Nadeesha Oliver Ranasinghe

Polymorphic Robotics Laboratory
USC Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292
Email: nadeeshr@usc.edu

I received my PhD in Computer Science from the University of Southern California in August 2012. I obtained a M.S. in Computer Science from the University of Southern California in 2003 and a B.Eng in Information Systems Engineering from Imperial College London in 2002. My research interests are in Robotics, Artificial Intelligence and Developmental Learning. My objective is to develop artificial intelligence techniques which enable robots to perform tasks in natural environments with minimal human supervision.

The evolution of human intelligence, in particular the development of skills in a baby has inspired me to design a life-long learning algorithm for a robot. With great insight from my advisor Professor Wei-Min Shen who has also been driven by this desire since his PhD we have developed a logic-based learning framework called Surprise-Based Learning (SBL). Currently, SBL has successfully been demonstrated on a few problems such as simple games, robotic navigation and video surveillance. Although there may be other algorithms that can solve such problems efficiently the key feature of SBL is its ability to detect unpredicted changes, evaluate possible causes and adapt its learned model to reflect the new situation in an unsupervised manner. This gives an autonomous robot some immunity towards runtime changes in its sensors, actuators, environments and goals, as unpredictable faults and failures are unavoidable throughout its lifetime. This algorithm is also capable of performing high level reasoning such as action recognition and gap filling in situations where we can guarantee that sensors, actuators, environments and goals donít change, yet intermittent interference such as noise and gaps exist. However, this progress is only the first step in the road towards my objective, as SBL has exponential growth in its search space for the cause of a surprise with respect to the number of sensors on the robot. I am currently working with my employer to further develop this technology for anomaly detection in real-world applications.

nadeeshr@usc.edu