Topologically-directed navigation

  • Authors:

  • David Rawlinson;Ray Jarvis

  • Affiliations:

  • Intelligent robotics research centre, monash university, melbourne, australia.;Intelligent robotics research centre, monash university, melbourne, australia.

  • Venue:
  • Robotica
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recent advances in simultaneous localization and mapping permit robots to autonomously explore enclosed environments and, subsequently, navigate to selected positions within them. But, for many tasks, it is more useful to immediately navigate to goals in unexplored environments, without a map. This is possible if a human director can describe the ideal route to the robot using grounded symbols that both parties can perceive directly. In this paper, a mobile robot is autonomously navigated to many locations in a cluttered laboratory environment by a variety of routes. A series of topological navigation instructions are provided in advance by the director, in a form that can be expressed verbally and translates easily to software representation. The instructions are based on the perception of spatial affordances available to the robot, namely nearby junctions and edges in a pruned Generalized Voronoi Diagram. The operator can generate the instructions by viewing or imagining the environment without any measurements. Only three to five instructions are needed to navigate anywhere in our laboratory. The instructions contain only topology. No spatial measurements or environmental data such as landmarks are provided to the robot.