View on GitHub

TransparencyRLfD

Expressive robot motions to improve the demonstrations in RLfD

Transparent RLfD

Demo

Robot motions in Learning from Demonstration tutorial

Related robot motion types: attentional motions? uncertainty motions? predicatable motions?

Possible influences: how the learning process evolve over time? the state visited, the values updated

Overview

drawing

Tentative schedule:

1. Demonstration (Jun 15th)

drawing

drawing

drawing

2. Model-free RL (Jul 1st)

Single state-action update

Autonomous Q-learning test on the simulation:

drawing

Multiple state-action update: update whole trajectory

To be decided:

To-do list:

Two different algorithms: to compare these two types of algorithms

3. Learning process evaluation (Jul 15th)

Q-learning:

4. Motion planning (Jul 25th)

To-do lists

  1. Need to find out the possible robot motions that would affect the human perception
  2. Need to implement the related motions
    • Gesturing motions: policy-based drawing
  1. Need to determine the learning model and evaluate the model effects towards the learning

Expected outcome

drawing

Final outcome

drawing

Matlab Reinforcement Learning