Research

Open and realistic dataset, captured directly on active construction sites using 360° camera with IMU measurements and floor plan priors, to evaluate and compare SLAM systems in real industrial conditions.

A robotic system that optimizes food product quality using powdered cappuccino preparation as a case study, combining optimization algorithms and computer vision to find ideal parameters for the best foam and improve results via closed-loop visual feedback.

A dataset of 1,000 indoor spatial Q&A pairs based on Replica scenes, balanced using a GIS-inspired taxonomy. It also includes an evaluation system that grades free-form answers with a vision-language model, and a baseline method (RAG3D-Chat) that achieves 67% accuracy.

A Unity3D-based simulation framework for Autonomous Underwater Vehicles (AUVs), enabling safe and efficient software testing, neural network training, and performance evaluation across diverse underwater scenarios. The system simulates key effects (e.g., buoyancy, caustics), supports sensor/actuator communication, dataset collection, and environment randomization for robustness testing.

A comparative analysis of underwater robot design based on the PWr Diving Crew project, documenting the evolution from simplified ROVs to AUVs with improved perception and autonomous task execution. The paper summarizes key design changes across generations and draws conclusions on structural optimization from practical experience.

A low-cost laser triangulation–based 3D imaging device for AUVs/ROVs to enable underwater sensing in low-visibility conditions. The paper presents a simplified design approach and validates it through above- and underwater experiments measuring object dimensions and distances.

A UDP-based teleoperation system enabling real-time two-way communication between an Android device and ROS. The paper presents an Android app and ROS package for controlling a Turtlebot3, validated in both simulation and real-world tests, with modular open-source code provided.