Perception-link behavior model : supporting a novel operator interface for a customizable anthropomorphic telepresence robot

The customizable anthropomorphic telepresence robot (CATR) is viewed as a potential medium for increasing the social presence in a computer-mediated communication medium towards face-to-face communication. However, there are issues in teleoperation that might deteriorate the efficiency and expressiveness of the CATR. These issues include decoupling, network condition, and limited situation awareness. Decoupling is an event where the operator forgoes part of his nonverbal modalities from a conversation. The second issue is the network condition that can be affected by many factors. For the last issue, the operator’s situational awareness of the remote environment is limited, and it might pose a danger to the remote participants. The outcome is to have a compatible interface for the CATR so that it can perform effectively without degrading its expressivity and overloading the cognitive load of the operator. From a review of the literature, the CATR interface should encapsulate the following: high expressiveness, low cognitive loads, have a deliberate safety mechanism, and ability to substitute. The CATR must express realistic nonverbal cues similar to the operator so that the CATR is representing the operator, in terms of personality and mood. In addition, the interface should not add to the operator’s cognitive load, and it should be intuitive for the operator to use. Next, the CATR should not intimidate or endanger the interactants without distorting the meaning of the gestures. Lastly, the interface should promote a seamless conversation, and the solution is to substitute any missing or undesirable modalities that best fit the current contextual information. None of the existing reviewed interfaces is complete and comprehensive for CATR. The aim of the perception-link behavior model (PLBM) is to overcome the above problems. The PLBM adopts a contactless natural interface as the acquisition approach, which allows the operator to gesticulate expressive gestures using a lower cognitive load. Furthermore, the PLBM employs an unsupervised approach to extract the distributed features, which is more consistent and complete. The PLBM transforms the input signal into a distributed output, which is more informational and expressive. The PLBM is also capable of reconstructing the input signal given the encoded data. In addition, the encoded data can be re-tuned to create a variance of the input signals. Finally, the PLBM can also associate the nonverbal cues from both the operator and audience in real time. This associating function empowers the operator to substitute any undesirable gestures with a coherence gestures retrieved from the knowledge-based. In short, the CATR with PLBM maintains a high-degree of social behavior and social cues; moreover, the PLBM also delivers a seamless communication without compromising the operator’s cognitive workload and audience’s safety. By optimizing and integrating the above modules, the PLBM has numerous advantages over other existing telepresence robot interfaces, e.g. imitation approach. The PLBM can conceal and replace unwanted gesture of the operator during a conversation. Secondly, the PLBM can insert missing data when there is a bad network connection. Lastly, the PLBM can re-tune the incoming data so that the CATR can deliberative and smoothly avoid a collision. In conclusion, the PLBM is superior to existing interfaces given the above scenarios.Doctor of Philosophy (MAE