Implementing vision,video and radar in unmanned aerial vehicles is now a task for multicore DSPs,writes Sandeep Kumar.
Unmanned aerial vehicles(UAVs)used by the military and in commercial applications such as fire-fighting are incorporating higher levels of digital signal processing to support automated navigation/landing and air traffic management systems.
The challenge for designers is to achieve this within the aircraft's power budget.
As a result,multicore digital signal processors(DSPs)are being considered as a design option for UAVs.The DSPs are used for the acquisition and processing of vision,video and radar data,and the establishment of secure,flexible ground communications in UAVs.
For reconnaissance and combat activities,military UAVs must integrate a wide range of subsystems including visual sensors,mission compute systems,radio communications and navigation systems within a very confined space.Small form factors and smaller modular enclosures complicate power supply,with some smaller systems having sub-200W power budgets.
Currently,this energy issue is resolved using large onboard power generation systems coupled with heat collection/rejection technologies that dissipate heat.
The need for these large onboard power generation systems and heat sinks can be minimised by using more power efficient multicore chips.For example,in a half-length PCI Express(PCIe)board,measuring less than 5in x 7in,has four multicore DSPs with performance of 500GFLOP at 54W power consumption.
The ability to quickly process large volumes of raw infrared and visible imaging,and full motion video data is a requirement for most UAVs.In each UAV,there can be installed between 10 and 100 CMOS cameras,with each one capturing up to several hundred megabytes of imaging data.
This combined with the drive towards high definition(HD)imaging and higher resolution,means that a substantial real-time processing capability is necessary.Furthermore,on-board image stabilization,noise reduction and compression must be carried out,as well as in-flight image enhancement/restoration and analysis.
DSPs can be used to implement advanced codecs e.g.JPEG2000,which provide optimal image quality,no inter-frame dependencies,and low latency when images are compressed and transmitted from the UAV to the ground in a'lossless fashion'(i.e.when all image data is restored when the file is decompressed during viewing).
They also have processing power to support existing video standards,for example H.264.They are programmable making it easier to support non-standard video formats.
Wide IO bandwidth allows packetised compressed video to be supported.A Gigabit Ethernet(GigE)or higher port is needed to handle multiple compressed channels transporting HD videos that typically require up to 10Mbit/s.
For high bandwidth video,PCIe and RapidIO should be considered.A PCIe Gen2,for example,can carry 5 gigabytes of data,which is sufficient for 4 channels of 1080i60,24 channels of D1,or 300 or more channels of QCIF at 30fps.In embedded systems,greater payload and continuous scalable video processing can be achieved with RapidIO.
Another source of data originates from on-board synthetic aperture radar(SAR).When used in conjunction with a ground moving target indicator(GMTI),SAR is able to locate targets or locations that cannot be detected by optical or infrared cameras owing to weather conditions(e.g.fog)or the terrain(e.g.heavily wooded areas).All the main requirements of SAR processing are handled effectively with advanced multicore DSPs,including:
Floating point–to provide the greater output precision,and larger dynamic range required for SAR
Variable size Fast Fourier Transforms(FFT)
'Corner-turns'or'matrix transpose'for an efficient SAR algorithm
A secure,continuous two-way communication datalink is critical for UAVs.This requires a high data rate channel for transmitting signals intelligence,video and radar,alongside a low data rate channel for'commands'and'status'.
The use of DSPs to provide secure,low latency,real time communications is well established,and they can be used to enable communications in UAVs that requires support for both low bandwidth proprietary channels,as well as broadband wireless air-interfaces e.g.LTE,WCDMA,and WiMAX.
Another feature of DSPs is the support for software compatibility/portability,incorporating open multicore programming models(e.g.OpenMP),open operating systems(e.g.Linux),and open tools(e.g.GNU C-compiler),which reduces the need for software investment.
Sandeep Kumar is product line manager at Texas Instruments
