"Researchers at Universidad Carlos III de Madrid (UC3M) have developed a new system which improves the ability of a GPS to determine a vehicle's position as compared to that of conventional GPS devices by up to 90 percent, and which can be installed in any vehicle at a very low cost.
The system, which is based on sensorial fusion, was jointly designed and developed by the Applied Artificial Intelligence Group (GIAA - Grupo de Inteligencia Aplicada Artificial) and the Systems Intelligence Laboratory (LSI - Laboratorio de Sistemas Inteligentes) at UC3M. The prototype incorporates a conventional GPS signal with those of other sensors (accelerometers and gyroscopes) in order to reduce the margin of error in establishing a location."
Seems like magic, when it's not really and already out on Diydrones! Welcome to arducopter 2.9.2 :)
Full article at http://www.gpsdaily.com/reports/A_system_that_improves_the_precision_of_GPS_in_cities_by_90_percent_999.html
Comments
@OlivierD
I like that they are conscious of fast log writes--there could be some interesting things to do with high resolution log playbacks. I like their candid attitude--no need to constantly second guess what it is they are doing/saying. I think a good portion of this software was initiated by this Danker guy in Australia who seems to know what he is doing & put in a huge effort long before 3Drobotics knew which way was up. The messaging protocol looks interesting & has been taken up by at least one other group.
http://wiki.openpilot.org/display/Doc/OpenPilot+Architecture
I've yet to find a similar document on the 3Drobotics storefront, er, I mean APM wiki site. I'm sure it exists somewhere, but all I've found is what looks like a monolithic blob in refactor mode.
Would be good if the different groups started standardizing and interoperating around the best written modules as was done with UNIX. For that matter, would be nice to see more standardized AP2Linux networking protocols as things are quickly moving to vision & other higher level software realms. Too bad 3drobotics has to ditch all that Swiss goodness to stay aligned with the Arduino fad. Such difficult decisions will impact their long term viability in the market for sure. That said, binding to scripting code is a good idea--autopilots seem very hardware centric in their culture.
Oops, am I coming across as a diydrone hater? Go go, 3drobotics go! Give that guy Tridge more raise, stocks, & say so--rsync is one of the most used tools in my box.
I guess next up for me will be a high flyover sniff of Autoquad and Openpilot.
J, I think TauLabs looks very promising, they are working on multiple STM32 platforms and moving pretty fast. Most of the members of this community were key developers at OpenPilot before things turned a bit strange over there, they formed their own group. Here is a post about one of the boards borne from that split:
http://www.diydrones.com/profiles/blogs/quanton-flight-control?id=7...
I bought my APM2.5 for the same reasons as you but got discouraged by the reports of fly aways, erratic behavior and how complicated it is to get it tuned in a fashion that you can feel 100% safe about for your first flight. A friend of mine has a CC3D OP board, first time setup took 12 minutes and it flies rock solid since. I know the APM can do this too but the setup is too complex at the moment and way too time consuming. It will eventually get there but by then, a lot of these STM32 boards will have passed it in terms of capability and raw processing power so I think I am going to put it up for sale. You are right about the code, if it is bullet proof, it is worth paying for to a certain extent.
"I also have an APM2.5 but haven`t flown it on my quads yet, I am concerned about the stability of the code..."
My situation as well. I bit at the shiny purple lure in the brochures before sniffing the waters. Do you know which of these groups is known for the highest quality code? I foresee tons of stuff going on top of these systems--SLAM, object avoidance, TOS cameras, radars, etc., & would like to learn the one that is most likely to survive a big gobble with the least bloat.
Agreed. APM needs to move to ARM soon. I have a PX4 and just about to order a Quanton, waiting on a REVO from Openpilot (that looks like it might take some time) but the future is in ARM based boards. I also have an APM2.5 but haven`t flown it on my quads yet, I am concerned about the stability of the code and don`t have time to spend hours tweaking anymore, work gets in the way and I simply no longer have the time to waste an entire day tweaking stuff.
I tried to get my hands on a CC3D but gave up. There is so much out there now though, PX4, VRBrain, Quanton, AutoQuad and many more. Tight code will come from a few of these projects and will not cost $500, I still find that a little too steep. I think VRBrain, PX4 and Quanton can do the same thing as AutoQuad sooner than later but for half the price.
http://autoquad.org/wp-content/uploads/2012/06/AQ_TimeLine2.png
Looks like they are 2 years ahead of the APM, & have better foresight in predicting the tech. That's worth $300 to me, considering I will spend 100's of hours tuning, debugging, building, etc.
Is the APM an inexpensive advanced system, or an expensive, hyped & marketed for the masses, system designed to make a few people a lot of money? The Openpilot is available for $100, has a modular, clean architecture, & there are mass produced read to fly complete systems in this price range as well.
I hear you but for that price, men, it needs to be able to make espressos as well!
Yes, I don't know if double+ the price = double++ the goodness. I'm just learning about what's out there. I'd like to find some comparative videos & software cleanness/modularity/architecture, etc. EKF is not the same thing as UKF. The swiss PX4 software already does EKF (not the APM port). Personally a couple of hundred bucks doesn't matter to me relative all the other costs & time investments. The code is by far the most important. By that, I am not referring to feature bloat.
Nice, except for the $530 price tag, that is a bit out of reach for most. I am sure AutoQuad is a very capable board but with all the new stm32 based boards coming out at one fifth of that price, except for the exceptional GPS accuracy, I don`t get it. Maybe the EKF code is very advanced but it still is a very steep price to pay. Is there some special hardware on this board that can justify $530?
Sensor fusion has been around for decades, & this group has not implemented a UKF as demonstrated in this article. The AutoQuad does, however:
http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A9096...
A practical SPK... That's something worth bragging about. Is anyone in this group working on multiprocessor systems?