Social navigation framework for assistive robots in human inhabited unknown environments


Engineering Science and Technology, an International Journal, vol.24, pp.284-298, 2021 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 24
  • Publication Date: 2021
  • Doi Number: 10.1016/j.jestch.2020.08.008
  • Title of Journal : Engineering Science and Technology, an International Journal
  • Page Numbers: pp.284-298
  • Keywords: Social navigation, Human-aware navigation, Human-robot interaction, Social robotics, Mobile robots, ROS, TRACKING


© 2020 Karabuk UniversityIn human-populated environments, robot navigation requires more than mere obstacle avoidance for safe and comfortable human-robot interaction. Socially aware navigation approaches become vital for deploying mobile service robots in human interactive environments, where the robot operates in interaction with human implicitly or explicitly. These approaches aim to generate human-friendly paths in human-robot interactive environments considering social cues and human behaviour patterns. This paper proposes a social navigation framework for mobile service robots, maintaining humans’ safety and comfort while navigating towards the goal location in human interactive environments. Our main contribution is that the presented social navigation framework is designed to be used in human interactive unknown environments. To achieve this goal, we use a variant of a pedestrian model called Collision Prediction based Social Force model (CP-SFM). This model is particularly developed for low or average density environments and takes the motion of the people tracked in the environment into account during the navigation. The model is employed as a local planner to generate human-friendly plausible routes for our service robot in corridor like indoor environment scenarios. A variety of different extensions and improvements of the conventional social force model are employed in the implementation stage. A novel improvement in producing multi-level mapping, identifying obstacle repulsion points and adopting CP-SFM for application in motion planning as local task solver is presented. The whole framework has been implemented as ROS nodes, and tested both in real world and simulation environments and successfully verified based on the obtained results.