3D Tracking with IMU

[Tyrant001MrX]

Thanks for your fast answer. If it's the latency, How I can improve it?I want to get the same speed of the cube in the video. Sorry for asking the same cuestion in my last post, my doubt was to obtaining it from the quaternions of your algorithm.

Thanks for everything, I hope not to being so disturbing for you.

[SebMadgwickResearch]

@Tyrant001MrX Latency (a pure delay) has nothing to do with my algorithm so I can’t help you here. It is probably due to ‘slow software’ in the same way that a PC game would appear “laggy” on an under-spec’ed machine.

[afazzo]

Impressing...I'm trying to implement your algorithm on PC using a BMA-180 plus IT-3200 IMU, and I'm quite optimist about the result. I have just a residual "vibration" in the estimated orientation, possibily due to the very simple forward Euler time stepping scheme you used.

I have a question about the position tracking... I wasn't able to find anything about it on your Ph.D thesis . Which algorithm did you use? Did you measured the error of the position at the end of the walk? Thanks.

[SebMadgwickResearch]

@afazzo My Ph.D thesis has not been published yet. I don’t know if the algorithm I used has a name or is documented anywhere. The final position error was ~1m after a ~100 m circuit including 2 floors (inc. 2 stair cases). I have since discovered that this is an extremely low error and is not representative.

[SebMadgwickResearch]

@afazzo My Ph.D thesis has not been published yet. I don’t know if the algorithm I used has a name or is documented. The final position error was ~1m after the ~100m circuit including 2 floors (inc. 2 stair cases shown in video). I have since discovered that this is an extremely low error and is not representative.

[Tyrant001MrX]

Sir, I'm so interested in your work because I'm trying to develop a system that measures sea waves height. I have the accurate attitude of the system using gyro and accelerometer measurments integrated in a Kalman filter. But I have problems de-rotate the system in order to change body axes to navigation axes(Euler) . What Is the best way to obtain the accelerometer forces in to navigation axes or horizontal earth surface axes?Can you show me an algorithm example using three attitude angles?

[SebMadgwickResearch]

@Tyrant001MrX It is achieved though a rotation matrix multiplication, you would want to avoid using Euler angles in such operations. I cannot explain the mathematics here but it is covered in any relevant introductory textbook; e.g. Kuipers’ “Quaternions and rotational sequences”, or Craig's “Introduction to robotics...”.

[Tyrant001MrX]

@SebMadgwickResearch Thanks, I will check those books in order to improve my knowledge in 3D inertial tracking.

[Tyrant001MrX]

@SebMadgwickResearch,Hi again, I'm testing your Quaternions IMU Algorithm, it works fine, but it seems to be more laggy than in the demo video, may be because I´m using a lower sample rate (100Hz, every 10ms), is that reason??. How I can obtain Earth referenced forces measured by accelerometers with the quaternions in order to eliminate gravity influence in the system?

Thanks again.

[SebMadgwickResearch]

@Tyrant001MrX If it seems “laggy” then this will be due to latency, not the update frequency. I have already responded to your “Earth referenced forces” question.

[raydks]

sir, what did you do to eliminate the influence of gravity?

[SebMadgwickResearch]

@raydks Fusing together the accelerometer and gyroscope data provides a measurement of the sensor’s attitude relative to the Earth surface. If you know the relative direction of the Earth then you know the direction of gravity and can subtract it.

[raydks]

@SebMadgwickResearch

sir, thanks for the answer, I try what u said and got a close result. I used the MARU u released in google project to calculate the gravity by multiplying the rotation matrix to the acceleration in each acceleration axis obtained at the beginning. However, there is still some DC gravity cannot be eliminated. Could u help? thanks a lot

[JackOhara91]

i have a question for you sir: can an IMU track the attitude of an airplane going 580km/hr by using update of its position and altitude from GPS unit "i cant use a gps alone because of the limited transfer rate of 10 Mbytes/s and it may skip some waypoints" the gps will update the initial frame from which the IMU is using dead reckoning..i'm one year away from my graduation project & i'll appreciate very much your help..very nice video by the way thanks for sharing

[SebMadgwickResearch]

@JackOhara91 It is common to bring together IMU and GPS data (and barometer etc.) to provide improved estimates of states (e.g. position, velocity, etc). It sounds like you are describing your answer in your question.

[JackOhara91]

@SebMadgwickResearch thank you for replying it's just i have little experiance in electronics i'm mechanical engineer but i'm learning electronics for this project & i dont know where to start in INS..can you tell me about some good books to design the algorithm and INS circuit, the reason why i asked my first question is because i didnt think its doable, so can you tell me where to start thanks in advance

[SebMadgwickResearch]

@JackOhara91 Yes, if the GPS is not comprised by extreme altitude and velocity. GPS updates are infrequent (e.g. 10 Hz) but these can be interpolated with the IMU measurements (e.g. at 500 Hz) to provide you with your "missing" data. Autonomous UAV projects are very popular right now, there is plenty of information available on the web and many dedicated websites; e.g. DIYdrones

[JackOhara91]

@SebMadgwickResearch thank you sir you've been both helpful and informative, i've bought a book "Strapdown Inertial Navigation Technology" seems like a good place to start, i've read a few pages and it looks alot like robot mechanics!

[dameron22]

would you mind sharing the algorithm?

[SebMadgwickResearch]

I am currently unable to share the code but I have explained the principles in replies to other comments.

[visotube]

Hello Boss U R Great Supper Video How Much Length U have Assigned For Every Axis To Get The Position

[garlest]

When calculate the position, I use the code below.

v_x = (a_x + pre_a_x) / 2 * sampling_rate + pre_v_x;

p_x = (v_x + pre_v_x) / 2 * sampling_rate + pre_p_x;

(a_x = acceleration, v_x = velocity, p_x = position, pre = previous)

But, have a lot of integration errors.

How do you do to integration?

Could you tell me how to integration....

Thank you.

[SebMadgwickResearch]

I explain the principle in my reply to other comments. I cannot go into any further detail here.

[garlest]

It seems to be detected with considerable accuracy. Great!

I think that you use the 9 DOF sensor for measuring of orientation and use only the accelerometer for measuring of position.

According to your post, you reset to zero speed constantly in order to reduce the integration error.

In order to reduce the integration error, are you using other algorithms?

If it is convenient to you, could you explain how to reduce the integration error?

thank you.

[SebMadgwickResearch]

The integrated acceleration (in the Earth coordinated frame) is 'forced' to zero each time the foot is detected as being stationary (via accelerometer). Accumulated biases in this velocity are interpolated and subtracted between stationary periods to provide the compensated velocity. This velocity is integrated to yield position.

[Shenkway]

many thanks for sharing this video! I want to make something very similar for my diploma thesis ! Maybe u could help me with rendering the graphics (3axis) in tracking.

[neslorelyks]

how easy would it be to include a model of the foot?  i would love to use this software if the box model could be replaced with a model of the device being monitored?

[SebMadgwickResearch]

Rendering the graphics of a foot (or anything) and virtual world be technically trivial but require time. It would make a great looking ‘end product’ but I don’t think I’ll be creating this any time soon.

[fatyfleur82]

great work, I want to know for how many time can you make the tracking with your IMU? i mean , it is operational withount drift for 1 hour or 2 or maybe more or less?

because i am working on orientation and position tracking for the meuseum

visit that can take more than 1 our, I want to know if your IMU can do

this task.(i am asking this question because i lnow that the gyro suffers from the drift and bias which changes with time and temperature)

[SebMadgwickResearch]

The translational position will drift continuously; it is just a matter of at what rate. Dead-reckoning over hours would require (expensive) precision sensors.

[fatyfleur82]

@SebMadgwickResearch

and what about heading? you can track it (without drift) for what duration?

[thebrooksy1]

what assumptions are you using to compensate for inaccuracy of the double integration of acc?

[SebMadgwickResearch]

That the foot translational velocity will periodically return to zero with each step.

[maylees]

@SebMadgwickResearch that's maybe the only way to fix it in IMU

[fernandohildebrand]

Wow! And I mean Wow! I was a fan before, now I´m speechless. But what´s the tech? Bluetooth + accel +gyros+ mags?

[SebMadgwickResearch]

I've added details in the video description. You can find more details on the x-IMU on the website.

[fax8]

cool! what did you use for making the graphs?

[SebMadgwickResearch]

MATLAB

[fax8]

@SebMadgwickResearch great! I never used this animated plot feature of matlab even if I know it pretty well.. something pretty interesting for the future. thanks for sharing this video!