Collaborative Autonomous Navigation of Quadrotors in Unknown Outdoor Environments: An Active Visual SLAM Approach

Abstract

The development of an integrated path-planning and Simultaneous Localization and Mapping (ASLAM) system, specifically designed for the autonomous and real-time guidance of quadrotors navigating through unexplored outdoor environments helps to map the generation of unknown natural resources. To achieve this goal, a path-planning methodology that leverages system observability is exploited for a quadrotor. This path-planning method is underpinned by the eigenvalues of the Gramian matrix, which are used as a measure of system observability degree, to increase the precision of the quadrotor’s estimated position. In SLAM, high accuracy in the quadrotor’s state estimation improves the accuracy of the map landmarks position estimation. To enhance the accuracy and fortify system robustness, implementing a centralized distributed architecture within a group of three quadrotors is advocated. In this setup, the role of a central hub for information fusion from all agents and determining the most observable path for the entire group is assigned to the leader quadrotor. An assessment of the proposed path-planning method against a random path-planning approach within a single-agent architecture is conducted across various scenarios. This evaluation compares the Root Mean Square Error (RMSE) of the quadrotor’s state estimation. The results illustrate a notable improvement in accuracy. Furthermore, a comparison is conducted to assess the performance of the multi-agent architecture in contrast to the single-agent architecture using the proposed method. The simulation and experimental results confirm a better accuracy in all scenarios and highlight the increased robustness of the cooperative architecture, particularly in fault scenarios, compared to a single-agent architecture.