Unmanned Systems Technology 004 | Delair-Tech DT18 | Autopilots | Rotron RT600 | Unmanned surface vehicles | AMRC | Motion control | Batteries

34 Focus | Autopilots that can introduce uncertainty into a control loop. It has been used as the heart of an autopilot running a dedicated scheduler as, with the right software, it can provide deterministic and highly effective control. However, this technology is 20 years old now, and the system design has evolved to an architecture where the time-critical autopilot code is run on the existing Coldfire processor while a separate board with a more powerful processor runs a Linux operating system. This allows customers to add their own software for their specific requirements and brings less risk of introducing bugs and failure modes that would have come with moving the code to a new, faster processor, and so maintains the control system’s reliability. Performance factors Regarding the performance of the autopilot there is a key balance between supporting many different types of platform and achieving the required timing for the control system. Some suppliers provide a range of autopilots optimised for different platforms, with versions for fixed-wing, rotary and hybrid designs for example. However, the autopilot has to be capable of handling different modes, even with a standard platform. Controlling the aircraft during take-off is very different from controlling it in flight, because the way the sensor inputs are assessed, and the resulting outputs that are used to control the aircraft, can themselves be very different. Developing an autopilot for a hybrid craft such as the Arcturus Jump is even harder, as it operates as a fixed-wing aircraft in flight and changes the motor position on take-off and landing to use the propellers in a vertical, rotary mode. There are fundamental differences between the control loops for these platforms. While the hardware and the operating system can be the same, the way the sensor input is handled is very different, and the autopilot has to take this into account. It does so by filtering Autumn 2015 | Unmanned Systems Technology few knots has a very different autopilot requirement from that for a jet-powered system or a helicopter-type platform. The autopilot controller has to handle a range of sensors, all feeding data into the comms ports on the autopilot board. One challenge is to support more interfaces while minimising the size and weight of the autopilot hardware. Some of the most popular commercial autopilots use fairly basic hardware technology, coupled with customised programming techniques and control algorithms. One such system is based on a relatively simple 8-bit processor with its own programming environment, called Dynamic C, which is optimised for control algorithms by adding new instructions to the standard C language. This means a dedicated compiler has to be used to turn the Dynamic C into the instructions for the processor. This has created highly efficient software that allows sensors to cross- check each other for accuracy and fault detection, and the autopilot uses failsafe algorithms to return to base and land automatically in emergency situations such as low battery or loss of comms with the ground control station. Another leading autopilot system uses the 32-bit Coldfire processor architecture that was launched in 1994 by Motorola Semiconductor (now Freescale Semiconductor, which is merging with rival NXP Semiconductor) to provide more performance from the highly successful 68000 family of chips. The Coldfire V2 architecture follows the same design strategy as the 68K, with a simple processor pipeline but without a memory management unit or other peripherals Piccolo III is the latest evolution of autopilot technology (Courtesy of Cloud Cap Technology) There is a key balance between supporting many types of platform and achieving the required timing for the control system