Socially Acceptable Pre-collision Safety Strategies for Human-Compliant Navigation of Service Robots

摘要

The human-symbiotic service robot raises new safety issues for autonomous navigation in the presence of people. Traditional obstacle-avoidance algorithms that do not differentiate between people and other objects in the environment may fail to secure human safety in both physical and mental aspects. In this paper, a framework for designing active and human-compliant navigation behaviors is proposed, in which a set of safety strategies is put forward to guarantee human physical safety and mental comfort. Based on a questionnaire study, several important pre-collision safety criteria are selected according to the subjects' interpretation in terms of social acceptability. These criteria are characterized quantitively and computed before being integrated into a unified cost grid for safe motion control. These safety strategies not only explore collision-free paths, but also incorporate social conventions into robot navigational behaviors, which will significantly improves humans' safety feeling, regarding the robot's understandable, predictable and polite behaviors. Moreover, a human-compliant robot navigation system for home-care applications is implemented and described in detail. By utilizing the system, a series of experiments were conducted to verify the effectiveness of the method in producing safe and friendly behaviors in different scenarios. The results of the navigation experiments and the human–robot interaction trials validate the effectiveness of the proposed method as well as the participants' subjective preference to the proposed human-compliant navigational behaviors.