Experimental evaluation of sliding mode and EKF observers for network delay compensation in bilateral control
Network delays in the feedback and control loop can give rise to loss of transparency and instability in teleoperation and bilateral control systems, especially with the use of the internet as the medium of communication. This is a current problem of bilateral control, with numerous approaches taken to improve the performance in the face of network delay. This study is based on the more recent approach of taking communication delay effects into account as disturbance, and addressing the problem via the design of a disturbance observer. To this aim, a novel sliding-mode observer (SMO) and an EKF observer are developed and tested for improved performance in bilateral position control systems. The SMO runs on the master side and estimates the slave position using delayed measurement feedback from the slave side. The estimated feedback is used in a PD controller, also running on the master side, the output of which is sent to the slave through the internet as control input for trajectory tracking dynamics. As another contribution to bilateral control, in this study, an extended Kalman Filter (EKF) is also developed to estimate the variable load changes on the slave side, which are further compensated by a “plus” term added on the slave side to the PD control input sent by the master. The designed SM and EKF based bilateral control approach is tested with experiments conducted on a slave system comprising a single link arm under gravitational load. For added challenge, the system is tested for bi-directional load and reference trajectory variations and for both constant and random measurement and control input delays of 1-2 seconds in all cases. The experiment results demonstrate that the designed SM and EKF based observers perform very well under no control input delay, and well- known slave parameters and initial conditions. The developed system performs reasonably well with bi-directional load and reference variations under parameter uncertainties, and definitely maintains stability even under simultaneous random measurement and control delays of 1-2 seconds, which is considerably more than the considered delays in current literature.