I am amazed by the tool that you make, I want to make a tool like this would you help me?

Hello. I saw you in School info system. I have some simple plans. How we contact?

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?

What is the name of this gimbal?can you send the detail

Can you map this to HID input ? Please get in touch!!

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

ola amigo que tipo de programa você usa pra aparecer na tela obrigado

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.

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.

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).

hi,may I know how you harsh vibration and and the shocking?many thx))

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?

Impressive performance! Is that approach kind of Unscented Kalmann with a probabilistic matrices definition?

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.

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?

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).

There is something really MICRO, google uFusion AHRS SYSTEM!!

Here is another my video with inverted pendulum. see current video description. Math rocks.

OMG YOU CAN NOW CREAT A BOM TO DESTROY THE HUMANITY LOL

Yes. I believe algorithm must be stable enough so that a person with disability (shaking, vibrations) could use it too.

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."

@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?

@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!

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.

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.

nice work. any code available?

Thank you