A Shared-Control Framework for A Human-Robot Front-Following Behaviour in Unknown Dynamic Environments

Abstract We present a modular framework for enabling a mobile robot to follow a user from the front inside unknown, dynamic and relatively obstacle-laden environments. This behaviour is developed in the context of robotic assistance and rehabilitation for people with mobility impairments. The system incorporates shared-control coupled to an efficient local planner with an intention-reading algorithm recognizing the user intent in undecidable path branches. The components and interconnections are discussed in detail. Field trials with real users are presented, showing that the assistive functionality of the framework encourages users to walk closer to their natural pattern, as compared to an approach with no intelligent features. Preliminary observations suggest that the assistive mode may reduce the cognitive effort required from users. Furthermore, the success rate of the system in identifying user intent is demonstrated to be very high. It is also shown that in certain situations the user-robot pair exhibits complex interaction dynamics, in which their roles are fluid and interchangeable.