Experimental study on the curve evolution method for noise filtering in the inertial state estimation

With more and more autonomous mobile robots being used for the tasks constrained by uncertainties, there is a need to utilize the on-board sensors to estimate the robot's states to enhance the robustness of the state estimation when the sensors relying on external beacons (e.g. GPS) becomes unreliable or even fail. A common type of on-board sensor is inertial measurement unit (IMU) which pick up signals related to the robot motion like acceleration and orientation. It is well known that drift occurs in the state (velocity and position) estimation from those measurements polluted with noises. Recently an innovative method has been proposed to filter out the noises in the IMU measurements based on the curve evolution technique. This paper presents an experimental study of the method in different working conditions with a commercial IMU.