The technological advances made in the past few decades have opened up while new possibilities in many areas. Computational power became cheap, manufacturing more precise. The field of autonomous crafts was not left unaffected by the changes. The process began with the military, where UAVs were used for reconnaissance. Later, the civil sector also started to use drones for aerial photography. Earthbound and airborne unmanned vehicles slowly appeared everywhere. Nowadays, almost every model shop sells remote controlled or robotic crafts for recreational activities.
However, the processing power and mechanical construction is not enough to create an autonomous craft. Software components are also needed. The commercially available, cheap sensors are usually noisy and not very accurate. In typical cases, a drone is equipped with at least a GNSS receiver and an IMU. To use them for navigational purposes, suitable filtering solutions were needed. The GPS measurement usually is too inaccurate and slow for control purposes, while the acceleration integral drifts away, but is precise for short time intervals. There are algorithms for combining the advantages of the two systems.
In this paper I will propose a simple algorithm for improving GNSS measurement accuracy by suppressing an error component. In a few words I will also describe the mathematics of the Kalman filter and show a possible implementation. The concerns of running the algorithms in an embedded environment are also examined for both cases.