Homebuilt ROVs Forum

Okay so we have all thought about putting a GPS on an ROV, but there are many complications; really the only way for a GPS to work would be to tow the GPS directly above the ROV.

Instead of relying on the GPS alone, would it be possible to use an IMU?
Take an initial GPS reading when we are above the water, use the IMU to measure all movements, and calculate the new position. When we resurface and if the GPS can get satellites quick enough, correct for error. It's dead reckoning, just like what submarines use.

For those who haven't heard of an IMU, here is a quick link(I just found out about them a few days ago, only have done initial research):
https://www.sparkfun.com/pages/accel_gyro_guide

Thanks for any ideas! Really excited to hear back on this one

Been doing more research into this, and this is how it would work:

1. Using a GPS, get the (almost) exact latitude and longitude coordinates. May have to wait a minute or so after turning on to get a lock.

2. After getting a lock, turn off the GPS, and turn on the IMU, which consists of a 3-axis Gyroscope measuring Angular VELOCITY, a 3-axis Accelerometer measuring lateral/vertical/horizontal ACCELERATION, and the 3-axis magnetometer which acts like a tri-axis compass. The magnetometer is for drift compensation of the Gyro, although I am not sure yet how that exactly works. Drift should be minimal anyway with an update rate on the gyro of 400 kHz, on the accelerometer 3200 hZ, and magnetometer 160 hz. To find the Velocity we just integrate (using Riemann sums, specifically trapezoidal because it is more accurate), and to find distance, we integrate AGAIN, using velocity this time, with initial velocity as 0 (we would wait for the GPS to have a lock before placing in water). To find the angle the ROV is moving at we take the angular velocity and integrate that once.

3: Now we just use vectors to find the distance/direction traveled from the original position. Ignoring the magnetometer because it is for correction, the slowest update speed for the important components is 3200 times a second, which should give minimal error.

Its called inertial guidance system and is at the moment not possible with sensors in our budget-class...

Theroetical easy to do, practically the IMUs we (can) use are drifting to much or not sensitiv enough.

I brought up the spec sheet for the components on that board, and the accelerometer measures about 125 LSB/G at 4G(I'm assuming that's the max we will experience) detecting inclination changes of 1°, giving me precision down to the milli-G, and the gyro measures 14.375 LSB/°/sec, giving accuracy to the tenth of a °/sec. I think that may be accurate enough . There is always some noise, but there are also open-source filters that have been created to eliminate noise. Is there someone who has done it yet? I know Ken is working on it HERE.

It seems we have access to sensors that are good enough for our purpose. However there is noise in every sensor, and we have to apply a filter to it. One of the more popular ones is a Kalman filter. I'm also looking at this IMU from PNI. They have it preloaded with their own algorithm (Wish I knew what it is, doubt they will tell me, although straightforward its a Kalman filter according to the website). There is a really nice whitepaper on their website that outlines the sensor/research they did. They used it for position finding too, not just finding the orientation (Euler Angle) of an object. This is only $100,and was used on an AUV by Cornell University grad students too. There is a way of doing all the calculations through matrices too, but I am also not sure exactly how that is done.

I have no clue what I am doing... still in the R&D phase but Doug over at SVSeeker has what looks to be a pretty cool setup for IMU.

He used this: http://store.3drobotics.com/products/arduimu-v3

Here is a video of it working on a mockup (using airplane propellers in the air instead of water to test the code)... pretty cool, almost an AUV.

Hi TwoBeaner. With regards to IMU based on multicopter sensors, I agree with ROVER3D's comments in his post earlier in this thread. Elaborating, I fear that because ROVs spend alot of time moving very slowly, drift would become too much of a problem. You could overcome accumulated errors in a way by logging IMU position when a static target comes into view (a target you are spending time checking out with camera for example), then re-load that position just before moving out of visual contact. Personally I would be choosing a pinger-based approach to navigation, or using such to assist the cheap IMU. http://en.wikipedia.org/wiki/Underwater ... ing_System Variations on this using GPS mounted on the topside boat would all be potentially quite cheap to do, and no problems with drift. Just my 2 cents as they say...
Last year I bought, along with some more motors and ESCs etc etc (still in packaging ) a couple of RC helicopter heading-hold gyros. These were for me to experiment with the "new" ROV's bow and stern thrusters, lateral and vertical. The idea is to make control easier for the operator, not having to watch and correct heading and to a lesser degree elevation. I figured that the heading-hold gyros could be integrated into the software easier than starting from scratch with bare gyros. These days about $15.

Ross