Biologically motivated distributed vision system for autonomous undersea vehicles

ABSTRACT The work described in this paper is a novel approach to undersea autonavigation; current applications are not truly autonomous and if a vision capability is available it is not used to navigate the robot; rather, this is done using sonar and acoustic tracking for dead reckoning methods or odometry which have limited accuracy and always require prior knowledge of the circumstances in which the robot will function. It is clear that, where conditions allow, the use of a suitable vision capability would make information about the environment available at a rate commensurate with the control frequency of the system. To date, there are still significant design and implementation challenges to overcome in the production of machine vision capabilities for autonomous robots and there remains a clear need for multi-disciplinary research to establish the physical principles and design strategies associated with the onboard integration of sensor systems. Traditional machine vision algorithms involve time intensive serial processing and in general cannot deliver the necessary real time capabilities. However, recent trends