Title: On Generalization and Benchmarking on Physical Robots

Abstract:

Robotics research has seen significant advancements; however, the field remains predominantly demo-driven, making direct comparisons between methods difficult without replicating them on individual setups. While many simulation benchmarks exist, they usually feature contrived datasets and do not accurately reflect real-world performance. In my thesis, we address these challenges by proposing a more realistic data assumption for physical robots. We evaluate the generalization capabilities of offline reinforcement learning and imitation learning methods within this framework. We also introduce an initiative towards establishing a real-robot benchmark. Initial results from the benchmark and the launch of a NeurIPS competition highlight the feasibility of such systems.

Committee:

Prof. Abhinav Gupta (advisor)

Prof. David Held

Sudeep Dasari

ZOOM Link: https://cmu.zoom.us/j/5010856764