Anthony O
- Research Program Mentor
PhD candidate at University of Michigan - Ann Arbor
Expertise
perception and computer vision for robotics, image processing, scene understanding
Bio
I’m a fourth-year PhD student at the University of Michigan, where I study robot perception. My research focuses on enabling robots to detect and track objects consistently in videos—a key ability needed for embodied artificial intelligence. I’m especially interested in bridging the gap between visual understanding and real-world robot action. Outside of the lab, I enjoy growing plants and experimenting with different methods of indoor gardening. I find plants to be a fun outlet that complements the structured problem-solving that goes with research.Project ideas
Can Large Language Models perform Task-Planning for Domestic Robots?
This project sets out to investigate the potential for large language models (LLMs) like ChatGPT to solve task-level planning for domestic service robots. While traditional planning algorithms rely on symbolic reasoning and structured representations, LLM-based planning algorithms have the potential to use abstract reasoning and generalize to diverse scenarios. We hypothesize these qualities will make LLMs useful in domestic robotics, where environments are highly unstructured. To evaluate this hypothesis, we implement both traditional and LLM-based planning algorithms and compare their resulting performance trade-offs in realistic robotic tasks.
Robot Perception in Adversarial Environments
all exhibit systematic errors in conditions such as extreme light, weather, or other `long-tail' (i.e. rare) samples. To evaluate this hypothesis, we develop a suite of test cases targeting adversarial environments that robots are likely to encounter and evaluate various state-of-the-art computer vision algorithms in each environment to quantify their respective failure rates.