Reducing Teleoperated System Data/Image Bandwidth Requirements using a Quantitative Model of UGV Operator Focus of Attention

The need to transmit video information greatly exceeds the current capability to transmit this information for any many military applications. This statement is absolutely true in the teleoperation of semi-autonomous robotic land vehicles where the remote operator needs advanced video compression techniques to support low data rate and supervised driving modes. This proposal advocates the development of a data compression scheme based on an active vision focus of attention model. The overall approach leverages the current understanding that the uniqueness of preattentive parallel features is the primary factor associated with determining visual focus of attention, and is coupled with a current understanding of the processing found within the motion detection channel of the human vision system to produce a technique that is computationally attractive. This overall approach directly leverages Cybernet's previous DARPA work developing an Interface Evaluation System (eye-tracking and operator monitoring). The outcome of this effort is directed at producing software which will function on the Portable OCU Cybernet is building for the DARPA Demo II program, and promises to significantly reduce the data bandwidth required to perform semi-autonomous vehicle navigation on this unit. We believe this project will provide significant technical leveraging to many UGVJPO projects.