Research

Bayesian Learning for Robotic Collaboration and Autonomy

Intelligence is critical for robots. In the past decades, deep learning technology has achieved great progress and has demonstrated its powerfulness, especially in perception tasks. However, classical deep learning techniques have limited capability to improve robotic intelligence, because the real world robotic applications require inference, face noisy data, and have limited resources for learning.   

RoCAL is interested in using probabilistic learning techniques, especially bayesian learning, to break these bottlenecks. Bayesian Learning learns the model, $\mathrm{P(Y|X)}$, from data. This technique is adaptive to perception and model noises and various sizes of training data and can deal with  casualty as the Bayesian network is directional.

RoCAL mainly focuses on three types of applications:

• Medical Information Processing

• RoCAL is interested in finding the information and identifying the key factors that closely correlate with surgical outcomes in endoscopic sinus and skull base surgeries.

• Supervised Autonomy in Robotic Surgery

• RoCAL is interested in correlating surgical outcomes with surgical planning and surgical motions and builds supervised autonomous surgical robots to maximize surgical outcomes.

• Simultaneous Localization and Mapping (SLAM)

• RoCAL focuses on building precise and robust graphs, through improving feature detection and data association reliability, adapting to environmental changes, and collaborative mapping.