State estimation of delays in telepresence robot navigation using Bayesian approaches

Telepresence systems allow a human operator to control and navigate a mobile robot around the remote environment and interact with their audiences through video conferencing. Telepresence robots suffer significant challenges during navigation due to communication time delays. If the time delays are not compensated to estimate the robot pose correctly in the remote site, the robot may crash due to inaccurate pose estimation by the operator. In this work, we propose a Bayesian approach to model such delays using state estimation techniques that are useful for robust navigation. Robot state estimation in dynamical systems is essential in real world applications, as the actual state is undetermined and sensors provide only a sequence of noisy measurements. Extended Kalman filter (EKF) generally acquires an estimate of the true state from noisy sensor measurements. However, when a filtering processor is attached to a network, there is a communication time lag. Additional time is required if there is a need to post process the raw sensor data for updating the state of the dynamical system. As a result a delay is introduced between the acquisition of measurement and its availability to the filter. This paper proposes state estimation techniques of delayed navigation of telepresence robots. Considering a small delay in the system, an augmented state Kalman filter (ASKF) [2] is proposed. As any delayed measurement carries information about a past state, the current state cannot directly be corrected only using the measurement. The past state corresponding to a delayed measurement should be determined before using the delayed measurement in the state estimation. The current state is then corrected after correcting the appropriate past state. We also assume that the delay is not precise and hence the uncertain delay is modelled using a probability density function (PDF) [2]. To our best knowledge, the proposed approach is first of its kind in compensating delay in telepresence robot navigation