A vehical is moving on a fix path from point A to point B. I'm continously receiving its GPS location, I want to continuously estimate its ETA at each time stamp with current received data. How can I do this? Can a kalman filter work?
A Kalman filter is useful to combine some expected model behaviour with some observations to get more exact information on the system's current state than any of these two sources can provide on its own.
For your case, the system state might be current position plus current speed, leaving open the question how much time is needed for the remaining distance. Will the vehicle move at constant speed? The speed that we observed over the last few samples? The average speed over the whole distance from the beginning to the current situation? Or if it was accelerating over the last few samples, will it continue accelerating?
So you have to decide on a behaviour model that answers these questions. You need that for the Kalman filter, and you can use it for the extrapolation part from the current situation to the target location.
A less sophisticated approach would be to just compute the average speed from the beginning up to the current place, and compute the time for the rest distance from that speed. I guess that approach would also be OK for all practical requirements, and much easier to implement.