Home        Store        Docs        Blog

Using current meters to localize a rov


(jakkala kalvik) #1

Did anyone try using a current meter like this one to localize a rov under water. Rotatory encoders on cars lets us localize a car. By combining the data from the encoders and imu you should be able to tell where a car is. So is it possible to do the same with a rov and current meter? The current meter will tell the speed of the water. So if we know how fast we are going will it be possible to predict actual location relative to the start position this way?


Airmar DX900+ for ROV?
(Harold Scadden) #2

There is a lot more involved than a speed sensor. We used something similar on the old style “Speed by Log” sensors on submarines, but you need to have course etc. and other items feeding into a dead reckon circuit. It is possible to do as the sensor logs like this are getting better.

I looked at the specification sheet and you are getting direction in X / Y axis. Now you need to deal with pitch and roll on the platform because the ROV is operating in a 3D environment. That is where a lot of other sensor integration would have to be used. You don’t have to process everything done on the ROV since you have a tether … use just have to get all of the data topside to do your calculations.


(Jacob) #3

Localizing a car in this manner is a lot easier than it seems compared to an ROV. The car localization algorithm has the benefit of many constraints on the car that do not apply to an ROV. A car can only move forward and backward (not laterally, or up and down), pitch and roll don’t matter too much either for these reasons. Importantly, a car’s wheels will not spin if it is stationary, as if it is driving on a treadmill, or the road is moving beneath the car. In contrast, an ROV may be stationary while the water around it moves, like when holding position in a current. This is a problem if you want to use the flow of water to determine your position.

This sensor will definitely provide data with utility, but to actually integrate it and implement a useful algorithm that uses the data will be a complicated task. As Harold suggests, it would be most useful when used in conjunction with other sensor data. I would suggest integrating it with the EKF that runs in ArduSub, but that is no trivial task.

-Jacob


(jakkala kalvik) #4

What if we use 2 of these sensors one in the front and other on the side of the rov, use EKF to clean the data, combine the imu information and depth sensor information. Think that will be enough sensors for localization, maybe use machine learning to estimate the location with info from the sensors should be easier than figuring out the math. Maybe we can measure the current in still water from both the sensors while moving the rov to get a estimate of expected output in still waters and then get data when the rov is moving in water with several known current directions and magnitude. That might be a good place to start right? And I am just asking cause I want to do something fun for my senior design project. figured I would ask here and see if its worth making well try to.


(Pierson) #5

Hey, so physics tell us that we can’t use current sensors to localize an ROV because it only tells you relative current. So if the rov is being displaced by a current, then that component of current velocity is not measured on the vehicle.

Below is a good write up on the techniques for localization that are commonly used and attempted. Perhaps I should append it to include why current sensors don’t work.


(jakkala kalvik) #6

Thanks for the reply. I see why current sensors are not good for localization.


(Pierson) #7

Everyone asks the question at some point. Myself included.