Relative state estimation based on hybrid kinematic model
XIA Rui-can
PEI Hai-long
Abstract：The accurate estimation of relative state typically relies on the precise angular acceleration of a reference platform,which is calculated by applied torque in aerospace scenarios.However,since unknown resistances are common in daily environments,the precise applied torque is almost inaccessible and the angular acceleration is generally approximated through differential angular velocity,resulting in decreased accuracy of relative state estimation.For this issue,this paper introduces a novel relative motion model called the hybrid kinematic model that solely relies on inertial measurements.By incorporating the uniformly accelerated linear motion model with vector kinematics,it accurately predicts relative states independent of the torque and inertia states.Besides,an extended Kalman filter(EKF)is introduced to seamlessly integrate the kinematic model with visual relative observations,thereby achieving precise estimation of relative states.The effectiveness of this approach is assessed using both synthetic and real data.Simulation experiments compare the EKF with advanced methods in terms of trajectory dynamics and relative observations loss,demonstrating that the proposed solution offers superior accuracy and stability.In practical experiments,this method is utilized to implement visual tracking of a six-degree-of-freedom controller in virtual reality applications,validating its exceptional millimeter-level positioning accuracy.The demonstration can be found at
Keywords：relative state estimationmulti-robot collaborationvirtual realityextended Kalman filtersensor fusion
Publication Date：2024-07-28
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：9( 1207-1215 )