What kind of machine learning method is suitable and efficient to identify pattern of water flow fast and accurate? The sensor signals collected are total pressure around a underwater vehicle. Thank you for sharing your views.
Sorry to be negative, but I doubt the correctness of your presumption: IMHO the water flow (aka water speed) is of lower concern for a fish, as it e.g. "knows" how much power it is investing to hold position resp. to advance. And - admittedly depending on the individual kind of fish - the movements of the fish itself will overlay any pattern created by the water flow as a fish in general has to be considered a non-rigid object.
For a fish, the vortexes created by predators/prey are the topic of interest that is primarily sensed by the lateral line.
Regarding your intention: I'd start with registering a lot of data of all sensors, moving the vehicle with different speeds and under varying environmental conditions (e.g. under the influence of cross currents with varying speeds).
Having gathered this data, continue with analyzing what to consider - and especially what to eliminate, and how. Though one would expect some kind of pressure profile depending on hull contour and speed, I bet there will be some unpleasant surprises within the registered data.
One thing should be clear: machine "learning" without putting prior knowledge into the learning algorithm does not really work.
@U. Dreher
Thank you very much for your answers. The lateral line of fish is usually to sense the moving target in the water. The procedure of processing sensor is described as you said. The sensor data I use is pressure data, which generally not so precise. As far as I know according to the identification of pressure datas for cfd simulation, the magnitude is not likely to be determined accurately. The direction of the flow is easy to be identified. And thank you for your advise on using machine learning, I will try to find more prior knowledge about research topic when using machine learning method. Another question is do you know more about how to help an underwater vehicle working better in water from the view of feeling environment better.
Hello,
if you feel that speed and position information would help to increase the usability:
as GPS/GNSS is no option, you might try to integrate a 3 axis accelerometer + 3 axis gyroscope MEMS sensor. (I somehow doubt that a 3 axis geomagnetic sensor would add reliable information, but you could at least try.)
Such 6 / 9 DOF sensors are available - fully integrated - e.g. from Bosch sensortec (www.bosch-sensortec.com).
Be aware that it is quite 'tricky': the sensors are delivering dv/dt (aka acceleration) signals. Thus speed calculation requires integrating the signals, position dead reckoning double integration. Thus measurement errors and noise will also be integrated/double integrated, requiring precision measurements and some noise/error cancellation techniques. But if it helps ...
Wish you success - and have fun!
@U. Dreher
Hello, thank you for your answer. I have added a 9 DOF MEMS imu into the measurement system. And the attitude of vehicle is obtained using AHRS method. The measurement is not so precise because the errors and noise are contained. Because of this, I resolve to lateral line of fish to help vehicle feeling underwater enviroment better. Lateral line can collect sensor informations outside, while IMU is a kind of internal sensor measurement unit which cannot interact with the enviroment outside. Thus, I need an kind sensor or system that is not so expensive and can add more information to the estimation of position of vehicle.
Depending on your application parameters, an acoustic triangulation might be applicable: setting up at least 3 transmitting beacons on the surface and some receivers on your sub would allow for acoustic triangulation - giving a 3D position relative to the beacons.
This method has its limitations (especially regarding operations offshore or in greater depths), but may come handy otherwise.
Regards
@U. Dreher
Hi, the acoustic triangulation method you mentioned might be applicable, but it is a little tricky for me to get related devices. I will try IMU to help me get the positiion of underwater vehicle, and maybe the direction of flow will provide more reliable information for estimation.
Best wishes.
- Badges
- Science topic