An Approach to Improving Attitude Estimation Based on Low-Cost MEMS-IMU for Mobile Robot Navigation

摘要

An inertial measurement unit (IMU) is an electronic device to measure vehicle states like attitude, orientation, velocity, and position. Recently, many low-cost micro electro mechanical systems (MEMS) IMUs have emerged for only several hundred US dollars [1]. These MEMS-IMUs usually consist of three-axis accelerometers, gyros and magnetometers. In comparison to high-end IMUs (usually used in aerospacecrafts, missiles, rockets and artificial satellites), an entire Inertial Navigation System (INS) can be implemented with smaller size/volume, lower weight and costs. In exchange, they have a rather low accuracy performance due to their large systematic errors such as biases, scale factors and drifts, which are strongly dependent on disturbance and temperature. Hence, the raw signal output of a low cost IMU must be processed to reconstruct smoothed attitude estimates. For many of the mobile robot navigation considered the algorithms need to run on embedded processors with low memory and processing resources. No sensor is perfect and no sensor suits all the applications. However these sensors have themselves distinctive characters, for example, although the gyro is not free from noise, it is less sensitive to linear mechanical movements because it measures rotation. The drift problem is a fatal weakness of gyro. However, the accelerometer does not drift. Therefore, by 'averaging' data that comes from accelerometer and gyro we can obtain a relatively better estimate of the vehicle than we would obtain by using the accelerometer or gyro data alone. The estimation accuracy of IMUs highly depends on the sensor fusion algorithm.