Deep learning techniques to conduct IMU based Position Estimation
The topic is available also in Hungarian.
The primary goal of the project is to provide the estimation of the position of the signals received from an IMU sensor. The reason why the accelerometer is involved is, e.g.,
- the instrument cannot be monitored optically, as it is being moved in the soil,
- the sensor is embedded in a smart watch (fall detection),
- the sensor is embedded in an UAV, its goal is to reach its target with high precision.
Theoretically speaking, the position can be calculated by the double integral of the acceleration. However, in practical applications, the actual sampling frequency and the noise can have a strong influence on the quality of the estimation.
The student will work on deep learning techniques that calculate the estimation of the position based on the acceleration history. Various deep learning algorithms are to be developed and evaluated. The involved techniques are, e.g., multilayer feed forward network, recurrent neural network.