A Generative Approach for Socially Compliant Navigation

Abstract

Robots navigating in human crowds need to optimize their paths not only for the efficiency of their tasks performance but also for the compliance to social norms. One of the key challenges in this context is due to the lack of suitable metrics for evaluating and optimizing a socially compliant behavior. In this work, we propose NaviGAN, a generative navigation algorithm in an adversarial training framework that learns to generate a navigation path that is both optimized for achieving a goal and for complying with latent social rules. Different from the reinforcement learning approaches who only covers the "comfort" aspect of socially compliant navigation, and inverse reinforcement learning approaches who only covers the "naturalness" aspect, our approach jointly optimize both "comfort" and "naturalness" aspects. The proposed approach is highly interpretable and demonstrates superior quantitative performance in sets of experiments. We also demonstrates qualitative performance on a ClearPath Husky robot and perform extensive experiments on real-world robotic setting. The video of qualitative robot experiments can be found in the link: https://youtu.be/61blDymjCpw