Design and implementation of a computer system for unmanned vehicles

Abstract

In this work, the design and implementation of a generic, computer system for unmanned vehicles (UVs) was tackled; more specifically, the UV targeted in this work was the micro unmanned vehicle (MUV) for non-safety-critical applications. As the title implies, this system is to be generic; therefore, it was not designed for any vehicle in particular. Also, unlike most of the literature, the term 'computer system' was preferred over autopilot system because the design here presented encompasses other features which are typically associated more with a computer system rather than an autopilot system. Nevertheless, since the primary role of such a computer system is indeed the autopilot function, in the following text, the terms 'autopilot system' or merely 'system' were used as well. Various computer systems for unmanned vehicles, both commercial and proposed in the literature were investigated and FPGA-DSP based systems were found to be the most promising. FPGAs, however, incur a price and power consumption penalty amongst other things. The limitations of the FPGA-DSP approach, combined with the relatively small number of autopilot systems intended for micro unmanned vehicles currently available on the market motivate this work. With this in mind, a new computer system architecture for MUV s has been proposed. A prototype of the proposed system was then designed and built, and its electrical integrity and functionality were verified. Field testing was then conducted in order to assess the performance of the system. A GPS-aided inertial navigation algorithm based on the Kalman filter was designed for this purpose and the system driven along a predefined course. The execution speed of the system could therefore be assessed and the collected data analyzed. From the obtained results, it is shown that the performance of FPGA-DSP systems can be achieved without the use of FPGAs. This results in a significantly cheaper design and a reduction in the power consumption of 26.67%. Moreover, since the proposed system does not require VHDL programming, it can also be claimed that the proposed system results in a simpler design process. These advantages were obtained despite the fact that the proposed system has a broader peripheral set.