The article explores the possibilities of using hand gestures as a control interface for robotic systems in a collaborative workspace. The development of hand gesture control interfaces has become increasingly important in everyday life as well as professional contexts such as manufacturing processes. We present a system designed to facilitate collaboration between humans and robots in manufacturing processes that require frequent revisions of the robot path and that allows direct definition of the waypoints, which differentiates our system from the existing ones. We introduce a novel and intuitive approach to human–robot cooperation through the use of simple gestures. As part of a robotic workspace, a proposed interface was developed and implemented utilising three RGB-D sensors for monitoring the operator’s hand movements within the workspace. The system employs distributed data processing through multiple Jetson Nano units, with each unit processing data from a single camera. MediaPipe solution is utilised to localise the hand landmarks in the RGB image, enabling gesture recognition. We compare the conventional methods of defining robot trajectories with their developed gesture-based system through an experiment with 20 volunteers. The experiment involved verification of the system under realistic conditions in a real workspace closely resembling the intended industrial application. Data collected during the experiment included both objective and subjective parameters. The results indicate that the gesture-based interface enables users to define a given path objectively faster than conventional methods. We critically analyse the features and limitations of the developed system and suggest directions for future research. Overall, the experimental results indicate the usefulness of the developed system as it can speed up the definition of the robot’s path.

This study proposes a new approach to improving laser sensor data collection through optimised sensor settings. Specifically, it examines the influence of laser sensor configurations on laser scanning measurements obtained by using a laser line triangulation sensor for transparent and non-transparent plastics, as well as aluminium alloys. Distance data were acquired with a three-degree-of-freedom positioning device and the laser sensor under both manual and automatic settings. Measurements were performed at the sensor’s reference distance and across a wide range of positional configurations. The results of extensive experimental tests highlight optimal sensor configurations for various materials and sensor orientations relative to the scanned surface, including both in-plane and out-of-plane angles, to enhance the reliability and accuracy of distance data collection.

Voxelization into a fixed voxel grid is used for simplicity of processing and data format for applications such as dynamic re-planing of a collaborative robot path around a dynamic obstacle. In this case, it may be undesirable that conventional voxelization methods generate more volume, which results in a longer generated path than needed if the real shape of the obstacle was used instead of voxels. The approach proposed in this paper utilizes triangular normals of the mesh formed by neighboring points in a depth image to adjust the occupancy of a fixed voxel grid, ensuring the voxels more accurately represent the true volume of the scanned object. An experiment with real hardware was performed to compare the proposed voxelization method with the conventional fixed grid voxelization method. The collected point clouds are voxelized using both methods, and the resulting voxel maps are then passed to a simple path solver to find the length of the robot path around the obstacle. The results indicate that our proposed method achieves better results in terms of shorter avoidance paths, as our method provides a tighter representation of the obstacle. The average increase in the length of the alternative avoidance path compared to the theoretical shortest path around the object was 15.7 % for the conventional method and 8.6 % for the presented method. Overall, the experimental results indicate the usefulness of the developed system, as it can shorten the robot cycle time when avoiding obstacles.

This paper presents a quantitative analysis of the impact of inaccurate positioning (deviations in position and orientation) of a 3D triangulation on surface coverage. Using simulated scanning of 50 heterogeneous models, the ideal sensor position maximizing surface coverage was first identified for each object. Subsequently, the influence of spatial deviation and angular inaccuracy was systematically examined in this optimal configuration. The results show high sensitivity of the system, where even a small deviation in position (25 mm) leads to an average loss of coverage of 7.99 %, and in the case of geometrically complex objects with a high degree of self-occlusion, the degradation of data is statistically significantly higher, which was confirmed by statistical analysis using paired t-tests. Orientation analysis revealed that tilts around axes perpendicular to the viewing direction have a significantly more destructive impact than rotation around the viewing axis. The simulation results were validated by real measurements using a triangulation-based 3D scanner mounted on a robotic workstation, showing good agreement between simulated and real data. The proposed methodology and the obtained data provide a quantitative framework for defining scanner positioning tolerances in real-world industrial applications and for the design of robotic inspection cells.

Depth cameras have become an integral part of a wide range of applications, including industrial inspection, reverse engineering, human–robot collaboration, and biomechanical analysis. Their practical use, however, is limited by a narrow field of view and self-occlusions, which often prevent complete surface coverage of the scanned object. This paper introduces a methodology for systematically determining the optimal placement of a depth camera and an auxiliary planar mirror to maximize surface visibility. To this end, a simulation tool was developed that models camera rays, their reflections from the mirror, and their intersections with objects. The proposed method combines geometric visibility analysis with an optimization procedure that searches for configurations with maximum coverage. Validation was performed on a set of twenty objects ranging from simple geometric primitives to complex reference models. While a standalone camera achieved an average coverage of 51.68 %, the addition of a single planar mirror increased this value to 85.46 %, with complete coverage achieved for several objects. A complementary user study demonstrated that manually chosen configurations exhibit high variability and do not outperform the algorithmic design. The results confirm that the use of mirrors provides a cost-effective solution for overcoming occlusion-related limitations without resorting to multi-camera systems, and that the proposed methodology delivers consistent performance across diverse object geometries. The approach is particularly applicable in industrial inspection, robotic vision, and other domains where complete geometric information is essential.

Hand localization in cluttered industrial environments remains challenging due to variations in appearance and the gap between synthetic and real-world data. Domain randomization addresses this “reality gap” by intentionally introducing randomized and unrealistic visual features in simulated scenes, encouraging neural networks to focus on essential domain-invariant cues. In this study, we applied domain randomization to generate a synthetic Red-Green-Blue–Depth (RGB-D) dataset for training multimodal instance segmentation models, with the aim of achieving color-agnostic hand localization in complex industrial settings. We introduce a new synthetic dataset tailored to various hand detection tasks and provide ready-to-use pretrained instance segmentation models. To enhance robustness in unstructured environments, the proposed approach employs multimodal inputs that combine color and depth information. To evaluate the contribution of each modality, we analyzed the individual and combined effects of color and depth on model performance. All evaluated models were trained exclusively on the proposed synthetic dataset. Despite the absence of real-world training data, the results demonstrate that our models outperform corresponding models trained on existing state-of-the-art datasets, achieving higher Average Precision and Probability-Based Detection Quality.