טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentSazbon Didi
SubjectSensory Based Navigation Motivated by Biological Systems
DepartmentDepartment of Computer Science
Supervisor Professor Ehud Rivlin


Abstract

            Inspired by the remarkable navigational skills of lower animals, referred as ‘biological systems’, this thesis studied various navigational tactics used by different animals in their everyday lives, and concluded that they strongly favor a task driven approach. Many tasks are solved right at the sensory level by utilizing task oriented ‘matched filters’, implying rapid real time solutions that only use important information while filtering out irrelevant information. This concept also suggest a flexible hierarchical structure of navigational capabilities: a set of basic navigational tasks with fast solutions forms the foundation to a set of high level, more complex, tasks.

            Four generic basic navigational tasks are defined here, as motivated by the requirements that any biological system must fulfill in order to move around in its environment and stay alive. The four navigational tasks are analyzed using the biological paradigm and show that the use of matched filters appears to provide a good approximation to the characteristics observed in the biological systems. From an engineering point of view, the usage of matched filters is attractive since it implies real time solutions that could be implemented in hardware. Furthermore, some of the algorithms and techniques developed here are implemented optically, hence supporting sensory level capabilities. The solutions to the basic navigational tasks presented here are highly flexible and can be used by other tasks as building blocks or alternatively, could be modified as necessary. Thus, this thesis, inspired by the biological systems and their extraordinary matched filters driven solutions, contributes in its concept of combining real time implementations and basic navigational tasks.