Our lab has a strong interest in developing a holistic computational solution for designing and optimizing cyber-physical systems (e.g., robots). Our research topics include: 1. Proposing continuous, discrete, or hybrid methods for design-space representation and parametrization. 2. Developing fast, reliable, and differentiable physics simulators for efficient evaluation of a design. 3. Researching novel numerical optimization techniques for co-optimizing a cyber-physical system's body and brain. 4. Exploring designs efficiently and effectively in the context of multi-objective performance metrics.
Our group has a strong track record of publishing graphics research papers at top-tier conferences (SIGGRAPH, SIGGRAPH Asia) and journals (TOG). Some topics that we are currently interested in include 1) Physics simulation of all kinds and its applications in machine learning, robotics, and computer vision, 2) Computer-aided design, shape models, shape analysis, and especially their industry-level applications, 3) Appearance modeling, (procedural) material design, modeling, rendering, and visualization, and of course, 4) computational design and fabrication of cyber-physical systems.
Our group actively works on different vision problems in the following domains: 1); Gaze tracking; 2) Novel view synthesis for 3D displays; 3) Multimodal learning (i.e., voice-face, pressure map-pose); 4) Computational photography (i.e., motion magnification, semantic soft segmentation); 5) Datasets for new learning tasks (CAD models, holograms, gaze, flash/no-flash photographs). Recent works have been published in ICCV, ECCV, CVPR, ICML, and NeurlPS.
Our research in fabrication explores new types of fabrication processes and systems that will have a broad impact on how humans design and build things. We combine computational design, machine learning to model, understand, and extend fabrication processes such as additive manufacturing, machine knitting, and conventional manufacturing processes. Our recent research in this area has included developing multi-material manufacturing systems to create 3D shapes out of high-performance functional and structural materials and new manufacturing methods to create garments that sense and classify human activity.
Our group has strong interests in a variety of Robotics research topics, including (1) efficient (differentiable) simulation tools for all kinds of robotic systems (e.g. rigid robots and soft robots), (2) machine learning algorithms (e.g. reinforcement learning) for robotics control, (3) computational design methods for co-optimizing both the structure/shape and the controller of various types of robots (e.g. robotic hand, terrestrial robots, underwater robots, aerial robots, etc.) for single or multiple objectives/tasks, (4) tactile sensing devices and their applications in robotic control systems. Recent work has been published in top-tier robotics and machine learning conferences and journals such as RSS, RAL, ICRA, ICML, etc.