A Data-Efficient Deep Learning Approach for Deployable Multimodal Social Robots

TL;DR

This paper presents a data-efficient deep learning approach enabling social robots to learn multimodal games like Noughts & Crosses with minimal data, robust perception, and competitive strategies, validated through human and automatic evaluations.

Contribution

It introduces novel algorithms for visual tracking and policy learning, demonstrating effective, data-efficient training of humanoid robots for multimodal game interaction.

Findings

High winning rates in game play surpassing DQN baselines

Robust visual perception is crucial for successful interaction

Approach can be extended to other games with similar success

Abstract

The deep supervised and reinforcement learning paradigms (among others) have the potential to endow interactive multimodal social robots with the ability of acquiring skills autonomously. But it is still not very clear yet how they can be best deployed in real world applications. As a step in this direction, we propose a deep learning-based approach for efficiently training a humanoid robot to play multimodal games---and use the game of `Noughts & Crosses' with two variants as a case study. Its minimum requirements for learning to perceive and interact are based on a few hundred example images, a few example multimodal dialogues and physical demonstrations of robot manipulation, and automatic simulations. In addition, we propose novel algorithms for robust visual game tracking and for competitive policy learning with high winning rates, which substantially outperform DQN-based baselines. While an automatic evaluation shows evidence that the proposed approach can be easily extended to new games with competitive robot behaviours, a human evaluation with 130 humans playing with the Pepper robot confirms that highly accurate visual perception is required for successful game play.