external world it is visible as a single logical device.
5. CAN Network CAN is a leading data link layer for embedded communication in passenger cars. It provides multimaster capability, which allows the design of distributed and redundant systems. Broadcast communication and unique fault confinement guarantees network-wide data consistency and reduces bandwidth requirements. Additionally sophisticated error detection functions which increases communication reliability. CAN is a broadcast communication with message-oriented transmission. Priority in the network is assigned to specific data rather than to a device. This is important when several stations compete for bus access
Fig. 2 "Atomic" nodes of distributed system
After such redefinition of onboard control system there will be no need to change any of additional modules of the UAV.
ISSN: 1790-2769
187
ISBN: 978-960-474-097-0
Proceedings of the 13th WSEAS International Conference on SYSTEMS
protocol for embedded control system called CANopen. The set of CANopen specification comprises the application layer and communication profile as well as application, device, and interface profiles. CANopen provides very flexible configuration capabilities. These specifications are developed and maintained by CiA members.
(bus arbitration). Always message with more important data will be sent first. The priority, at which a message is transmitted, is specified by the identifier of each message. The priorities are defined during system design and cannot be changed dynamically. The identifier with the lowest binary number has the highest priority. Content oriented addressing scheme guarantee high flexibility of the system. It's easy to extend existing CAN network without making any software and hardware modifications to existing nodes. Data transmission is also not based on the availability of specific types of stations, which allows simple servicing and upgrading of the network . Thanks to broadcast schema data are consistent in all nodes connected to the network Fig. 3. When one node , called producer, sends data with predefined identifier, all other nodes called listeners or consumers can 'consume' the data if it is needed for node proper operation.
6. Conclusion After tests performed on beta version of qB Distributed System, very positive results were obtained. Single board platform that was working as UAV onboard computer was redesigned. Now it operates as distributed system and there is no difference in system stability of operation. So the main goal was successfully achieved. System was designed with use of CAN data bus. All main features of this network, such as multimaster capability that allows the design of distributed and redundant systems, broadcast communication and unique fault confinement that guarantees network-wide data consistency and sophisticated error detection functions were very helpful in system development. Those functionalities allowed to redesign the system and keep all constrains connected with real time operation modes of control system what was obligatory from the beginning. New flexible software and hardware platform is an entry point for any embedded control system design. Now, taking into account all advantages of real time operation together with flexibility of distributed systems, new system can be ported to almost any mechanical platform with minimal work that have to be done to accomplish such a task. Any changes to hardware layer are allowed and easily doable. Any sensor or actuator can be replaced without bigger and difficult changes in hardware layer. New extensions for example wireless communication module based on WiMAX technology or biometric sensors to secure UAV model against unauthorized use can be added as 'atomic' nodes without any problems. Flexibility of hardware layer was achieved, together with portable and scalable software written in ANSII C language that works on top of real time operating system µC/OS-II it becomes very powerful and universal embedded platform designed to control Unmanned Aerial Vehicles.
Fig. 3 CAN broadcast transmission
Bus access conflicts are resolved by bit-wise arbitration of the identifiers involved by each station that want to get access to the buss. Hardware arbitration minimize possibility of any collisions in the network in same way maximizing the throughput of the data bus. Design of the real time distributed system is a little simplified with use of CAN network because it introduces mechanisms that guarantee data consistency and allow to assign priorities to messages what has direct impact on the bus load. Additionally thanks to CAN in Automation (CiA) nonprofit organizations profiles of all nodes are standardized same as CAN-based higher-layer
ISSN: 1790-2769
188
ISBN: 978-960-474-097-0
Proceedings of the 13th WSEAS International Conference on SYSTEMS
7. Acknowledgment The research presented here were done as a part of research and development project no. O R00 0022 06 and have been supported by Ministry of Science and Higher Education funds in the years 2008 - 2009.
References: [1] The Autonomous Robotic Vehicle Project (ARVP) http://www.arvp.org/ [2] MICRIUM – µC/OS-II – Real Time Operating System http://www.micrium.com/ [3] Haarbrink R.B., Koers E., Helikopter UAV for photogrammetry and rapid response, InterCommission WG I/V, Autonomous Navigation, 2005. [4] Eisenbeiss, H., A mini unmanned aerial vehicle (UAV): system overview and image acquisition. International Archivesof Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XXXVI-5/W1, 2004. [5] Eisenbeiss, H., K. Lambers and M. Sauerbier, Photogrammetric recording of the archeoligical site of Pinchango Alto (Palpa, Peru) using a mini helicopter (UAV). International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XXXIV5/C34, pp. 238-243. 2005 [6] Herwitz, S.R., L.F. Johnson, R.G. Higgins, J.G. Leung and S.E. Dunagan, Precision agriculture as a commercial application for solar-powered unmanned aerial vehicles. AIAA 2002-3404, Portsmouth, VA., 2002. [7] Unmanned Aerial Vehicles Roadmap 2002 – 2027,Office of the Secretary of Defense, Washington D.C. December 2002. [8] Kozak Kamil, Nawrat Aleksander, Development of the system for simulation and control of the UAV, Advanced Robotics, January 2009
ISSN: 1790-2769
189
ISBN: 978-960-474-097-0
