Particle Filter Based Attitude Estimation with Inertial Measurement Units
In this paper,a new attitude and orientation estimation approach using particle filter in connection with inertial measurement units (IMU) is developed.Particle filters,also known as sequential Monte Carlo methods,are model estimation methods based on simulations.IMUs can determine a body’s attitude in an inertial coordinate system.The IMU used in this work takes the measurements of magnetic fields,accelerations and angular velocities.The orientation is estimated in the form of quaternion by means of multi-sensor fusion.A particle filter,in particularly a bootstrap filter algorithm,is implemented for this sensor fusion and state estimation.Particle filters propagate the state,the attitude quaternion.The associate importance weights are updated with the measurements.Furthermore,if a significant degeneracy is observed,resampling techniques are applied in order to retain the particles with high importance weights.To evaluate the functionality and the attributes of particle filter algorithms for attitude estimation,the experimental results of a particle filter are compared with those of an extended Kalman filter,a sigma-point Kalman filter and reference data.In addition,a systematic method for the setting of the parameters in these estimation algorithms is also provided.
particle filter inertial measurement unit sensor fusion bootstrap filter residual resampling.
Chen Zhao Wolfgang Günthner Heinz Ulbrich
Institute of Applied Mechanics Technical University Munich,Munich,Germany
国际会议
International Conference on Modelling,Identification and Control(模拟、鉴定、控制国际会议)
上海
英文
2008-06-29（万方平台首次上网日期，不代表论文的发表时间）