Generalizing to New Tasks via One-Shot Compositional Subgoals

no code implementations • 16 May 2022 • Xihan Bian, Oscar Mendez, Simon Hadfield

In addition to improving learning efficiency for standard long-term tasks, this approach also makes it possible to perform one-shot generalization to previously unseen tasks, given only a single reference trajectory for the task in a different environment.

Imitation Learning

Robot in a China Shop: Using Reinforcement Learning for Location-Specific Navigation Behaviour

no code implementations • 2 Jun 2021 • Xihan Bian, Oscar Mendez, Simon Hadfield

Robots need to be able to work in multiple different environments.

Multi-Task Learning Reinforcement Learning (RL) +1