As expected, time delays threaten the performance and comfort of telepresence systems. Such latencies have already been discussed in terms of visual rendering (Section 7.4) and virtual world simulation (Section 8.3.2). A networked system causes new latency to be added to that of the VR system because information must travel from the client to the server and back again. Furthermore, bandwidth (bits per second) is limited, which might cause further delays or degradation in quality. For reference, the average worldwide travel time to Google to back was around 100 ms in 2012 (it was 50 to 60ms in the US) . Note that by transmitting an entire panoramic view to the user, the network latency should not contribute to head tracking and rendering latencies.
However, latency has a dramatic impact on interactivity, which is a well-known problem to networked gamers. On the other hand, it has been found that people generally tolerate latencies in phone calls of up to 200 ms before complaining of difficulty conversing; however, they may become frustrated if they expect the robot to immediately respond to their movement commands. Completing a manipulation task is even more difficult because of delays in hand-eye coordination. In some cases people can be trained to overcome high latencies through adaptation, assuming the latencies do not substantially vary during and across the trials . The latency poses a considerable challenge for medical applications of telepresence. Imagine if you were a doctor pushing on a scalpel via a telepresence system, but could not see or feel that it is time to stop cutting until 500 ms later. This might be too late!
Steven M LaValle 2016-12-31