A&A Ph.D. student Niyousha Rahimi received a UW + Amazon Science Hub Award to support her doctoral research in robotics and AI.

Niyousha Rahimi, A&A Ph.D. student

In 2018, after graduating with a master’s degree in mechanical engineering at the UW, Niyousha started her doctorate in A&A, conducting research in the Robotics, Aerospace and Information Networks (RAIN) Lab. Part of her focus includes developing systematic approaches for embedding learning algorithms in robotic control systems with safety guarantees. These systems are challenging to analyze and develop because robotic control systems often operate with real-time streaming data in uncertain environments, but techniques such as clustering, matching, feature recognition, and neural network training require batch data/off-line processing.

Using data to learn a map from complex observations to physical quantities (e.g., from images to the position of an object in the environment) simplifies the planning problem when the trained neural network is adopted as a perception module in the control loop for autonomous navigation. For continuous perception maps and nonlinear systems, Niyousha showed that robust control could be used to guarantee generalization: ensuring that the closed-loop sensitivity is bounded proportionally to the continuity of the perception errors certifies stability.

Niyousha and her collaborators also formulated a new system theoretic concept to capture how stability-type questions can be addressed over finite-time horizons using real-time streaming data. This theory of data-guided control would allow means of analyzing stability-type requirements without relying on the underlying dynamic model.

The UW + Amazon Science Hub fellowship provides three quarters of funding to pursue independent research projects. With this fellowship, Niyousha can put her research into practice. In the future she hopes to further investigate nonlinear theory, computational sensing, & safe autonomous navigation in crowds.

Niyousha's faculty adviser, Professor Mehran Mesbahi, says, "I am sure Niyousha will advance foundational techniques at the interface of learning and control with applications to safety critical autonomous systems under this fellowship."