Zero Shot Learning on Simulated Robots

TL;DR

This paper introduces a zero-shot learning approach for robots using a self-model to transfer skills across tasks, significantly improving data efficiency and enabling new task learning without additional data collection.

Contribution

The work presents a novel self-model based method that enables zero-shot transfer of robot skills to new tasks without further environment data.

Findings

Self-models enable accurate state prediction across robot morphologies.

Training on self-models is more data-efficient than real environment training.

Robots can perform new tasks zero-shot using the learned self-models.

Abstract

In this work we present a method for leveraging data from one source to learn how to do multiple new tasks. Task transfer is achieved using a self-model that encapsulates the dynamics of a system and serves as an environment for reinforcement learning. To study this approach, we train a self-models on various robot morphologies, using randomly sampled actions. Using a self-model, an initial state and corresponding actions, we can predict the next state. This predictive self-model is then used by a standard reinforcement learning algorithm to accomplish tasks without ever seeing a state from the "real" environment. These trained policies allow the robots to successfully achieve their goals in the "real" environment. We demonstrate that not only is training on the self-model far more data efficient than learning even a single task, but also that it allows for learning new tasks without necessitating any additional data collection, essentially allowing zero-shot learning of new tasks.