Unmanned Systems Technology 010 | nuTonomy driverless taxi | Embedded computing | HFE International marine powertrain | Space vehicles | Performance monitoring | Commercial UAV Show Asia report

32 E mbedded processing is at the heart of all autonomous systems. A microprocessor core supported by a range of peripherals provides the essential computing resource for developers, but there are several elements required to make use of that computing power in a driverless car or a UAV, and these are all linked together. Several different ways of providing computing power for autonomous control algorithms and sensor processing are emerging, combining design blocks, physical implementation in silicon and the board design and form factors. The design of the processing blocks provides the core computing resources, from standard CPU processor cores to digital signal processor (DSP) and graphics cores. The blocks also include the interface protocols to provide high- bandwidth links, particularly for cameras and Lidar sensors. Deep learning and neural network algorithms for artificial intelligence are now increasingly making use of arrays of graphics processing units (GPUs) and the latest DSP cores. The implementation of the cores in silicon determines the power consumption, the speed of the interfaces and even the choice of the processing units themselves. The board form factor is often determined by that power envelope and the bandwidth and connector requirements, as well as the space available and the weight restrictions in the vehicle. Aerial systems have tighter weight and power restrictions than ground vehicles, although Space, Weight and Power (SWaP) is an increasingly important factor for electric autonomous vehicles to get as much mileage as Nick Flaherty gives a guide to the wide range of embedded computing solutions and standards for autonomous systems Board-level decisions Nvidia is using two processors that are aimed at building supercomputers on a board for controlling driverless cars (Courtesy of Nvidia) October/November 2016 | Unmanned Systems Technology