About this document

Sparse Diffusion Policy for Robot Learning by lkc.thu17 is a document available to read on EtoBox.

The Sparse Diffusion Policy (SDP) introduces a flexible and efficient framework for multitask and continual learning in robotics by utilizing a Mixture of Experts (MoE) architecture within a transformer-based diffusion policy. SDP selectively activates experts for specific tasks, significantly reducing computational costs and preventing catastrophic forgetting when learning new tasks. Extensive experiments demonstrate that SDP excels in multitask scenarios, maintains high performance in continual learning,

Author

lkc.thu17

Language

EN