Skip Navigation

Research Tasks

Learning Homogeneous Knowledge Representation from Heterogeneous Data for Qualitative and Quantitative Reasoning in Autonomy

Learning Homogeneous Knowledge Representation from Heterogeneous Data for Qualitative and Quantitative Reasoning in Autonomy
A robot learning a manipulation task to assist a researcher.
As part of DARPA’s Simplifying Complexity in Scientific Discovery (SIMPLEX) program, this task seeks to develop advanced autonomy for human-robot collaboration that would facilitate a wide variety of applications such as notional human-robot squad patrols, assistive robots, and smart prosthetics. The focus of this research is on developing a model-based cognitive architecture that unifies knowledge representation and cognitive reasoning across many heterogeneous regimes such as grasping and manipulation, task planning, visual perception, human activity understanding, theory of mind, and human-robot dialog. These models are learned from observations of humans naturally performing similar tasks, and transferred to a mobile humanoid robot platform.
Point of Contact: Brandon Rothrock
Sponsored By: DARPA