Robustness and Safety of Machine Learning Components in Autonomous Systems (DSI-SRP)
This DSI-SRP fellowship funded Ulysses Yu to work in the laboratory of Professor Xenofon Koutsoukos in the Department of Electrical Engineering and Computer Science during the summer of 2019. Ulysses graduated in 2020 with majors in Computer Science and Mathematics.
The project funded by this fellowship aimed to to investigate perception in machine learning and how it can be utilized effectively and safely in autonomous systems. Autonomous systems such as robotics systems are generally based on a “perception, reasoning, and action” loop. Being able to observe and identify objects in the world is pivotal to an autonomous agent. Deep neural networks provide amazing results for object identification based on vision data and active sensing modalities such LiDAR. However, deep models can be hard to train. Another significant challenge is to characterize their performance and robustness. The goal is to investigate performance and robustness of convolutional neural network used for perception in autonomous systems. Image and LiDAR data will be generated using powerful 3D simulators of the environment and used for training of different neural networks architectures. The project will emphasize robust learning algorithms for investigating how such perceptors perform for previously unseen data.
In addition to receiving support through a DSI-SRP fellowship, this project was supported and facilitated by the DSI Data Science Team through their regular summer workshops and demo sessions.