Localization using Average Landmark Vector in the presence of clutter

Topological maps are used by biological species for navigation in complex and cluttered environments with very primitive sensory inputs and limited computational abilities. The core aspect of topological maps are the properties (usually some environmental features) used to define the nodes of the map. In this paper we use a well known technique adopted by desert ants called visual piloting for autonomous robot navigation. This technique works by extracting a useful feature from the environment and calculating the Average Landmark Vector (ALV) with all such features. This is a first step in using environmental features with ideas inspired from ants/bees, and will form the basis of building and using topological maps for robot navigation. We present ALV in terms of a gradient descent mechanism. We also present results from an implementation on the robot platform. We extend current results to the case of navigation in a cluttered environment. Furthermore we extend a well known technique from robotics, the Vector Field Histogram, to navigate using the ALV.