I am amazed by the tool that you make, I want to make a tool like this would you help me?
@serhatvarli38693 жыл бұрын
Hello. I saw you in School info system. I have some simple plans. How we contact?
@RegiDoIt3 жыл бұрын
You will find my email at en.ktu.edu/contacts/?searchQuery=Urniezius&searchCaller=header
@serhatvarli38693 жыл бұрын
@@RegiDoIt thank you very very much
@serhatvarli38693 жыл бұрын
@@RegiDoIt What hours can you answer to emails? Because I sent you an email
@tavorobotas4 жыл бұрын
Ar jus naudojate IBVS ar PBVS metodika ? Panašu į image based bet išvedate vaizdą su optinėm distorcijomis tuomet image based netikslus. O stereo kokiam tikslui?
@RegiDoIt4 жыл бұрын
Šitame video tik IBVS. O kituose video su robotu: kombinaciją abiejų.Gal labiau PBVS, nes rekonstruojam erdvę (SfM), kuri yra ant grindų paguldytų juostų konfigūracija. Third party stereo nenaudojam, turim savo low-level high-speed. Stereo reikia, nes kaip žinosi savo position, kai jau erdvė žinoma, todėl po SfM mes stereo naudojam savo positijos tikslinimui ir tada jau pagal PBVS vairuojam, bet SfM dalyje yra IBVS dalis be abejo.
@muhammadnouman23234 жыл бұрын
What is the name of this gimbal?can you send the detail
@RegiDoIt4 жыл бұрын
It is STorM32 BGC v1.3 based.
@umairaziz43994 жыл бұрын
have you bought the body of the gimbal or made it yourself?
@RegiDoIt4 жыл бұрын
Everything was in the same purchase. In this specific case STorM32 BGC and motors were there, mounted. In case you are thinking of something else, contact me.
@oooodles35 жыл бұрын
Can you map this to HID input ? Please get in touch!!
@logeshramasamy18256 жыл бұрын
CAN U PLS SHARE THE CODE I AM DOING A PROJECT OF SIMILAR KIND.I WAS ABLE TO READ ACCELEROMETER DATA BUT I WAS UNABLE TO CLEAR NOISES IN SIGNAL
@leopondernascimento99098 жыл бұрын
ola amigo que tipo de programa você usa pra aparecer na tela obrigado
@RegiDoIt8 жыл бұрын
For user interface using my own C# program.
@RegiDoIt9 жыл бұрын
I added another video with high speed decomposition of free fall from accelerometer measurements. Video is called "High speed decomposition of free fall and motion accelerations in harsh vibrations and shocks" which uses the method of maximization of relative entropy.
@GeankDOORS9 жыл бұрын
Hi, I'm working on a similar project, although using furuta swing-up strategy and cascade PID for stabilization. I'm having problems with the stabilization control "catching" the pendulum after the swing-up brings it to the control zone (+-25º) and the "angle zero" gets lost, I see in your video that the pendulum even after several turns still is able to be stabilized, how do you do that? do you use some logic to reset the encoder that measures pendulum angle? and if so, how do you do it? Any help would be appreciated.
@RegiDoIt9 жыл бұрын
GeankDOORS Yes, I do. Encoders usually have third channel called Z which can be used to recalibrate your drifted reference. But it is a small pulse, so be careful to catch it. It produces a pulse when there is full rotation. So after full rotation you know exact reference.
@GeankDOORS9 жыл бұрын
RegiDoIt Oh I see, my encoders don't have the index channel so I guess I'll need to work on some logic on simulink to do the angle reset. Thanks anyway.
@RegiDoIt10 жыл бұрын
Yes, the idea of the video is to show that gyro's readings are useful when decomposing accelerometer's dynamic vector from the static vector which must be a free fall one. Because there is a direct *linear* and parabolic relationships between these two types of measurement when two accelerometers are present. I have no good answer for one accelerometer case yet. For Kalman filtering basics resource refer to a paper titled as "The Kalman Filter Revisited Using Maximum Relative Entropy" (by Giffin and Urniezius) which will guide to more advanced literature. The main idea to optimally join Hamiltonian dynamics (model) with sophisticated filtering (data).
@jettsui215810 жыл бұрын
hi,may I know how you harsh vibration and and the shocking?many thx))
@RegiDoIt10 жыл бұрын
There is thesis summary available online with details. Search for "Research and Development of Dead Reckoning Localization Method" by Urniezius. Vector orientation math is put in detail in a paper "Iteration free vector orientation using maximum relative entropy with observational priors".
@mohamedthekeys993710 жыл бұрын
let's say I want to draw basic shapes in the computer using one accelerometer, how the error will be? is using 4 or 5 accelero and one gyro useful for the elimination of the accumulated error?
@dvlachy10 жыл бұрын
Impressive performance! Is that approach kind of Unscented Kalmann with a probabilistic matrices definition?
@RegiDoIt10 жыл бұрын
Accurate or precise?! The more information you incorporate, the more precise answer you get. The more information sources you choose, the higher accuracy you get.There are tradeoffs on both contexts. But, yes, in general the more info you pick and collect into knowledge, the longer you postpone the accumulation of errors. To me it is hard to answer your last question. Provide an example and describe your question better.
@Basketballer13810 жыл бұрын
is it possible to gain accurate information of velocity using a combination of accelerometers and gyroscopes? I understand that acceleration is the derivative of velocity... which the first derivative is equal to dx/dt = velocity and these calculations would be subject to accumulative error but would the use of these components give me accurate indication of velocity even if it is only initially until cumulative error kicks in?
@RegiDoIt10 жыл бұрын
First thing to start with is looking at regular 3D vector as the probabilistic one. For first math reference see our paper "Iteration free vector orientation using maximum relative entropy with observational priors". For deeper algorithm description my PhD thesis is the second candidate. Its title is RESEARCH AND DEVELOPMENT OF DEAD RECKONING LOCALIZATION METHOD (2011).
@princesasery11 жыл бұрын
There is something really MICRO, google uFusion AHRS SYSTEM!!
@RegiDoIt11 жыл бұрын
Here is another my video with inverted pendulum. see current video description. Math rocks.
@tedytorres112 жыл бұрын
OMG YOU CAN NOW CREAT A BOM TO DESTROY THE HUMANITY LOL
@RegiDoIt12 жыл бұрын
Yes. I believe algorithm must be stable enough so that a person with disability (shaking, vibrations) could use it too.
@johnwmckown12 жыл бұрын
RegiDoIt, do you think mice implemented with motion sensors will be accurate enough for the cut and paste operations that are routine in word processor applications? By "accurate enough" I really mean "as conveniently usable as a standard table mouse."
@compliantrobotics12 жыл бұрын
@regiDolt Thanks for the encouragement. But I am also very impressed with your imu's accuracy. I have spent a great deal of time researching how to remove drift (through combining the accelerometer readings with the gyroscopic readings like you said) but have had little luck in finding an accurate filtering method or pairing equation. "Kalman" and other filters have not been accurate enough. Could you point me to any good resources or forums you used to derive the estimates' expressions?
@RegiDoIt12 жыл бұрын
@skylergreenman : No invensense software. Only their gyro ITG-3200. I have derived estimates' expressions myself. Removing a drift of gyro is done by utilizing magnetometer or accelerometer data or both, because magnetometer and accelerometer together directly give you an orientation in a 3D space. I saw your gloves project. Very interesting and keep it up. It will be interesting to compare it with my 3D mouse features in the future, when you have something. Good luck!
@compliantrobotics12 жыл бұрын
Did you use Invesense's software and hardware to make the 3D mouse work? And how did you get that software? If you created the software, what "math formulas" did you use. I am working on a 3D mouse project and need to figure out how to remove the drift and gravity from accelerometer and gyroscope readings. Thanks and well done.
@RegiDoIt12 жыл бұрын
Code would contain just math formulas regarding the actual algorithm behavior. Nothing very special. I assume the idea behind it would be more interesting. The approach is probabilistic analysis, i.e. more specifically entropic updating. The main idea is that you take the model and "follow" it. If your model is relevant, your approach eventually reduces the overall uncertainty.