Search
Search Results
Found one item.
1 - 1 of 1 items
Digital Human Models (DHMs) are becoming an important tool in robotics, enabling reproducible and controlled validation of computer vision algorithms. Such algorithms include human detection, tracking, and following, which play a key role in autonomous navigation, trajectory planning, and safe robot-human interaction in shared spaces. However, standard simulators are typically limited in the realism of virtual actors, the diversity of their appearance, and the variability of their behavior. These limitations reduce the reliability of the results and make it harder to transfer algorithms to real-world robots. This paper presents a set of customizable DHMs for the Gazebo simulator. The developed models feature variability in clothing, anthropometric parameters, and appearance, and include a library of walking animations simulating various human movement scenarios, including changes in speed and direction. To evaluate the effectiveness of the proposed approach, virtual experiments were conducted in a realistic office environment using the TIAGo Base mobile robot with differential drive, equipped with a 2D lidar and a monocular camera. The efficiency of the user-following algorithm was evaluated in terms of average distance traveled and the number of target loss events (the number of false positives). The results showed that the proposed DHMs make it possible to reproduce complex perception conditions, including occlusions and dynamic obstacles, while providing scalable and systematic validation of algorithms. The obtained data confirm the significance of the developed DHMs for testing robotic systems oriented toward human interaction in conditions close to real-world settings