Toward Socially Aware Robot Navigation in Dynamic and Crowded Environments: A Proactive Social Motion Model

Safe and social navigation is the key to deploying a mobile service robot in a human-centered environment. Widespread acceptability of mobile service robots in daily life is hindered by robot's inability to navigate in crowded and dynamic human environments in a socially acceptable way that would guarantee human safety and comfort. In this paper, we propose an effective proactive social motion model (PSMM) that enables a mobile service robot to navigate safely and socially in crowded and dynamic environments. The proposed method considers not only human states (position, orientation, motion, field of view, and hand poses) relative to the robot but also social interactive information about human-object and human group interactions. This allows development of the PSMM that consists of elements of an extended social force model and a hybrid reciprocal velocity obstacle technique. The PSMM is then combined with a path planning technique to generate a motion planning system that drives a mobile robot in a socially acceptable manner and produces respectful and polite behaviors akin to human movements. Note to Practitioners - In this paper, we validated the effectiveness and feasibility of the proposed proactive social motion model (PSMM) through both simulation and real-world experiments under the newly proposed human comfortable safety indices. To do that, we first implemented the entire navigation system using the open-source robot operating system. We then installed it in a simulated robot model and conducted experiments in a simulated shopping mall-like environment to verify its effectiveness. We also installed the proposed algorithm on our mobile robot platform and conducted experiments in our office-like laboratory environment. Our results show that the developed socially aware navigation framework allows a mobile robot to navigate safely, socially, and proactively while guaranteeing human safety and comfort in crowded and dynamic environments. In this paper, we examined the proposed PSMM with a set of predefined parameters selected based on our empirical experiences about the robot mechanism and selected social environment. However, in fact a mobile robot might need to adapt to various contextual and cultural situations in different social environments. Thus, it should be equipped with an online adaptive interactive learning mechanism allowing the robot to learn to auto-adjust their parameters according to such embedded environments. Using machine learning techniques, e.g., inverse reinforcement learning [1] to optimize the parameter set for the PSMM could be a promising research direction to improve adaptability of mobile service robots in different social environments. In the future, we will evaluate the proposed framework based on a wider variety of scenarios, particularly those with different social interaction situations and dynamic environments. Furthermore, various kinds of social cues and signals introduced in [2] and [3] will be applied to extend the proposed framework in more complicated social situations and contexts. Last but not least, we will investigate different machine learning techniques and incorporate them in the PSMM in order to allow the robot to automatically adapt to diverse social environments. © 2017 IEEE.