Ant-inspired navigation algorithm based on visual landmarks

Abstract

This paper describes a method for mobile robot navigation that is similar to the navigation mechanism of social insects. Unlike other bio-inspired methods that mimic certain morphological features of animals or separate natural mechanisms, the proposed approach is based on the phenomenology of the behaviour of some ant species during collective foraging. This method does not require a map, but allows a robot to memorize a path from the "nest" to the food resource based on visual landmarks, compass data (the skylight compass in insects) and the time component. The path is represented by a sequence of movements that takes the robot from one support landmark to another, as in ants. The remembered route allows the robot not only to return to the starting point, but also to repeat the route. The path built by the robot is not optimal, and the robot repeats this path approximately. However, the method does not require precise positioning and high accuracy of the sensors. The developed route description format is compact, because it only contains a sequence of directions to support landmarks, distances to them, and the amount of time the robot spent on the way from one landmark to another. A key feature of this solution is the implementation of route transmission from the scout to the forager via low-speed communication channels, for example, using the RC-5 infrared communication protocol. The method has been tested in simulations and on real robots in an indoor testing ground. In addition, due to specific architectural and technical solutions focused on modeling behaviour and the use of dimensionless parameters, the transition from simulation models to the management of technical objects (robots) can be made without the physical modeling stage, which makes development easier and cheaper.