Robotics projects and Physical AI
MobiLab combines academic research, student training, and validation in real environments. Our projects integrate perception, control, manipulation, autonomy, and remote operation for safer work in complex environments.
Physical AI research
Physical AI connects artificial intelligence to the physical world. At MobiLab, this means robots perceive environments, make decisions, and execute tasks with sensors, actuators, and control on real hardware. Related work includes multimodal perception, mapping, navigation, locomotion, manipulation, human-robot interaction, teleoperation, simulation, learning, and sim-to-real transfer.
Applications
Search, rescue, and emergency response
With the Volunteer Fire Department of Joinville, the laboratory investigates quadruped robot capabilities for fires, rescue, person localization, load traction, and nozzle control. Work includes 3D LiDAR, thermal and depth cameras, obstacle traversal, and remote decision support.
Autonomous infrastructure inspection
In collaboration with Joinville's infrastructure department, MobiLab explores autonomous inspection of urban tunnels and underground drainage using the Unitree B2W. The work addresses confined environments, long missions, weather resistance, and visual and spatial data collection.
Humanoid robotics
Humanoid platforms support research into locomotion, robotic hands, manipulation, human-robot interaction, machine learning, and autonomous systems in environments designed for people.
Mobile robotics, mapping, and simulation
Localization, mapping, sensor fusion, navigation, ROS 2, Gazebo, and MoveIt 2 connect simulation to hardware experiments. The laboratory also collaborates with the PERRO Lab and Thompson Rivers University in Canada on quadruped robotics, sensor fusion, navigation, infrastructure inspection, and emergency response.
Evidence and updates
Each project should document its team, partners, platform, methods, results, software versions, publications, datasets, limitations, and safety procedures. See MobiLab news and the MobiLab GitHub organization.