Extended Kalman Filter State Estimation for Aerial Continuum Manipulation Systems

The primary goal of this letter is to address the state estimation problem for dual-arm tendon-driven aerial continuum manipulation systems (ACMSs). While the state estimation problem for conventional rigid aerial manipulation systems (AMSs) has been addressed, parameter estimation remains a significant challenge for the recently introduced ACMS platform. Compared to AMSs that utilize arms’ encoder data, ACMSs with flexible arms are not equipped with such sensors. As a result of the requirement for external sensors such as vision systems, measurement challenges may arise in ACMSs cases. Additionally, the dynamics of ACMSs are substantially more complicated, coupled, and nonlinear, posing additional barriers to tackling the estimating problem at hand. This letter proposes integrating deep neural networks with the extended Kalman filter (EKF) technique to enable real-time applications of the method. Simulation results demonstrate the performance of the suggested learning-based EKF approach.