Search

Search Results

• VISUAL NAVIGATION OF UNMANNED AERIAL VEHICLES USING SEMANTIC TERRAIN DESCRIPTIONS

  N.V. Kim, N. V. Udalova, N. Е. Bodunkov, D.S. Girenko, N.А. Lyapin

  2025-04-27

  Abstract ▼

  The article addresses the problem of visual navigation for unmanned aerial vehicles (UAVs), which
  involves the automatic determination of the current position of the UAV (coordinates in the ground (local)
  coordinate system) based on the comparison and identification of descriptions of the current images (CI)
  received on board with reference descriptions stored in the form of a digital map in the memory of the
  UAV's onboard computer. The aim of this work is to improve the efficiency of visual navigation methods in
  terms of increasing computational performance, robustness, and accuracy of image identification algorithms
  in complex and changing observation conditions by using semantic descriptions of observed scenes.
  In this work, semantic descriptions are understood as descriptions that include classes of objects observed
  in the scene, their attributes, and relationships between them. The preparation of semantic descriptions of
  the map is carried out at the pre-flight preparation stage of the UAV using pre-trained neural networks for
  semantic segmentation. Semantic descriptions of the received CIs are generated on board the UAV. The
  use of neural network algorithms allows this process to be implemented in real-time for a wide range of
  observation conditions (different times of day and year). The use of semantic descriptions of the map and
  CI reduces computations compared to traditional pixel-by-pixel matching of raster images. Semantic descriptions
  are compared by matching object classes, their attributes, and relationships. The work presents
  a general algorithm for visual navigation, the main stages of the methodology for forming semantic descriptions,
  and the algorithm for comparing and identifying semantic descriptions of CIs and map descriptions.
  A hierarchical algorithm for comparing and identifying images based on the sequential application
  of semantic and raster descriptions of observed scenes is proposed. It is shown that the use of the procedure
  for comparing semantic descriptions of CIs and maps by the classes of objects present significantly
  reduces the computations necessary for image identification

• AN ONTOLOGICAL APPROACH TO THE CREATION OF ROBOTIC COMPLEXES WITH AN INCREASED DEGREE OF AUTONOMY

  S. М. Sokolov

  42-59

  2022-04-20

  Abstract ▼

  The aspects necessary for the implementation of robotic complexes with an increased degree of
  autonomy (RC with IDA) in practical work are considered. The distinctive features of such complexes,
  the needs of the corresponding intelligent information control systems (IIСS) are indicated. The requirement
  of situational awareness is highlighted and, as a consequence, the need for a diverse system
  of knowledge representation, means of perception of the external environment and comparison of operational
  information with models and a priori information about this environment. In addition, it is pointed
  out the need to automate the processes of creating RC with IDA, accessibility, and simplification of their
  use. In order to answer these questions, the paper proposes to use the concept and mechanisms of ontologies
  in relation to autonomous robotics. Examples of existing solutions in this area are given. In robotics,
  ontologies are used to define and conceptualize knowledge accepted by the community, using a
  formal description that is machine-readable, shared, and contains the flexibility to justify this knowledge
  in order to derive additional information. Ontologies are of considerable interest for multi-agent systems
  for organizing interaction between agents and with other systems in heterogeneous environments,
  the possibility of reuse and support for the development of new RCs. The author describes the construction
  of an ontology proposed by the author in such an applied field as information support for targeted
  movements of autonomous ground vehicles based on technical vision systems. All consideration is conducted
  in the configuration space of the information and control systems of the RC with IDA. This space
  allows you to aggregate a large number of different technologies used in the construction of RC. The
  embodiment of a particular system in this space corresponds to the "assembly point". The coordination
  of the forms of knowledge representation in the IICS is ensured by the consistent consideration of planes
  in this space. As a connecting link – a means for automated translation of descriptions of descriptive
  ontologies into descriptions of functional, machine-readable ontologies, the use of the language of information-
  motor actions and interpretive navigation commands is proposed. In conclusion, the shortterm
  prospects for the development of the described approach are considered, and wishes are expressed
  to the domestic community of roboticists.

• ORGANIZATION OF GOAL-DIRECTED MOVEMENTS OF VEHICLES USING VISUAL LANDMARKS

  S.M. Sokolov, N.D. Beklemishev, А. А. Boguslavsky

  2021-04-04

  Abstract ▼

  The report considers the solution of the navigation problem with the help of a technical vision
  system that determines the position of the mobile vehicle relative to the landmarks indicated
  in the surrounding space. Navigation by landmarks is the most objective criterion for the location
  of a mobile vehicle in the surrounding space. The method of measuring the parameters of the ratios
  that characterize the location of the mobile vehicle relative to the landmarks is almost independent
  of other navigation measurements. Data input for correcting coordinates and other motion
  parameters can be performed not continuously, but at some discrete, and, in general, quite
  rare moments of time. The general scheme of the solution is considered: from setting up, to receiving
  navigation information. The integration of the obtained data with data from other navigation
  tools is briefly described, and the key problems and parameters of the VS that affect the accuracy
  of the obtained results are analyzed. The key point in this method is the solution of a system of
  equations describing the position of robotic complexes relative to the specified landmarks. This
  system is solved by a modified Gauss-Newton method for a nonlinear redefined system of equations.
  By replacing the left side of each equation with its differential at the point of initial approximation,
  linearization is performed. The values of the unknowns in the redefined system of linear
  equations for which the sum of the squared residuals in the equations is minimal can be obtained
  either by the SVD (singular value decomposition) method or by using the system's symmetrization.
  At the same time, SVD is more resistant to the accumulation of computational error, but it is
  somewhat more demanding on computer resources and more difficult to implement. We used the
  symmetrization solution as a simpler one. The resulting system is solved by the square root
  (Cholesky) method. To detect landmarks in the VS, two types of VS modules are used – panoramic,
  based on a camera with a fish-eye lens, and stereo. The proposed method allows us to solve the
  problem of clarifying the parameters of motion by separate, sparse measurements of the proper
  position and speed relative to landmarks in the surrounding space. Independently and in combination
  with other navigation tools, the described approach provides high-precision determination of
  navigation parameters in various driving conditions. The results of field experiments with the
  model of the proposed system in motion under various conditions are described. The ways of improvement
  and development of the considered approach are discussed.

• ESTIMATION OF THE SPATIAL POSITION OF AN ON-BOARD CAMERA BY COMPARING AERIAL IMAGES AND SATELLITE IMAGE DATA

  А.Y. Budko , Т.А. Gaida , Z.А. Ponimash

  2026-02-27

  Abstract ▼

  The article describes a method for estimating the spatial position of an onboard camera of an aircraft. This method involves comparing aerial photographs and georeferenced remote sensing (RSS) data by using a neural network detector to detect stable spatiotemporal reference points in both datasets. This method then solves the well-known Perspective-n-Point (PnP) problem for estimating rotation and translation matrices that minimize the reprojection error based on the correspondences between 3D world points and 2D points of their projections onto the onboard camera matrix. This approach can be used to solve the pressing problem of aircraft localization in the absence of global navigation satellite system signals. Road intersections are selected as stable spatiotemporal reference points that are clearly visible in RSS data and aerial photographs. Other local semantic image patterns characteristic of a particular area may serve as an alternative. Since direct comparison of remote sensing and airborne images is difficult due to significant differences in shooting conditions, the use of robust landmark detectors based on artificial neural network (ANN) algorithms is proposed. To train the robust detector, a mixed dataset was created using satellite and airborne imagery. The mixed dataset was labeled using a 3D Gaussian function normalized to unity with a apex at the intersection center, the graph of which is projected onto a 2D mask of the training set. The parameters of the Gaussian function are calculated based on the radius of the circle enclosing the intersection. Using a normalized 3D Gaussian function with a apex at the geometric center of the intersection projection allows the network to predict the probability of each image pixel belonging to the intersection, with a maximum at the intersection center, which increases positioning accuracy due to more precise georeferencing of the landmark point in the global 3D dataset. A U-Net-type artificial neural network was trained as an intersection detector. A differentiable analog of the Dice metric was used as a training quality metric. AdamW, coupled with a CosineAnnealingLR cosine learning rate planner, was used as an optimizer. The final section of the paper presents the results of comparing satellite data and airborne imagery using the proposed method.