Chinese fire up world's 'most powerful' drone brain Chinese UAV outfit DJI is trumpeting the release of what it describes as the "most powerful computer designed for drones" – the "Manifold" embedded computer packing a quad-core ARM Cortex A-15 processor. The Manifold With NVIDIA's Tegra K1 and you've got the power to "transform aerial platforms into truly intelligent flying …
What I am getting at is this.
This is a UK based tech site and this is a story for for a Chinese drone, we have only been given one price, in US dollars.
My question, why?
Why do I read a UK site if I have to go off and convert the price to a UK amount?
I might as well go back to the uk-not-so-uk tomshardware.co.uk or whatever.
Seriously vast performance advantage. Except for the flying component, they've hit every box of my dream machine. (I'm into ground based and sub-surface, waterborne and submarine.) The K1 makes for a nifty signal processor with nVidia tooling.
BTW first use of airborne is reporter/paparazzi in a box. And they thought the recent fires here in California we had were hard to deconflict? Child's play.
This a a Raspberry Pi on steroids. The original Pi is built down to a price and is barely more than an obsolete phone chip on a breakout board. The Pi series is fine for the money. But if you have more money and want a more powerful ARM with much more I/O then this is it. I wish it cost 1/2 as much though.
At a bit lower price this would be very nice for all kinds of projects where the Pi lacks I/O or CPU power.
That has enough grunt to run most algos for abominably precise targeting and "under radar" flight which used to be a privilege of the likes of Raytheon or Almaz systems. Anything you can think of - terrain guidance by profile matching (cruise missiles), last stage guidance by visual, IR, UV or radar sig. You name it.
All you need is to mount it on one of these https://www.youtube.com/watch?v=potAETW-VG8 or https://www.youtube.com/watch?v=sa-TSNeTK-A instead of a puny electric quadcopter along with a laser altimeter and a few cameras for the terminal guidance. A few kg of high explosive optional.
Thanks for the correction. I meant the particular Oerlikon goods which supposedly neutral Switherland shipped to all warring parties in WW2. Namely this: https://en.wikipedia.org/wiki/Oerlikon_20_mm_cannon. While it is not the best in class today (that title goes to the Soviet/Russian ZPU), you can pretend that it is an antique garden ornament commemorating Swiss neutrality in WW2.
Probably not, most military and aerospace control systems use established HW with known predictable errors and appropriate responses to those issues and countermeasures implemented.
They will also run an aerospace specific RTOS, no need to run Linux.
Most of the processing power in this is to run the OS, 99% of control systems don't actually need an OS of this magnitude.
E.g. The flight control system on the first generation F16's was built in the 70s using that era's technology.
This post has been deleted by its author
Way too predictable:
Jack Ryan:
Has he made any Crazy Ivans?
Capt. Bart Mancuso:
What difference does that make?
Jack Ryan:
Because his next one is going to be to starboard.
Capt. Bart Mancuso:
Why? Because his last was to port?
Jack Ryan:
No. Because he always goes to starboard in the bottom half of the hour.
Errr, what? AI doesn't work even on huge, big (and heavy) supercomputers or mainframes. I'll take a wild guess and assume that line is straight from the press release with little or no interpretation by a journalist writing for a highly tech aware audience.
EDIT: Oh, hang on. I just noticed. It's Lester. Carry on then, as you were.
I'm a great ARM fan but in a drone power is not really a problem - if it is Adapteva Parallela gives you best bang for watt.
Is there a reason why everyone uses batteries in drones when its really easy to get watts from glow plugs which are much better power to weight and pretty reliable. And no they are no noisy if you feed the exhaust into a decent silencer.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
