List view

• Expand Graph-based model to learn from different topologies

  Don't really know the problems that are gonna be faced here

  No due date
  0% complete0 open0 closed

• Implement model for Graph State Observation (same embodiement)

  One of our contributions. Implement model that leverages a structured latent space for generating the actions. ** Demonstrations and actions should still have same embodiement:

  No due date

  •3/3 issues closed
  100% complete0 open3 closed

• Implement model for Vector State Observation

  To be directly benchmarked with [Diffusion Policy](https://github.com/columbia-ai-robotics/diffusion_policy). Model to learn from observation in vector form, and generate vector form actions. ** Demonstrations and actions should still have same embodiement:

  No due date

  •1/1 issues closed
  100% complete0 open1 closed

• Implement Pipeline

  Implement pipeline for training, testing and evaluating models with the chosen datasets. ** Consider how the developing model can be evaluated against the benchmarks

  No due date

  •4/4 issues closed
  100% complete0 open4 closed

• Implementation of First Model

  Implementation of first model, to be plugged into the pipeline in order to obtain the first benchmark results.

  No due date

  •5/5 issues closed
  100% complete0 open5 closed

• Implement Baselines

  Use baseline models from previous work to be used on the environment of this project. * https://github.com/columbia-ai-robotics/diffusion_policy * https://github.com/microsoft/Imitating-Human-Behaviour-w-Diffusion/tree/main * https://github.com/intuitive-robots/beso * The ones from https://robomimic.github.io/docs/introduction/implemented_algorithms.html

  Overdue by 2 year(s)

  •

  Due by July 20, 2023

  •5/5 issues closed
  100% complete0 open5 closed