Unfortunately I've determined that the Aero RTF kit doesn't meet my needs and I've decided to return it. Although my 'ROI' on explaining why here is probably sub-zero, I want to describe the reasons so that Intel can use the information to learn/improve and developers are forewarned about what they're getting into.
I ordered the Aero RTF with the expectation that the 'RTF' kit would at least be a stable platform with basic sense & avoid capability. It is neither of these. My goal was to familiarize & experiment with the RTF kit until achieving a basic comfort level, then transplant the components onto one of the larger drone prototypes that I'm working on. However, this experience hasn't met my expectations in any way. To briefly enumerate why:
My apologies if this is getting TL;DR. To summarize, I believe Intel has a good thing going with RealSense. Unfortunately it appears that the typical large-company middle-management CYA decisions have hamstrung the product for users like me. I'm going to take a close look at Yuneec's products to see it it's reasonable to graft it's controller onto my airframes. Apparently these vehicles not only have RealSense hardware onboard, but have software to make it useful.
Thanks for reading this. I hope someone here finds it valuable. Good luck!
First I have to say that I've never got a chance to play with the RTF kit yet. But I've been paying very close attention to Joule and Aero recently. The main reason why I still haven't decided to actually buy and try either of them is my bad experience with Edison. And Joule/Aero/RTF are just too expensive to be tried and left in the dust.
I think Intel is producing very attractive hardware. Edison/Joule/Aero all have great potential for different types of IoT applications. But as Dave indicated, I also feel the software/documentation support is really poor. I'm not an expert on IoT or embedded systems but I think I have decent knowledge/experience to work with this kind of single board computers. I have to say I often get frustrated when playing with Edison and trying to make it work the way I would like it to. If you're just trying basic applications, maybe the support is enough. But if you want to go further, it becomes much more difficult. The support guys here are polite and patient but I always feel the company policy or whatever is making a very clear barrier between them and the developers here. One example of the bad support from Edison is the MCU SDK. It's understandable that there are limitations in the latest release. But it's totally unacceptable to not update the SDK for over a year AND not provide any information on the roadmap or future plan. This easily makes people think Edison is dead and no valuable time should be spent on this platform anymore. Joule and Aero are relatively new and I truly hope they won't follow the same path as Edison. I really look forward to seeing some changes.
I get frustrated by sparse of non-existent documentation as much as the next developer. I wonder though if it is a bad habit that IT folk have developed over the last couple of decades as the pace of the world has increased: instant knowledge gratification. At the dawn of the personal computing industry, programmers and engineers made new discoveries and innovations by inserting different values into variables written by other devs to see what would happen when they changed.
That was how the whole concept of 'POKE' cheats for infinite lives, level skips, etc in 8-bit cassette games in the early 80's came about, and how companies who wanted to make games for Nintendo and Sega without an official licence created dev kits through reverse-engineering consoles via calling registers to find out what function did what.
In my own case, most of my knowledge about RealSense came from spending a couple of years picking apart Intel's sample scripts, putting bits in and taking stuff out, and seeing how I could expand their capability or change its behaviors. That taught me eventually how to build my own custom code and mechanisms for RealSense.
"Plug and play" development (looking up a code sequence in a manual and tweaking it a little) is great when you're in a tight spot time-wise. But we still need a bit of good old fashioned blind dabbling and pulling machines apart too in order to keep technology fresh and moving forwards.
I get what you're trying to express. But I think neither of the two extremes (plug-and-play vs. digging out every bit on your own) is a good thing. And I feel it's reasonable to ask for necessary support for a product. I totally agree with you that one needs to spend time before he/she can really use a tool well and should not expect everything to be ready for use out-of-box. But time budget is also a very important factor to consider. I think one purpose of the Joule/Aero/Realsense platforms is to accelerate the development of various IoT/robotics applications. People here have different backgrounds and may have different concentrations. The computing board may be just one element in the whole project. You will never get anything done if you have to reinvent every single part of your system. At least for me, I'm just trying to find a good balance. I like to spend time learning and doing experiments with a new widget but I also need to get my work done. Of course, it would be great if your work happens to be aligned well with the digging process you described or you don't have a tight deadline to meet for it.
It was in no way meant as a criticism of those who needed documentation, RxDUty. As I said in my opening, I appreciate good documentation too. I was just having a philosophical muse directed at no person in particular, and showing my age too. When deadlines are a factor, good information is priceless. The "old guy" part of me just tries to hang onto his appreciation of experimentation too, lest humans become like the Eloi in the far future of HG Wells' The Time Machine: laying around having fun whilst the libraries of knowledge crumble to dust and the Morlocks eat them for lunch *grins*
Thanks for your very well-argued response!
No worries. I knew you had no bad intentions in the earlier reply. I believe most people here are just trying to learn things and get work done, not fighting with others with different opinions. Personally, I still think the hardware from Intel is very attractive for many IoT/robotics applications, even though Atom didn't play well in the mobile device market. To be really competitive with ARM alternatives, Intel still has a lot to improve.
I have the RTF package. The comments represent a rational range of user experience and expectations.
I got with Edison from the moment it came out. There was a bunch of handwringing on the cost of the board. For $50 you got wifi, ble and powerful MCU with 2gb of memory -If these features matter to you, this is the best buy in the industry.
It was also programmable out of the box using Arduino IDE, node.js or C++ using free IDE's and python w/o any IDE.
The platform is rock solid too - Edison # 1 runs indefinitely pulling weather reports from the Internet and local conditions from a TI BLE SensorTag then displaying them on 7-segment LED's. Edison # 2 runs a robot with video camera and other sensors. Edison # 3? - 12V through battery + does not have a good outcome. If you aren't blowing these things up from time to time you are not really having fun. Just ordered up another one today for prototyping.
Spotty docs - yes. An unusable phantom MCU (quark) - yes.
These are tradeoffs I will gladly make for the level of quality of the product itself. Intel made a substantial investment they are not recouping from the sales price in the name of making sure the Intel name is on a healthy % of the 20B IoT CPU's projected. I really like being on that side of the equation.
Like the RTF Package, the Joule and RealSense - these products are for those of us that want to work ahead of standard documentation and product lifecycles that can add 1 or more years to channel delivery.
When I compare the RTF package capabilities to what is out there - I think the case can be made for at least $1000 in cost savings in addition to the wait to get things to a basic level integration and documentation.
I actually got the RTF pack in part to experiment with RealSense as the aero is pretty kickass by itself although will likely succumb to a Joule based system to prototype with.
Not trying to be cavalier to the very rational criticisms outlined here - but honestly - I would rather have the RTF pack now than wait a year before they got it all straight. YMMV.
I just got my refund yesterday. On top of having a dead USB port (which basically means no updates, development, etc.) my 40-years-in-tech spidey-sense was telling me the RTF drone was a hastily put together program with very few resources to support a complex product. The tipping point was when they posted that document admitting that all most all of the hardware functionality comes without any working or even example code, including the range finding technology that they are touting in the press. Seemed like a very one sided value proposition to me and $1200 is a lot of money for drone hardware these days.
Experimenting and prototyping pre-market-ready technology is not for everyone.
"The Intel Aero Ready to Fly Drone is a platform for developers and is intended to be modified by developers according to their professional judgment. Intel has not established operating limitations for this drone development platform, or tested any configurations other than the base configuration as shipped. Developers are responsible for testing and ensuring the safety of their own configurations, and establishing the operating limits of those configurations. "
Intel was pretty upfront about this so I am puzzled how your "40-years-in-tech" got so wrong an impression of what you bought not to mention the many hours of time you spent posting about it.
Fully aware it was a developer package thanks. In fact I am going back to developing on the open source stuff which is much cheaper and has a much larger community.
What were my hints? Well first there were the multiple delays where folks at "the store" would send me email promising it absolutely will ship by x-y-z and but it didn't. After the 2nd time this happened they gave me the option to cancel and I almost did.
Then it showed-up dead and when I tried to get support help I got nothing. In fact it appeared to me that there wasn't a lot of depth on the technical support team to be brutally honest. I invested a $100 in additional hardware including multiple USB hubs and cables just trying to get the USB to react to .
So to be honest if my machine had worked out of the box I might have dived-in and poked around the code but that wasn't an option for me.
For what it's worth - probably not much since this is likely my last post here - a few parting random thoughts thrown together on my lunch-break:
Lunch break is over. I hope Intel managers read this and understand and correct the bifurcation that appears to being created here. And that developers also understand they might be investing in a technology cul-de-sac unless they're aware of these issues and have plans to navigate the less than clearly outlined issues at hand.