THE INFORMATION AND NAVIGATION FIELD CONSTRUCTING SYSTEM FOR UGV AND UAV IN AN URBAN ENVIRONMENT

Abstract

The recent increasing demand of heterogeneous groups of robots (UAV and UGV) with increased
autonomy when conducting special operations in industrial and urban environments is
substantiated. The urgent task of forming, based on data from UAVs on-board computer vision
systems, an information and navigation field that ensures autonomous targeted safe UAVs and
UGVs movement in shielded areas of an urban environment is formulated. The formation of a
generalized geometric model of the external environment can be achieved by specifying a set of
target positions in terms of the working area, which the UAV must visit in a given sequence and
return to the starting point. In the process of visiting achievable target points, a generalized geometric
model of the external environment is formed and the current coordinates of the UAV are
determined. Methods and algorithms for constructing various models of the external environment
and solving navigation tasks are described, which ensure planning and executing of targeted safe
movement trajectories in real time according to on-board data, which is the basis of autonomous
control, including the heterogeneous robots’ groups control. Autonomous control systems for the
movement of UAVs and UGVs are based on methods and algorithms for identifying semantic objects
(supporting surface planes and vertical walls), which abound in urban environments, and
extreme navigation using 3D images (point clouds), obtained from lidar or depth cameras.
The results of experiments on the information and navigation fields creation and solving navigation
tasks based on on-board computer vision data in a real industrial-urban environment are
presented, confirming the effectiveness and practical value of the proposed methods and algorithms.
The use of a single information and navigation field, on the one hand, significantly increases
the autonomy of a group of robots due to the ability to independently plan actions when
performing complex operations, and on the other hand, increases the situational awareness of
robot operators by providing information about the working space in a convenient form