Prediction-based methods for teleoperation across delayed networks

The remote nature of telepresence scenarios can be seen as a strongpoint and also as a weakness. Although it enables the remote control of robots in dangerous or inaccessible environments, it necessarily involves some kind of communication mechanism for the transmission of control signals. This communication mechanism necessarily involves adverse network effects such as delay. Three mechanisms aimed at improving the effects of network delay are presented in this paper: (1) Motion prediction to partially compensate for network delays, (2) force prediction to learn a local force model, thereby reducing dependency on delayed force signals, and (3) haptic data compression to reduce the required bandwidth of high frequency data. The utilized motion prediction scheme was shown to improve operator performance, but had no influence on operator immersion. The force prediction provided haptic feedback through synchronous forces from the local model, thereby stabilizing the control loop. The developed haptic data compression scheme reduced the number of packets sent across the network by 90%, while improving the quality of the haptic feedback.