Embedded Files
We are developing autonomous robotics research with theory, software, and experiments.
Our current algorithmic focus is developing combinations of model-based search and model-free learning, where we use real-time computation to make robust and adaptive standard deep reinforcement learning pipelines (train a model offline, freeze weights, and deploy online).
In particular, we are continuing to develop these core algorithmic and application areas:
Planning with Feedback Systems of Search and Learning
Efficient Decision-Making Representations
Space Robotics
Planning with Feedback Systems of Search and Learning
How to combine real-time search process ("System 2 thinking") with offline learned models ("System 1 memory") to get fast training and fast inference?
Efficient Search Representation
How do we construct efficient search trees for a complex, uncertain, and changing physical world?
Space Robotics (experimental facility coming soon!)
Fault Diagnosis Identification and Recovery (FDIR) to unexpected failures in time-sensitive scenarios is practically important and requires new technology for fast and general adaptation.
In-orbit Servicing, Assembly, and Manufacturing (ISAM) is a foundational technology to extend the lifespan of satellites, assemble massive life-seeking telescopes in space, refuel and repair spacecraft on journeys to distant locations
Autonomous small body exploration is important because "Small Bodies Tell the Story of the Solar System" and, sometimes, they even tell stories from outside the solar system.
Page updated
Google Sites
Report abuse