Multi-agent Algorithm for Orientation and Navigation of Autonomous Robots in Mountainous Areas

Abstract

AbstractIn unstructured and dynamic environments, which include mountainous areas, interaction with a robot is not always possible. Due to the variability and unpredictability of the environment, it is also difficult to control robots. Most of the tasks that arise in the conditions of mountainous areas are solved through manual or semi-autonomous control. This approach provides reliability in dealing with unforeseen circumstances by combining human cognitive decision-making processes with the capabilities of a robot. However, its efficiency is low and requires significant human effort to remotely understand the data received from the sensors and control the robot. Therefore, the task of researching and developing intelligent solutions that use various sensor modalities and input data sources to calculate the localization of the robot in mountainous conditions is very relevant. The paper presents the structure of the sensory subsystem of an autonomous robot for working in mountainous areas. An algorithm for the formation of a multi-agent system of locative events and relations between objects for the construction of metric, topological and semantic maps is presented. An algorithm for processing data from the navigation subsystem of a mobile robot for working in mountains based on a multi-agent neurocognitive architecture has been developed. Thanks to the obtained multi-agent space around the robot, a hybrid map of the area and the trajectory of the robot movement are built within the framework of the multi-agent neurocognitive decision-making system used.KeywordsAutonomous robotMulti-agent systemsIntelligent systemsNavigation and orientation